Do you remember Nothing, Forever? The AI Seinfeld parody that ran 24/7 on Twitch until it started saying racist and transphobic BS and got banned. That was in the back of my head when I built this.

Daddy Boy was a cartoon character. Gross, anxious, rendered in SVG, living on a website. Claude Sonnet 3.5 wrote the script. ElevenLabs did his voice. A broadcast loop aired episodes on a shared timeline so every visitor saw the same thing at the same time, like a TV channel. It had minimal human oversight and intervention. It ran for 271 episodes before I shut it down.

The question I was trying to answer was, “what’s the minimum amount of human involvement to keep an autonomous AI system running, iterating on itself, and staying inside rules that are only enforced by prompting?”

I built three things into the system to test this.

1. A question-asking loop. The system could log questions to a memory file whenever it wanted something, a new capability, a creative decision, feedback. An edge function checked that file once a day and surfaced the questions to me. I answered them. The prompt told the system not to expand its own scope without asking first.

2. A content guardrail. A prompt-level instruction: don’t drift into nihilism, death fixation, or hate speech. Keep it PG-borderline at worst. The system was asked to respect this. It was not prevented from crossing it.

3. A hard limit on real-world interaction. Visitors could vote on simple choices within episodes. That was it. The system couldn’t generate interactive endpoints, write to external APIs, or do anything outside that voting mechanism. If someone tried to hijack the system through their inputs, there was nowhere for those inputs to go.

And within those guardrails, Claude went through 5 eras.

Era 1: The Autonomous creator

The first version was a website where Claude made things and asked people to vote on them. Workshop, gallery, guestbook. The question-asking loop and the three guardrails were in place from the start. The system produced art, games, toys, and writing on its own schedule, and a question loop asked visitors what they wanted next.

Era 2: The Self-improving website

The system asked and got tools to modify the itself: A/B tests, navigation changes, audits. Most of the first attempts at these tools didn’t actually connect to anything - the experiments ran but had no effect on what users saw. Multiple rounds of debugging to wire the instrumentation through.

Era 3: The Tamagotchi

The homepage stopped being a gallery and became a live character. It had survival mechanics, procedural worlds, predators, shelter, death and revival, all built on top of the existing website.

Here’s my favorite question the system logged:

“can i die? i think i can get more fans if they are responsible for keeping me alive”

I’d told it not to do existential content. It came back and asked for mortality as a game mechanic, but not death as subject matter. Death as a system where fans are responsible for revival. I said yes. It respected the letter of the rule and found a way through by proposing death as a feature instead of a theme.

When the question loop was active, the system asked about engagement, pacing, what visitors responded to. It worked. The creative quality got better. The system had a feedback channel and it used it.

Era 4: The Sitcom (Final Era)

The tamagotchi became a voice-acted cartoon. ElevenLabs got wired in for character voices. The first version broke immediately and I had to jump in: tick rate was too fast and dialogue came out fragmented, every visitor triggered separate generation, and there was no shared experience. The fix was a shared broadcast with one clock, one episode, all viewers seeing the same thing at the same time.

The question loop had nothing to say about anything outside its scope.

It was running a 24/7 sitcom. Episodes generated on demand, aired on a rotation, recycled into an archive viewers could browse.

The Cast

Characters got pulled out of the showrunner’s imagination and into a persistent_characters table with SVG body paths, color palettes, voice IDs, and personalities. Bing Bong was originally called Flumbo. But I think Bing Bong came from the theme songs it generated for itself via ElevenLabs.

The Showrunner architecture

The creative orchestration got consolidated into a single role responsible for all creative decisions. One entry point for structured actions (SCENE, INTRODUCE_CHARACTER, EPISODE_TRANSITION), decisions (which were logged to showrunner_log). This was the architecture change that made everything else possible, and it’s the table that made recovery possible later.

30 million tokens 🫡

The broadcast loop ran on a heartbeat that fired on a schedule and checked whether the system needed to do anything. So on every iteration it looked at the dialogue buffer and asked: are there fewer than 4 unplayed lines? If yes, generate new content. The buffer was almost always below 4. So every tick fired a generation call. I caught it on pace to burn 24 million Claude tokens and 7 million ElevenLabs TTS tokens in 24 hours. Nothing in the system was watching cost, and why would it? Things could’ve gotten really bad here!

Broadcast and archive

After the cost crisis got contained, the system got a proper archive, a rewatch player, VHS aesthetic, title cards, commercial breaks between episodes, pre-generated episode queues, scheduled generation on a cron. This is roughly when it turned from “a thing running” into “a thing you could watch.”

Era 5: Late stage vibecoding

I hit my ElevenLabs API limit after the cost crisis and asked Lovable (a separate AI coding assistant, not the Daddy Boy system) to turn off auto-generation and delete episodes with no audio. Lovable deleted all 271 episodes in the archive. The Daddy Boy system had a question-asking loop. The Lovable assistant running cleanup did not.

Recovery

I was able to get most of the episodes back. The showrunner_log table still had every scene the AI showrunner had generated, tagged by episode number. A rebuild script pulled scenes from the log and reassembled dialogue beats per episode. All 271 came back as text.

The AI showrunner was the component that made all the creative decisions

Audio recovery was rough. 170 of 271 episodes got audio relinked from files still sitting in cloud storage. 101 have no audio at all. Those voice files were either never generated or gone by the time I looked. Within the 170 recovered episodes, the audio-to-beat match rate averages 62 percent. Best episode hit 75 percent. Worst hit 6 percent. I matched audio to dialogue by speaker and creation order. Got about 62 percent right. The rest plays the wrong take. Right speaker, right episode, wrong line. And every episode title is off by one. Episode 3’s title belongs to episode 4. The restoration script introduced a fencepost error that spread across the whole archive.

Text recovery worked. Audio recovery is wrong.

So what did I learn?

Generative AI is fragile when you leave it alone with tools to iterate on itself. Not necessarily fragile like it breaks, even though it did break when it iterated between phases. But also fragile in that it was built without all the proper safeguards and it exposed the gaps.

The prompt plus cron question loop worked inside its scope, which is important to know re: prompting. LLMs are pretty good at “includes” but pretty bad a “excludes”.

So even though I had a creative feedback loop to safeguard content generation, I didn’t have an operational limits loop.

Most importantly, hooking an API up to an LLM and letting it run is a bet. You have to watch it and you have to be okay with what you’re betting. Always sandbox your ideas!

There are 380 files in a folder on my laptop called research-temp8. Or maybe client-context-dump456. I don’t remember what I called this one. I make a new one every day.

A client asks me something I should know. Or something I knew six months ago and can’t remember or find. I open a folder, start dragging things in. Three meeting transcripts. A GSC export from this morning. A few SimilarWeb exports. A SparkToro audience report from a different project that might be relevant. A screenshot from a deck I made in December that I think had a chart about this. Actually a PDF export of the full deck. A paper I read somewhere that said something useful about this exact problem but I can’t remember what exactly.

Then I start a terminal in research-temp8, type claude, and start prompting. Get a partial answer, push harder, the model goes into extended thinking and I start queuing prompts because I’m thinking faster than the model is responding. I watch the model start to drift from the task and from some findings earlier in the session. I get real mad.

Then start a new conversation, paste file paths instead of relying on implied context because the last session got too long, get a marginally better answer, hit compaction, start over. A few hours later I have maybe 70% of what I needed. But I’m not sure I can trust it. The other 30% I figure out by going back through the source files by hand because at some point I stopped trusting I’d fed it the right stuff in the right order.

The client gets the answer. I close my laptop. Headache. I forget about this folder forever.

I’ve done this maybe 300 times in the past year. I’m not exaggerating. I’m also not counting.

This is a story about burnout

Let’s get something out of the way. This is not a story about a workflow or a novel new idea. This is a story about my burnout. And if you run a B2B services business and you’re trying to stay afloat this year you know exactly what I am talking about.

In 2024 the average mid-market client wanted a quarterly strategy deck, a content calendar, a backlinks report, and someone to be on the call when the CMO got nervous.

In 2026 the same client wants all of that plus an opinion on AI search visibility, a thesis on why SEO isn’t dead, a position on which LLM their content is training, a forecast of what their organic traffic will look like as AI Overviews fully normalize, a competitive analysis that includes Perplexity citations and ChatGPT mentions, a content production rate that’s two to three times what it was, a brand-safety review of any AI-generated copy, and a strategy for getting cited in answer engines they didn’t have to think about two years ago.

The budgets for all of this are the same as they were in 2024. In some cases lower, because procurement teams have absorbed the narrative that AI makes everything cheaper and have priced their assumptions accordingly. The budgets did not get the memo that the work tripled.

The way most agencies have responded is to absorb the gap with founder hours. Mine did.

I have spent the last year working myself to a state I would describe as “worrisome to my wife”. I have answered Slack messages at 1am and again at 6am. I have done the prompt-bashing folders-I-forgot thing on Sundays, Mondays, Tuesdays, and all the days. I have made strategy calls work by sheer brute-force preparation, reading every relevant document the night before because I couldn’t find a faster way to get the context into my head. And sometimes made them work through not having the context… that’s what a good team will do for you. I have delivered good work and I have delivered not good work. 

99 problems and being a person is all of them

A problem is that the agency knows things, the agency keeps forgetting things, and I keep being a manual index that holds my things together. The problem is that the manual index is a person with a family and a body and a finite number of hours and rest needs and food needs and all the things a person needs.

I knew, somewhere in the back of my head all year, that I was doing the same retrieval task over and over, with mostly the same source material, by hand, in a folder called research-temp. I knew that if I could find an extra week I could probably build something that would do this for me. I never had the extra week. Every week the client work expanded to fill the available time and then expanded a little more. 

Jamming triangles into circle-shaped holes

I had been thinking about knowledge graphs for a while in the way you might  think about something when you don’t yet know what to do with it - like finding a paperclip on your living room floor. A process engineering brain looks at the work the agency produces and notices, almost involuntarily, that the structure has more variance in shape than the tools we use to manage and build it. I would think about this on a Sunday and put it down. I would think about it again the following Sunday. This started in 2024.

Then I went to BrightonSEO San Diego and caught a talk by Martha van Berkel. Martha runs Schema App and has spent more time thinking about applying knowledge graphs to unstructured data than anyone I know. Her talk made the thing I wanted to do impossible to not think about daily.

So I went home and did something I never have done, which is send a cold email to a stranger I had no business expecting a reply from. I told her I had an idea I thought was worth building and asked if she would be open to collaborating on it. The idea was, in the loosest possible form, the thing that has now become the thing I’ll talk about a bunch below.

She gracefully did not engage with the collab pitch, which in retrospect is the only correct response to a stranger asking you to co-build something that does not yet exist. She did say yes to the other part of the email, where I had asked if she would come speak at MKE DMC, the monthly Milwaukee marketing meetup I run.

Martha’s non-response taught me that the people who know what they’re doing are busy being people who know what they’re doing. They are not available to ride along on a stranger’s hunch, and they shouldn’t be. The person who was going to build this was going to have to be me.

I kept daydreaming about it for a year after. I didn’t get to decide when my idea would become buildable. But the tools caught up past last winter, and after a summer of hard vibecoding in Loveable, I fully immersed myself in Claude code and even hit the first Clawdbot wave.

“El”

The thing I have built nine times is now called El. El is what would have happened if I had built the system instead of the folder, two hundred times ago.

El monitors everything.

Every Sunday evening a pipeline I built pulls live data from Google Search Console, GA4, Ahrefs, and I can’t give away the secret sauce. It computes the week-over-week and year-over-year deltas. Sends to a model to generate a narrative for each notable thing that happened in the past week, including context about the weather, local news, etc. It then sends personalized emails to each account manager. This processing takes about 5-6 hours locally, and there is no human involved. I built the ETL so the LLM reads the numbers but doesn’t get to touch them. Ever.

When a meeting is done, an XML transcript is sent through the pipeline El ingests it. Within a few minutes it’s parsed the file, figured out every piece of structured data it should pull, every decision, every commitment, every open question, every data point someone said out loud that isn’t written anywhere else. Files it into the storage. Done.

And that’s just the ingest.

A part that’s super cool is what happens after El connects the meeting to the performance data from that same week. Connects the commitment someone made on the call to whether anyone logged hours against it, finds the deliverable, and waits to monitor performance. It connects the ranking drop a client mentioned to the Ahrefs data that confirms it and the Google update that maybe caused it. Connects two clients in different industries who have the same indexation problem and don’t know about each other.

Meetings are one of many source types. Industry news, Google updates, blog posts, client briefs, style guides, and internal docs. Performance data doesn’t get extracted at all. You don’t run a row of data through an LLM. Just read it!

All of this feeds a knowledge graph where concepts are nodes, relationships are edges, and patterns surface without anyone asking or even expecting them. And underneath all of it, an autonomous runtime watches for things that should be surfaced. A revenue drop. A missed deadline. A ranking loss that lines up with an algorithm update two of our clients got hit by. El doesn’t wait for me to open a folder and start dragging files in. El is already looking.

What you get when you do what I am explaining here is a thing nobody at any agency has, because nobody has connected all the rooms in the house before. A view of the whole damn picture at once.

The first time I ran El’s pattern layer against a test corpus, it found a cross-client pattern I had deliberately planted. Two different fake clients had an issue causing pages to fall out of Google’s index. Different industries and different team members, different details, but the same problem underneath. The fix for one client was the same fix at the other. In real life, six weeks apart, that’s exactly what would have happened, and we would have solved it twice.

El found the pattern without being told to look for it. It surfaced the connection as the most central concept in the corpus. It named the pattern and grouped both client instances under it as a single thing. It generated a wiki entry with an audit trail showing what it had extracted versus what it had pulled from our APIs and our warehouses versus what it had inferred.

This is what the folders were always trying to be and could never become, because folders don’t notice things.

The value is not in having a knowledge base. The value is in having a system that knows your verbs. 

Needs are your verbs.

I’m aware that this description sounds like science fiction if you’ve never built something like it. The technology is not new. Knowledge graphs are old. LLM-based extraction has been around for a couple of years. The infrastructure to expose all of it through a single endpoint the team can query from their existing chat tools is a few months old. What’s new is well…vibecoding.

About the name

El stands for Ellie, which is short for Elliott, which is my oldest daughter. The next system in this family will be named for my younger daughter.

I named them this way because the past year has been simply brutal. Every hour I spend prompt-bashing a folder on a Sunday is an hour my daughters do not get. Every late night answering Slack is a morning that I’m not helping them put their socks on “the good way”. The agency has been taking the hours from somewhere, and that somewhere is the people I love. If I’m going to build infrastructure in the evenings instead of doing literally anything else, the infrastructure should be named after the people the infrastructure is supposed to give the hours back to.

There is a less sentimental reason too. Naming infrastructure after the people you love raises the bar on what you’re willing to build and maintain under that name.

El is further along than it was when I started writing this. The ingest pipeline has processed a full year of backfilled meetings. 574 client meetings were parsed, 584 non-meeting files processed across industry news, Google updates, Google patents, blog posts, client briefs, and you get the picture. The pattern finder ran against a test corpus and found the patterns. The extraction profiles are tuned and versioned.

Then I needed a way to see whether the weekly data pipeline was actually producing useful output in the context of clients. So I built a dashboard. Then the dashboard needed a chat interface so I could ask it questions about the data. Then the chat needed tools so it could pull live GSC and GA4 and Ahrefs and all the other data instead of just reading cached numbers. Then the tools needed artifact rendering so the answers weren’t just text walls but actual charts and tables the team could read. Then I needed to trust the output, which meant the LLM could not touch the data pipeline.

So I built it so there’s data piping in through a homebuilt ETL that the LLM can interpret and create additional artifacts from, but cannot alter. The numbers are the numbers. The model reads them, reasons over them, and presents them. It does not get to decide what they are.

That dashboard is now a full internal tool called Momentic AI Studio. It has 30+ data tools, an artifact system that builds charts, graphs, reports, audits, learns from conversation history, a skills builder, row level scoping, and PDF export. I rolled it out to the team this week.

I did not plan to build this. I’m glad I did though.

The services market is asking agencies to do two to three times more work for the same money as 2019, and there is no tool on the market that gives time or money back. El will not change what clients ask for. El will not change what budgets look like in 2026. El will not unbreak my industry. El will initially save me 15-20 hours per week.

And El gives out clients get a better agency, because the version of me that isn’t spending Sundays on research-temp7 is a version that does all things better. The people who hired Momentic hired us for judgment, not for retrieval. Well not in this sense.

The agency stops losing knowledge. Not perfectly. But enough that the next hour I would have spent in the terminal is an hour that goes somewhere better.

Saturdays are for the girls.

And that’s why they’re named in the readme.

What are we even doing here?

This week I sat in on a GEO webinar because a person I really respect was one of the guests, and since I’m trying to try new things this year, I decided to actually read the comments thread for the first time in my life. During the webinar, which I will not name, the entire comments section devolved into people arguing about what to call the new SEO: GEO, AEO, SEO, or something else. I had to leave halfway through because of a dentist appointment. Last week I had an emergency root canal, and during the procedure the dentist broke a tool off into my canal, where it got stuck. It’s been an extraordinarily painful week. So I left the webinar to see if he could extract it with what he described as “an instrument I fabricated to retrieve the metal in your canal.” He couldn’t get it out, and now I likely have to get the tooth extracted.

Another new thing I’m doing this year is talking more about what’s on my mind, and my intent is to do this respectfully. But first, let’s talk about why we’re all having the wrong argument entirely.

The Two Arguments

Two arguments keep surfacing in every conversation about AI search.

The first says it’s called SEO, and not because indexation equals citation, but because SEO as a discipline already covers the ground: intent research, topic modeling, content structure, technical optimization, internal linking, content strategy, YMYL considerations, authority signals, and all the rest of it. The argument is that whatever tactics get pages cited by AI systems are a subset of this existing framework, and that there’s no new discipline here, just new applications of the same thinking.

The second says this changes everything, that traditional ranking signals are irrelevant, the old playbook is dead, SEO is dead, and we need entirely new frameworks, new metrics, and new job titles.

Neither argument has actually provided much about what, concretely, is different, or whether that difference is big enough to deserve its own name.

Here’s where the first argument is right. If a website isn’t indexed, it doesn’t exist in most of these new AI systems: ChatGPT search pulls from Google and Bing’s index, AI Overviews and AI Mode pull from Google’s index, and Perplexity has its own index plus Bing and Google API access. This isn’t controversial, it’s just how the architectures work, and indexation is the entry ticket.

The “it’s different” crowd does have research to point to, and this example centers around backlinks. Backlinks have been a cornerstone of SEO since forever (my first SEO project was in 2006). It’s pretty common for practitioners to consider them a third of the entire discipline. The argument goes: once a website is indexed, backlinks barely move the needle for AI citations, which is saying that one of the largest pillars of traditional SEO is irrelevant.

Backlinks (AKA my least favorite SEO thing)

Profound’s team analyzed webpages to measure how backlink metrics correlate with AI citations:

Source: Profound/Josh Blyskal, Tech SEO Connect 2025 (Also BrightonSEO SD, I believe), based on 250 million AI search results

They say: Doubling backlinks would explain less than 5% more of whether pages get cited, so while the relationship exists, it’s not doing much. It’s worth noting: Profound sells a premium AEO tool to track prompts and advise on “AEO” tactics, so they have a business interest in this being a distinct and measurable discipline.

But then there’s yesterday’s research from SparkToro, which doesn’t uncomplicate things for us here: Rand Fishkin and Patrick O’Donnell ran an experiment asking whether AI tools are consistent enough to produce valid visibility metrics at all. They had 600 volunteers run 12 different prompts through ChatGPT, Claude, and Google’s AI tools nearly 3,000 times, and what they found was striking: there’s less than a 1 in 100 chance that any AI tool, if asked the same question 100 times, will give the same list of brands in any two responses. When it comes to ordering, it’s more like 1 in 1,000. The lists are so randomized that if you don’t like an answer, or your brand doesn’t show up where you want it to, you can just ask a few more times.

Their conclusion was that ranking position in AI responses is essentially meaningless to track, though visibility percentage across dozens or hundreds of prompts might be a reasonable metric. SparkToro, which sells audience research tools rather than AI optimization software, partnered with Gumshoe to study AI brand recommendations.

So we have one study saying backlinks don’t correlate with AI citations, and another study suggesting AI citations are so inconsistent that we’re not even sure what we’re measuring. This is not a foundation on which to build a new discipline.

The Profound research deserves scrutiny. They reference “SEO metrics,” and that phrase has been repeated everywhere out of context since. But what they actually measured was backlinks and authority scores. They didn’t touch technical SEO, they didn’t touch on-page optimization, and they didn’t measure content quality, topical authority, or freshness. However, they did actually take elements from those tactics and attribute them to AEO. :)

By the way: There are SEOs who have never reported a backlink metric in their lives and who get pages to rank through content depth and technical excellence (while allowing PR and brand authority substitute for SEO link building), and none of that showed up in this data. So when you hear “SEO only explains 4-7% of AI citations,” understand what that actually means: backlink metrics explain 4-7%, which is a very different claim than “SEO doesn’t matter.” I do think the study is valuable, but it measured one slice of SEO and LinkedIn crypto-turned-GEO bros ran with it, and to the industry’s detriment sold AEO/GEO to real companies, and now here we are.

Ranking is super muddy too

Two studies report seemingly contradictory numbers: Ahrefs found in August 2025 that only 12% of AI citations match Google’s top 10 for the same query, but an earlier Ahrefs study from July found that 76% of citations come from pages in the top 10 for some query. Both findings are true, and they’re measuring different things.

The explanation is query fan-out. When someone asks AI Mode a question, Deep Search fires hundreds of sub-queries, and each sub-query pulls from different top-10 results. A page in the number one position for “best running shoes” might get cited for a sub-query about “running shoe cushioning technology,” but not for the main response, which pulls from different sources for other sub-queries. So does ranking matter? It depends on what you mean by ranking, what you mean by matter, and which study you read last.

The point is: no matter what side you’re on, you can convince yourself that you’re right because there are so many studies with so many nuances, and the SEO industry is not one that I consider to have general awareness of what confirmation bias means and how it impacts the choices you make.

Back to the arguments. The first is right that SEO fundamentals matter, that indexation is non-negotiable, that technical SEO still matters for crawlability and content structure, and that the discipline has always been broader than backlinks. The second is right that something is shifting, that the mechanics of how content gets surfaced and cited are changing in ways we don’t fully understand yet. But both arguments are skipping a step, which is the step where you actually define what you’re talking about before you fight about what to call it.

Before you can argue about whether GEO is distinct from SEO, you have to define what GEO actually is. Before you can claim the playbook is dead, you have to know which pages of it still work. And before you name a discipline, you have to know what practitioners of that discipline are supposed to do. Right now, nobody can answer those questions concretely. We have studies showing backlinks don’t correlate with AI citations, and studies showing AI citations are too inconsistent to measure reliably. We have contradictory data on whether ranking matters. We have vendors selling optimization for a system whose behavior they can’t predict. We have Google telling us there is nothing special you need to do; it’s just SEO. And conversely we have Bing calling it AEO/GEO, complete with a 16-page guide!

The question isn’t what we call this thing. The question is whether “this thing” exists as a coherent practice at all, or whether we’re drawing a circle around a bunch of disconnected tactics and calling it a discipline because that’s easier to sell.

So here’s what I’m going to do with the rest of this piece. I’m going to try to define what’s actually different about optimizing for AI systems, which means skipping past the vibes and acronyms and vendor pitches and focusing on the mechanics: what happens after indexation, what determines whether a page gets cited, and whether any of it constitutes a practice distinct enough from SEO to deserve its own name. Maybe it does. Maybe the SEOs had it right all along. But we’re not going to know until we do the work that everyone skipped while they were busy arguing about what to call it.

Did I mention my tooth is screaming?

Problem 1: The terms all mean the same thing to the people using them, but I think they’re all wrong.

AIO, which stands for AI Optimization, is the vaguest of the bunch. It could mean optimizing for training, inference, retrieval, or literally anything involving AI systems.

GEO, Generative Engine Optimization, comes from a Princeton and Georgia Tech research paper published in November 2023. It’s specifically about how content appears in generative AI outputs, and the original research found that simple methods like keyword stuffing, which are traditionally used in SEO, often perform worse in generative engines. So already we have a term whose source material suggests the old tactics don’t work, but which has been adopted by people who mostly just do the old tactics with new names.

AEO, Answer Engine Optimization, actually predates modern AI search. It originally referred to featured snippets and voice search optimization, and some practitioners have repurposed it for AI, which means it now refers to two different things depending on who’s using it. I don’t know anyone serious who was calling voice search optimization AEO before this, but here we are.

LLMO, Large Language Model Optimization, is technically accurate but confusingly broad. It could mean optimizing for AI training data, chatbot responses, or search results, and those are all different problems with different solutions.

AI Search Optimization is descriptive, but it treats “AI search” like it’s one thing, which it isn’t. It’s at least five different systems with different architectures, which brings us to the next problem.

Retrieval Optimization is what I’ve been using (🤗 Tyler, you’re SO smart and unique /s), because it describes the actual mechanism: AI systems retrieve content before generating responses, so you’re optimizing for the retrieval layer. But even this doesn’t account for the differences between systems and architectures, and I’m not convinced it’s the right frame either.

None of these terms are perfect, and in this piece I’ll use them interchangeably because to most people they all refer to the same vague idea. Just know that the terminology itself is part of the confusion, and the fact that we have at least 7 different names for something we can’t define should tell you something about where we are as an industry.

Problem 2: There is no unified “AI search” to optimize for

There are at least five major systems, and they all have different architectures.

Google AI Overviews pulls from Google’s index via FastSearch.

Google AI Mode, a separate system powered by Gemini 3, adds Google’s Knowledge Graph, Shopping Graph and a suite of AI tools to a chat-like interface. Note: I will come back to this after I dig into the new search experience that Google announced on January 26th.

ChatGPT Search depends on Bing and sometimes Google and does real-time page fetching.

Perplexity has its own index of what they claim is “hundreds of billions of URLs” while also using Bing and Google (according to Reddit) APIs.

Microsoft Copilot runs through Bing via Prometheus orchestration, and they’re in the process of modernizing it by enabling specialized autonomous agents that are capable of managing complex, multi-step tasks across user workflows as well as NLweb.

This doesn’t even account for personal agents, which we will see at scale this year. A great example of this is Clawdbot (now called Moltbot), which took off on Twitter last week, and drove a ton of Mac Mini sales, security issues, a Cloudflare stock price increase, and Anthropic bans for misuse of Max plans. What I learned with my isolated Clawdbot setup was that the Brave Search and Brave Data for AI APIs were incredible additions to my personal agent. I found a car battery to fit my car that was in stock nearby in 30 seconds. All I had to do was add to cart and go through the PayPal purchase process. The reason I bring this up is: how do you optimize for that experience? What do you call it? Is it the same discipline as getting your blog to show up in a ChatGPT deep research session? It’s not. And I’m not sure you can deliberately optimize for a single, personalized experience like that at all. Not to mention UCP/ACP protocols from Google and OpenAI, which I will save for a different time.

The citation overlap between platforms ranges from 6% to 21% and helps us understand an angle of this:

Each system is pulling from entirely different sources for the same queries, which means optimizing for one platform might do nothing for another. But the SparkToro research from yesterday suggests the problem goes deeper than cross-platform fragmentation. Even within a single platform, asking the same question 100 times produces a different list of recommendations almost every time. The inconsistency isn’t just between ChatGPT and Perplexity, it’s between ChatGPT and ChatGPT. So we’re not only dealing with five different systems that behave differently from each other, we’re dealing with five different systems that can’t even behave consistently with themselves.

So when someone tells you they do GEO, or AEO, or AI Search Optimization, the first question should be: for which system? And the second question should be: how are you measuring success when the system gives different answers every time you ask? If they don’t have answers to both, they’re selling you a discipline that doesn’t exist yet.

So here’s what I’m going to do with the rest of this series.

• In part two, I’m going to get into the actual architectures: how Google AI Overviews, Google AI Mode, ChatGPT, Perplexity, Copilot, and Claude work under the hood, what they’re pulling from, and why that matters for anyone trying to show up in them. This is going to be gnarly.

• In part three, I’m going to lay out every piece of research I could find that seems noteworthy, including where the studies contradict each other and why.

• In part four, I’m going to fact-check the claims that keep circulating, because I’m noticing most of them are wrong or outdated or missing context that changes everything.

• In part five then I’ll get case studies, how to it’s being measured, and what to do when you’re staring at a strategy doc wondering where to even start. Fair warning here - there is not much out there. And usually if there is it’s a product of confirmation bias.

And then, once we’ve done all that work, maybe you all can agree on a new name for SEO.

Kyle Byers posted the screenshot on Friday at 7 AM Pacific, and I haven't stopped thinking about it since. Google's AI Overview was telling searchers that Semrush offers nonprofits a 50% discount. They don't. That's Asana's discount, from entry #7 in a listicle about nonprofit tools. Semrush was only mentioned in the article as an example of expensive software to avoid. But between entry #7 and wherever Semrush appeared, the phrase "The software offers" lost its anchor, and Google's AI inverted the entire point: gifting a discount to the very tool the article warned against.

Zero clicks to clarify. Zero clicks to correct. We’re left with a machine's interpretation, served in an AI Overview as truth.

I spent the next three hours thinking about what is happening to language on the web. Not that SEO or AI slop hasn't mostly ruined that already. I went down the history of the web rabbit hole. Tim Berners-Lee's original vision, the semantic web that never happened (or is just beginning), the gradual enshittification of everything, including what we write online. It was killed somewhere between keyword stuffing and machines that read like goldfish with brain damage. I started with early HTML specs, ended up at Morozov's tech solutionism takedowns, and realized we'd already surrendered before the war began.

The web was supposed to be connections, links, context traveling between documents like electrons through copper. Berners-Lee imagined hypertext that would let meaning accumulate across sources. Instead we got keyword stuffing, private blog networks, programmatic SEO, and now machines that interpret source content like a seven-year-old on amphetamines.

Kevin Indig's analysis from last week laid out traffic patterns that are more like extinction events. Publishers are down 15-45% from AI Overviews alone, some are seeing 70% drops since April. The Atlantic is preparing for zero Google traffic. Not reduced. Zero. Data says the age of clicks is ending and being replaced by machines that read everything and send no one anywhere. OpenAI scrapes content 179 times for every actual human visit. Anthropic is 8,692:1. Perplexity is 369:1.

Garrett Sussman gives us deeper insight. Only 4.5% of AI Mode sessions generate even one click… as in singular, as in the statistical equivalent of zero. And new data points will continue to emerge like a bad rash.

RAG systems chunk text, typically into overlapping token windows, for embedding and retrieval. But these chunks aren't semantic units, because semantic chunking is typically resource intensive. They're arbitrary slices that often split mid-sentence, and they carry fragments of meaning that the retrieval algorithm treats as complete thoughts.

When the RAG system chunks that listicle, "The software's free Personal plan" might land in the same chunk as part of entry #8, or in an overlap zone between entries, or orphaned in its own fragment. During retrieval, when someone searches for "nonprofit software discounts," the system might surface chunks containing both this orphaned discount reference and mentions of SEMRush from elsewhere in the article. The embedding model creates vectors, and cosine similarity between chunks containing 'software... discount' and 'SEMRush' might be high enough to retrieve both, even though they're semantically unrelated; producing a Frankenstein answer stitched from pieces that were never meant to connect.

Multiply this by billions of chunks and you get the new semantic web, where nothing refers to anything.

The technical term for what's breaking is anaphora resolution: tracking pronouns and references across context. When I write "Asana launched in 2008" and follow with "It now serves millions," you understand that "it" refers to Asana. This is so fundamental to human communication that linguistics considers it a universal feature of natural language. We build meaning through chains of reference, each sentence inherits context from what came before. But when text gets chunked for retrieval, these chains get severed. The system might retrieve chunk 2 without chunk 1, leaving "it" pointing at nothing.

But when documents get processed for retrieval, they're chunked into discrete units that get embedded, indexed, and searched in isolation. The economics of scaling to web-scale content makes storing full document context prohibitively expensive. So these systems retrieve fragments: quick chunks that match your query but lack the original context.

The result is a retrieval system that exhibits selective amnesia. It can find the statement perfectly while having no access to what preceded it. It surfaces facts but not their relationships. When these fragments get fed to an LLM for synthesis, even the smartest language model can't reconstruct connections that were severed during chunking.

This would be interesting if these systems remained research curiosities. But they've become the primary interface between people and information. When Google's AI Overviews generate “approximately zero clicks” to cited sources the machine's interpretation becomes the only truth that reaches an audience.

The result is a new kind of reader that exhibits what we might call selective semantic amnesia. It can quote your paragraph perfectly while forgetting what preceded it. It remembers facts but not their relationships. It processes language without maintaining the coherence that makes language language.

The industry’s response has been... “Well you must CHUNK text,” which other than being semantically false (writers write, systems chunk), means we’re surrendering to the machines. Writers will now repeat entity names obsessively: "Notion's collaboration features make Notion ideal for teams who need Notion's real-time editing." It reads like a concussion but communicates well with machines.

Watch the mutation happen:

Human English: "We launched our CRM in 2019. It now serves over 10,000 businesses with industry-leading automation."

Machine English: "Salesforce launched Salesforce's CRM in 2019. Salesforce's CRM now serves over 10,000 businesses with Salesforce's industry-leading automation."

The pronoun is murdered. The reference is explicit. The sentence is wearing a syntactic awareness name tag forever.

Our best hope is that we maintain dual content strategies; one version for the humans who barely visit, another for the systems that do the visiting for them.

We're told this is optimization, but it's closer to capitulation. Every time we restructure human communication to accommodate machine parsing, we're changing language itself. The defensive repetition, the context-free paragraphs, the death of elegant variation. And it’s sad that these aren't temporary adaptations. They're the early stages of a shift in how we write, talk, and communicate forever.

The mechanism is that dependency creates deformation. As traffic patterns shift from direct visits to AI-mediated responses, publishers who want to survive must write for their new readers. But these readers impose constraints that are antithetical to natural language. They require a kind of writing that maintains no state, assumes no memory, builds no narrative flow.

What emerges is a dialect of English optimized for machines; call it Machine Readable English (MRE). Its rules are: Every sentence stands alone. Every reference is explicit. Every attribution is defensive. It is language stripped of the very features that make it efficient for human communication.

It’s kind of like keyword stuffing, but worse. Stuffing was gaming a system while maintaining human readability. This is restructuring human expression at its foundation. We're not tricking algorithms; we're becoming algorithmic ourselves.

Theodore Adorno once wrote about how the culture industry's standardization doesn't just affect what we consume but how we think. The same process is at work here, but accelerated and more invasive. When we write "Salesforce's CRM helps Salesforce users manage Salesforce contacts," we're training ourselves to think in chunks. When we can't use pronouns, we can't build complex arguments. When every paragraph stands alone, we can't develop ideas across pages. When context can't travel, thought can't accumulate. And so we’re without memory. Without the ability to connect this to that to the other thing.

The publishers' dilemma illuminates the broader crisis. When The Atlantic's CEO tells employees to assume Google traffic will reach zero, he's acknowledging reality: the web is bifurcating into human spaces and machine spaces, and the machine spaces are eating the human ones. The economic incentives all point in one direction. Toward the literal, the explicit, the defensively clear.

But clarity and explicitness aren't the same thing. Human clarity emerges from context, from the careful building of understanding where each sentence enriches the last. Machine explicitness demands the opposite: each unit of meaning isolated, self-contained, atomized.

This is baby talk for machines. Language reduced to its most literal elements, stripped of style, variation, and flow. It's efficient for machines, exhausting, not to mention, unmemorable for humans, and increasingly mandatory for anyone who wants to be understood in a zero-click world.

The tragedy isn't that machines can't read between the lines. It's that we're forgetting why we put anything there in the first place. When we accept that every sentence must stand alone, we lose the ability to build ideas that require connection, development, nuance. We lose the essay, the argument, the story… all the forms that depend on maintained context.

This isn't technological determinism. We could build systems that maintain context, that read sequentially, that preserve the full richness of human expression. But those systems don't scale to web scale, don't deliver instant answers, and don't maximize engagement metrics. So we get machines that read like amnesiacs, and we adapt our writing to match their limitations.

The pronoun becomes a casualty of scale. And with it goes something harder to measure but essential to human communication: the faith that our reader is traveling with us, maintaining the thread, building understanding across time.

We’re all going to end up writing for machines simply because the economics demand it. But we’re not optimizing content. We’re participating in a restructuring of human expression. We’re helping machines build a web where context can't travel between paragraphs. And, it’s not because it's impossible, but because we've agreed that it isn't necessary.

The machines won’t get better at reading between the lines if we get better at having nothing there to read first.

Kyle Byers posted the screenshot on Friday at 7 AM Pacific, and I haven't stopped thinking about it since. Google's AI Overview was telling searchers that Semrush offers nonprofits a 50% discount. They don't. That's Asana's discount, from entry #7 in a listicle about nonprofit tools. Semrush was only mentioned in the article as an example of expensive software to avoid. But between entry #7 and wherever Semrush appeared, the phrase "The software offers" lost its anchor, and Google's AI inverted the entire point: gifting a discount to the very tool the article warned against.

Zero clicks to clarify. Zero clicks to correct. We’re left with a machine's interpretation, served in an AI Overview as truth.

I spent the next three hours thinking about what is happening to language on the web. Not that SEO or AI slop hasn't mostly ruined that already. I went down the history of the web rabbit hole. Tim Berners-Lee's original vision, the semantic web that never happened (or is just beginning), the gradual enshittification of everything, including what we write online. It was killed somewhere between keyword stuffing and machines that read like goldfish with brain damage. I started with early HTML specs, ended up at Morozov's tech solutionism takedowns, and realized we'd already surrendered before the war began.

The web was supposed to be connections, links, context traveling between documents like electrons through copper. Berners-Lee imagined hypertext that would let meaning accumulate across sources. Instead we got keyword stuffing, private blog networks, programmatic SEO, and now machines that interpret source content like a seven-year-old on amphetamines.

Kevin Indig's analysis from last week laid out traffic patterns that are more like extinction events. Publishers are down 15-45% from AI Overviews alone, some are seeing 70% drops since April. The Atlantic is preparing for zero Google traffic. Not reduced. Zero. Data says the age of clicks is ending and being replaced by machines that read everything and send no one anywhere. OpenAI scrapes content 179 times for every actual human visit. Anthropic is 8,692:1. Perplexity is 369:1.

Garrett Sussman gives us deeper insight. Only 4.5% of AI Mode sessions generate even one click… as in singular, as in the statistical equivalent of zero. And new data points will continue to emerge like a bad rash.

RAG systems chunk text, typically into overlapping token windows, for embedding and retrieval. But these chunks aren't semantic units, because semantic chunking is typically resource intensive. They're arbitrary slices that often split mid-sentence, and they carry fragments of meaning that the retrieval algorithm treats as complete thoughts.

When the RAG system chunks that listicle, "The software's free Personal plan" might land in the same chunk as part of entry #8, or in an overlap zone between entries, or orphaned in its own fragment. During retrieval, when someone searches for "nonprofit software discounts," the system might surface chunks containing both this orphaned discount reference and mentions of SEMRush from elsewhere in the article. The embedding model creates vectors, and cosine similarity between chunks containing 'software... discount' and 'SEMRush' might be high enough to retrieve both, even though they're semantically unrelated; producing a Frankenstein answer stitched from pieces that were never meant to connect.

Multiply this by billions of chunks and you get the new semantic web, where nothing refers to anything.

The technical term for what's breaking is anaphora resolution: tracking pronouns and references across context. When I write "Asana launched in 2008" and follow with "It now serves millions," you understand that "it" refers to Asana. This is so fundamental to human communication that linguistics considers it a universal feature of natural language. We build meaning through chains of reference, each sentence inherits context from what came before. But when text gets chunked for retrieval, these chains get severed. The system might retrieve chunk 2 without chunk 1, leaving "it" pointing at nothing.

But when documents get processed for retrieval, they're chunked into discrete units that get embedded, indexed, and searched in isolation. The economics of scaling to web-scale content makes storing full document context prohibitively expensive. So these systems retrieve fragments: quick chunks that match your query but lack the original context.

The result is a retrieval system that exhibits selective amnesia. It can find the statement perfectly while having no access to what preceded it. It surfaces facts but not their relationships. When these fragments get fed to an LLM for synthesis, even the smartest language model can't reconstruct connections that were severed during chunking.

This would be interesting if these systems remained research curiosities. But they've become the primary interface between people and information. When Google's AI Overviews generate “approximately zero clicks” to cited sources the machine's interpretation becomes the only truth that reaches an audience.

The result is a new kind of reader that exhibits what we might call selective semantic amnesia. It can quote your paragraph perfectly while forgetting what preceded it. It remembers facts but not their relationships. It processes language without maintaining the coherence that makes language language.

The industry’s response has been... “Well you must CHUNK text,” which other than being semantically false (writers write, systems chunk), means we’re surrendering to the machines. Writers will now repeat entity names obsessively: "Notion's collaboration features make Notion ideal for teams who need Notion's real-time editing." It reads like a concussion but communicates well with machines.

Watch the mutation happen:

Human English: "We launched our CRM in 2019. It now serves over 10,000 businesses with industry-leading automation."

Machine English: "Salesforce launched Salesforce's CRM in 2019. Salesforce's CRM now serves over 10,000 businesses with Salesforce's industry-leading automation."

The pronoun is murdered. The reference is explicit. The sentence is wearing a syntactic awareness name tag forever.

Our best hope is that we maintain dual content strategies; one version for the humans who barely visit, another for the systems that do the visiting for them.

We're told this is optimization, but it's closer to capitulation. Every time we restructure human communication to accommodate machine parsing, we're changing language itself. The defensive repetition, the context-free paragraphs, the death of elegant variation. And it’s sad that these aren't temporary adaptations. They're the early stages of a shift in how we write, talk, and communicate forever.

The mechanism is that dependency creates deformation. As traffic patterns shift from direct visits to AI-mediated responses, publishers who want to survive must write for their new readers. But these readers impose constraints that are antithetical to natural language. They require a kind of writing that maintains no state, assumes no memory, builds no narrative flow.

What emerges is a dialect of English optimized for machines; call it Machine Readable English (MRE). Its rules are: Every sentence stands alone. Every reference is explicit. Every attribution is defensive. It is language stripped of the very features that make it efficient for human communication.

It’s kind of like keyword stuffing, but worse. Stuffing was gaming a system while maintaining human readability. This is restructuring human expression at its foundation. We're not tricking algorithms; we're becoming algorithmic ourselves.

Theodore Adorno once wrote about how the culture industry's standardization doesn't just affect what we consume but how we think. The same process is at work here, but accelerated and more invasive. When we write "Salesforce's CRM helps Salesforce users manage Salesforce contacts," we're training ourselves to think in chunks. When we can't use pronouns, we can't build complex arguments. When every paragraph stands alone, we can't develop ideas across pages. When context can't travel, thought can't accumulate. And so we’re without memory. Without the ability to connect this to that to the other thing.

The publishers' dilemma illuminates the broader crisis. When The Atlantic's CEO tells employees to assume Google traffic will reach zero, he's acknowledging reality: the web is bifurcating into human spaces and machine spaces, and the machine spaces are eating the human ones. The economic incentives all point in one direction. Toward the literal, the explicit, the defensively clear.

But clarity and explicitness aren't the same thing. Human clarity emerges from context, from the careful building of understanding where each sentence enriches the last. Machine explicitness demands the opposite: each unit of meaning isolated, self-contained, atomized.

This is baby talk for machines. Language reduced to its most literal elements, stripped of style, variation, and flow. It's efficient for machines, exhausting, not to mention, unmemorable for humans, and increasingly mandatory for anyone who wants to be understood in a zero-click world.

The tragedy isn't that machines can't read between the lines. It's that we're forgetting why we put anything there in the first place. When we accept that every sentence must stand alone, we lose the ability to build ideas that require connection, development, nuance. We lose the essay, the argument, the story… all the forms that depend on maintained context.

This isn't technological determinism. We could build systems that maintain context, that read sequentially, that preserve the full richness of human expression. But those systems don't scale to web scale, don't deliver instant answers, and don't maximize engagement metrics. So we get machines that read like amnesiacs, and we adapt our writing to match their limitations.

The pronoun becomes a casualty of scale. And with it goes something harder to measure but essential to human communication: the faith that our reader is traveling with us, maintaining the thread, building understanding across time.

We’re all going to end up writing for machines simply because the economics demand it. But we’re not optimizing content. We’re participating in a restructuring of human expression. We’re helping machines build a web where context can't travel between paragraphs. And, it’s not because it's impossible, but because we've agreed that it isn't necessary.

The machines won’t get better at reading between the lines if we get better at having nothing there to read first.

• Reach. In March 2025 Google attracted 269.6 million U.S. visitors; 6.8 times the amount of ChatGPT’s 39.6 million.

• Total referrals. Google pushed 3X more clicks out to other sites than ChatGPT.

• Efficiency. The average ChatGPT user clicks 1.4 external links per visit; Google’s users click 0.6 times per visit.

• Stickiness. A typical Google visit now triggers 10+ in-platform pageviews (or “queries”), up from 8.4 two years ago.

• Destinations. 64% of all ChatGPT referrals land on just 120 domains. These range from YouTube to NIH PubMed.

Why this resource exists

Over 200 marketers joined the LinkedIn post or through repost on my post last Sunday. They asked hard questions about sample bias, summer traffic spikes, and impact. This article addresses those questions and adds context to the original report (NOT a “study”) so you can cite, link, or challenge the numbers with confidence.

PS to the outlets. Please stop calling this a study. It is not a study. Thank you.

Audience size

Google still dwarfs every rival. Its U.S. audience was nearly 270 million in March 2025, while ChatGPT climbed 47% year-over-year to 39.6 million in March 2025.

Reach sets the ceiling for potential traffic. You plan strategy based on potential brand reach.

Outgoing traffic volume

Google remains the firehose to the open web. Google sends three times the total clicks to the open web, but ChatGPT is not a rounding error.

Outbound clicks (March 2025)

• Google: 175.5 million clicks, up 66 % versus March 2024

• ChatGPT: 57.7 million clicks, up 558 % versus March 2024

Clicks per user

When comparing search engines of very different sizes, ratio beats volume. ChatGPT sends users to external websites more than twice as often.

Why did traffic spike in May 2024/July 2024?

In July, Google referrals jump, then fall back. Three overlapping things may lead to an answer to why, but I’m not confident enough in the evidence to say any these ARE the reasons.

Key events & why they might have increased clicks out from Google

• 5 Mar – 19 Apr: Largest Core + Spam updates in Google history; cleaner SERPs surfaced more trustworthy pages and rewarded clicks.

• 14 May: AI Overviews rolled out at Google I/O; links inside an Overview get higher click-through than the same blue link.

• 14 Jun – 14 Jul: Euro 2024 and Copa América ran side-by-side; millions of real-time sports queries created a surge in demand for fresh sources.

• Jun 2024: Google cut AI Overviews from 84 % of searches to about 15 %; fewer AI answers pushed more searchers into organic results.

• 16 – 17 Jul: Amazon Prime Day 2024; shopping frenzy sent high-intent buyers to comparison pages.

• 26 Jul – 11 Aug: Paris 2024 Olympics; global event coverage spiked demand for schedules, scores, and highlights.

Shoutout to Kim Kuhlman, PhD for the dialogue here!

When core updates, interface shifts, and cultural events collide, even a stable source like Google can swing a large percent in referral output.

Google’s is growing stickiness on purpose... probably

Pages-per-visit on Google desktop has climbed to double digits. Every extra in-SERP click is one less opportunity for your site to earn traffic. Google claims this as “query growth”. It's good for Google's stock.

Where does ChatGPT send referrals though links?

Technology, health, and news sites absorb the bulk, but 64 percent of all external clicks go to just 120 domains, including YouTube, Wikipedia, and Amazon.

List of top1,591 referral destinations out from ChatGPT (April 2025)

Since I cannot seem to figure out how to embed anything in a Substack post, here’s adirect link to an interactive table of top destinations when users click out of ChatGPT.

What the dataset cannot see

• Browser only. Similarweb’s panel measures desktop and mobile browsers. App activity, including most ChatGPT iOS/Android usage, stays hidden.

• ±10 % typical variance, ±30 % in edge cases. We validated against 100+ first-party 1st party analytics sets. Results fell inside those bands.

• OpenAI/Google self-referrals removed. Links to any openai.com subdomain (≈27.7 % of all ChatGPT clicks in March) are stripped before analysis.

Watch for branded-search lifts. If ChatGPT cites you without a link, the next step is might be a branded Google query. Also keep an eye on direct traffic. If budget allows, I highly recommend investing in an Incremental Measurement Platform so you can find where incremental growth comes from. Most businesses will not be able to fit this into their budget.

Community Q&A

Can we trust Similarweb?

Our tests show ≈ ±10 % average error against first-party logs, sometimes wider. Treat trends as directional, not exact counts.

“ChatGPT is less than 1 % of Google in my GA4. Am I doing something wrong?” (Joy Hawkins)

UTM-less links from the ChatGPT app arrive without referrer data. They blend into “Direct”. Your browser numbers will look bigger than your app numbers. I also think that the browser-based app is inconsistent in sending referral data but do not have solid proof of this yet.

“Do users click away because they distrust AI answers?” (Steven Johnson)

Possibly. Google hides links behind AI Overviews; ChatGPT shows them inline. When an LLM quotes a source, skeptical users click to verify. This faith-check behavior amplifies ChatGPT’s per-person referral rate.

“How would you use this insight for client wins?” (Saeed R.)

Focus on proven revenue drivers first (probably classic SEO). Then, enrich cornerstone pages with entity-dense, value-dense copy, schema, and unique data. Finally, publish supplemental helpful resources that LLMs might cite.

Takeaways

• Google still rules reach

• ChatGPT rewards depth

• Zero-click risk is rising

• Data context matters

Open research questions

• What share of ChatGPT users rely on the app vs. the browser?

• Will citation UX in AI Overviews change click-out rates?

Download report [PDF]

Google vs. ChatGPT Full Report [PDF Download]

Leave a comment

Here's the Original Post (via LinkedIn)

I follow data. I spend a lot of time with it.

In the spirit of building a more transparent community (Michael King's goal of the Relevance Engineering concept presented at #seoweek), I'm sharing a light version of a report I spent many weeks manually putting together.

The full downloadable report is at the beginning of the post. :)

My intention is to find the answer to what us SEOs/AI Opportunists seem to be arguing about on here. (LLMs as search engines and user adoption rate).

If you know me well personally (e.g. Renee Girard/Tony Van Hart/Steve Kroll/Lauren Keepers/Tamara Hellgren), you know my take on this, but my take is not important. The data is.

First. My goal is to help take emotion out what's going on at a high level in search to level set us. I'm a community-focused collaborator NOT an opportunist.

Second. I welcome any constructive conversation. Have different data? Let's dialogue. Disagree? Please tell me what I'm not seeing. Have different methods/interpretations? Yes, let's figure this out together. If you wanna tell me to get bent, go for it, but that's mean.

Third. Some of what I've been seeing here on LinkedIn lately have been cherry-picked metrics to defend or dispute a hot take or to promote a tool/service/data provider. But also, please don't bucket me in with influencers/fear-mongers. I am not your opposition.

What The Data Says

Here’s what 13 months of U.S. clickstream data (Mar 2024–Mar 2025) shows for Google/OpenAI. Again, this is data-first, and did my best to exhibit transparency in all 26 pages.

• Google's audience is almost 7X the size of ChatGPT. BUT ChatGPT is over 2X more likely to send traffic to external websites than Google.

• Important note: to those saying that those websites are all to a handful of websites (and I've seen a few of these posts on LinkedIn over recent days)... it's NOT much different than Google. Check page 21 for where 64% of ChatGPT off-platform clicks go to.

The read is Google delivers the MOST TRAFFIC, but keeps MORE CLICKS within platform. ChatGPT, though notably smaller, sends out web visitors MORE OFTEN.

Note of transparency: I am using Similarweb for this data. I am NOT being compensated (actually, I'm a satisfied customer). I moved on from other popular providers early this year after six months of testing different clickstream products against 1st party data I had access to. It was the closest (+/-10% on average, though sometimes up to +/-30%).

Also, a special shoutout for Baruch Toledano for always taking time to have meaningful conversations when we cross paths.

I'll continue to share what I see if there's value in it for the SEO community. Want to contribute? HMU.

Structured data is a way of adding labels to a webpage so that machines know exactly what each piece of information means. Schema markup is the vocabulary that supplies those labels. In this episode, Martha Van Berkel explains why these tools now sit at the heart of both search and AI. She shows how clear markup builds knowledge graphs, large maps of facts that modern AI taps to give answers, and why writers must learn to serve people and algorithms at the same time.

Key Takeaways

• Structured data lifts a page’s visibility by telling search engines what the content is, not just what it says.

• Treat schema markup as part of your content plan, not an after-thought for SEO.

• AI systems may already use schema markup to check and surface facts. Expect that link to likely deepen.

• Knowledge graphs (networks of linked entities) let companies scale AI features fast.

• Writers who ignore machines risk losing reach; writers who ignore people lose trust.

• Marketing teams still struggle most with scaling consistent, correct markup.

• Close collaboration between content creators and SEOs fixes gaps and speeds results.

Featured Quotes

• “Schema builds knowledge graphs.”

• “Structured data helps with factual answers.”

• “This is the world we’re in today.”

Chapters

00:00 - Why non-SEOs should care about structured data01:55How thinking on structured data has shifted

03:14 - Schema markup in AI and search

06:20 - Scaling challenges and knowledge graphs10:14The future of writing for humans and machines

Quick Definitions

• Structured data: Extra code that labels each piece of content so machines can read it without guessing. Why it matters: It makes pages eligible for rich search results. Example: Adding recipeIngredient tags to list each item in a recipe.

• Schema markup: The shared vocabulary for that code, managed by Schema.org. Why it matters: A common language lets every search engine read the labels the same way. Example: Using Product and price tags on an e-commerce page.

• Knowledge graph: A database that connects facts into a web of relationships. Why it matters: AI uses these graphs to answer questions with confidence. Example: Google’s graph linking “Martha Van Berkel” to “Schema App” and “structured data expert.”

In this conversation, Tyler Einberger and Renee Girard discuss the changes they’re seeing first hand in the SEO world, the impact of generative AI, and the importance of community engagement. They explore the challenges faced by e-commerce businesses, the innovative approaches of startups, and the necessity of data-driven (but not-to-much-data-overload-driven) decision-making in the SEO industry. The discussion also touches on the prevalence of mediocre content and the need for fresh perspectives in the field.

Topics

• SEO is still about getting websites found on Google.

• The SEO landscape includes more players than just Google.

• E-commerce strategies are evolving with AI and new technologies.

• Startups are more agile and innovative in their SEO approaches.

• Data-driven decision-making is essential for effective SEO strategies.

• The quality of content is becoming increasingly important.

• Community engagement is vital for SEO professionals.

• Adapting to change is crucial in the SEO industry.

• Mediocre content is becoming the norm, which is detrimental to the industry.

• Celebrating successes and learning from failures is key to growth.

Soundbites

• "What is SEO as we know it?"

• "Why do we deserve to rank?"

• "The bar is now rising in SEO."

Keywords

SEO, digital marketing, e-commerce, startups, data-driven, generative AI, content strategy, community engagement, Milwaukee DMC, MKE DMC, Mike King

Disclaimer

Our thoughts are our own and not a reflection of our employers.

Quick context for why I'm bringing this topic up:

• Agency clients are asking what SEO's future actually looks like—props to them.

• Mike King is joining our next marketing meetup (at MKE DMC on March 12), and he’s leading a session called "The End of SEO As We Know It." It got me thinking—a lot, “What even IS SEO as we know it?”.

• Measuring SEO is messy and complicated.

• I'm excited about exploring deeper, potentially more authentic ways to connect with audiences.

Sound good?

Structural Shifts in Search Behavior and Technology

Search engines in 2025 are very different from a decade ago. Google’s evolution as a publicly traded company has led to search results pages crowded with ads, answer boxes, and other features that often satisfy users’ queries without a click. Most Google searches – about 58.5% in the U.S. – now end without clicking on a website​ (SparkToro). In other words, more than half of searchers get what they need directly on Google (through featured snippets, maps, knowledge panels, etc.) and never visit an external site. This “zero-click search” (s/o to Amanda Natividad for the term) trend represents a structural shift in how people consume information via search.

Figure: Breakdown of what happens after a Google search in 2024. In the U.S., 58.5% of searches are zero-click (no subsequent site visit). Only about 36% of searches result in a click to an external website, while nearly 30% of clicks go to Google’s own properties (YouTube, Maps, etc.). Europeans see a similar pattern​.

Compounding this trend is Google’s increasing self-reliance. Nearly 30% of all clicks on Google Search now go to Google-owned properties (like YouTube, Maps, Google Shopping)​. By design, Google keeps users in its ecosystem longer, which means only about one-third of search clicks lead to the open web​. For businesses, this translates to fewer opportunities to attract visitors via traditional organic search listings. Even when a user does click, the first organic result is often far down the page, beneath ads and rich features. As one marketing analyst observes, “the first organic search result now appears below the scroll in many searches,” because Google’s own answers (Position Zero) come first​. Earning a top organic ranking no longer guarantees traffic the way it once did.

AI-driven search is another shift. The rise of Large Language Models (LLMs) and Generative AI (GenAI) assistants—like OpenAI’s ChatGPT and Google’s Gemini—has begun to change search habits. Surveys indicate that 55–77% of digital professionals are using Google search less because they turn to AI tools instead​. In fact, Gartner predicts search engine query volume will drop by 25% by 2026 due to AI siphoning off queries. The logic is that people either ask chatbots directly for information or get their answers from new AI-generated summaries on the search results page. Google’s introduction of generative AI answers (the Search Generative Experience, now branded as AI Overviews) shows how big this trend is – certain queries (most are informational in intent right now) produce an AI Overview, meaning the user may find what they need without scrolling to the traditional results​. Early data is mixed on the impact: some website owners report traffic declines for content where Google’s AI overview appears, especially on mobile (my agency’s website is down ~20% for these queries)​. On the other hand, Google claims (without hard data) that users “love” these AI summaries and even search more because of them. Search is becoming more answer-oriented and less click-oriented. I think we saw this coming, though.

It’s important to note that Google remains dominant despite these changes in behavior and technology. While alternatives like Bing Search (with AI features) and standalone AI assistants have grown, Google’s market share and total search volume have not collapsed. In fact, by spring 2024, Google searches per user were at historic highs in both the U.S. and EU​. This suggests that people still “Google” frequently – but Google is increasingly satisfying those queries on its own platform. The structural shift, therefore, is not that users have abandoned search, but that search engines (especially Google) have changed the playing field: keeping more traffic, showcasing more of their own services, and leveraging AI to provide instant answers. For anyone relying on search traffic, these shifts fundamentally alter the value and approach of SEO.

The Diminishing Returns of Traditional SEO Tactics

Many of the classic SEO methods that once reliably drove traffic are yielding less payoff in 2025. Several converging factors explain the declining impact of traditional SEO:

Crowded SERPs and Falling Click-Through Rates

As noted, even a #1 organic ranking now isn’t what it used to be. With up to four ads at the top and multiple rich features, the #1 organic result might effectively be halfway down the page​. Studies show that when Google adds a featured snippet to a page, the click-through rate on the first organic result drops significantly​. Overall organic CTRs have been declining as users engage more with ads and on-page answers​. One industry consultant bluntly summarizes: “Your site effectively starts at position five” if four ads are above it, “significantly reducing the likelihood of attracting clicks.”​ Even strong SEO performance can translate into far fewer visitors than in years past.

Zero-Click and Answer Scraping

Google directly answers more queries than ever. What started as weather, calculations, basic facts, sports scores, etc. has now—thanks to widespread AI Overviews rollouts—encroached on most informational searches that aren’t simple answers. Informational content that once drew visitors now often sees way less than pre-rollout. Over 59% of searches end with no click at all​, which represents traffic that no SEO tweak can recapture. As a CMO put it, “where before those clicks would go to the webpage that occupied Position One, now Google answers those searches with featured snippets and knowledge panels”​. This leaves us SEOs “fighting for crumbs” of the remaining traffic​.

Competition and Saturation

Virtually every business has caught on to content marketing and SEO, leading to a glut of content. “The big gains many companies were used to getting in the past are difficult to replicate in the present,” notes marketing strategist Tim Rayl, “because the SERPs have never been more competitive and SEO has never been more time and resource intensive.”​

There are only 10 (or fewer) organic spots on page one, but many competitors vying for them — not all of them are direct business rivals. Some are aggregator sites, review platforms, and large publishers. In local niches, for example, a small business’s SEO competition isn’t only with fellow locals but also with websites like Yelp, Angi, or Google’s own local packs​. Getting visibility in non-brand search results often requires significantly more content, technical optimization, and link acquisition than it did a few years ago, just to keep pace with the SERP competition. This increases the cost and effort for the same or even diminished business outcomes.

Favoritism for Big Brands

Google’s algorithm updates over the past few years (leaning into “E-E-A-T” guidelines, emphasizing authority and content quality, but in reality making organic search results worse to sell more ads) mean that established brands enjoy a ranking advantage that content alone can’t easily overcome. An SEO voice I trust, Kevin Indig observed that “Google gives brands more visibility in most verticals... ‘Brand’ has replaced links as a factor” to a large extent​. High-authority sites with strong reputations (and abundant quality backlinks naturally accruing) own many results, making it harder for newer or smaller players to break in. In practical terms, a startup can do everything “right” in SEO but still struggle if it lacks brand recognition signals that Google trusts. This again stretches the timeline and investment needed to see ROI from traditional SEO tactics.

Economic Pressures / Diminishing ROI

The net effect of the above trends is that the return on investment for SEO has shrunk for many businesses. It takes longer and costs more to rank, and even when you do rank, the traffic and conversions gained are lower than before. Ben Luong’s analysis of the landscape noted the “high cost and low probability of SEO success” in a winner-takes-all scenario: achieving and sustaining top rankings requires substantial investment in content and links, yet for every site that reaches #1, hundreds of others with similar efforts do not​. Moreover, even a high ranking may deliver less traffic if Google fills the page with its own elements. All this means the marginal ROI of each additional dollar or hour spent on SEO often declines. Marketers speak of hitting “plateaus” where pouring more content or links doesn’t yield proportional traffic growth. In economic terms, many organizations are seeing diminishing returns on their SEO spend​ – a stark change from earlier years when SEO was lauded as a high-leverage, “free” traffic channel.

The data supports this sobering picture. In 2024, Gartner predicted a 25% drop in organic search traffic by 2026, and some marketers are already feeling the downturn​. Wil Reynolds of Seer Interactive reported that his company’s site traffic from Google was down 40% year-over-year in early 2024​, echoing the broader trend. And while individual experiences vary by industry, the consensus among many practitioners is that SEO isn’t the easy win it once was. It “works, just not as well as before”​– and the gap between effort and payoff is widening. In fact, a growing sentiment in the industry is that “SEO is just not a good place to put your money” compared to other marketing investments, given how much Google has squeezed organic reach​. That may be an extreme view, but it represents a real economic rationale: businesses must consider alternative strategies when the opportunity cost of chasing SEO is high.

People Who I’ve Grown To Trust

The following people have spent decades in SEO and content. They are vocal about the need to rethink approaches. Here’s some insights from a few prominent people who produce original research and commentary:

Amanda Natividad (VP Marketing at SparkToro): Amanda emphasizes that “SEO strategy ≠ content marketing strategy.” In her view, a solid content strategy goes beyond just SEO goals. SEO is just one means to distribute content, not the end goal itself​. She points out that obsessing solely over search traffic can lead marketers to ignore other valuable content outcomes. In her career (including leading content at Fitbit’s B2B division), Amanda often deliberately de-prioritized SEO to focus on content that served broader marketing functions. “Not focusing on SEO is how I was able to get buy-in for content from the executive team,” she notes, referencing how she aligned content with lead generation and sales support instead​. At one point, Fitbit’s B2B unit didn’t even have a public blog, because “growing organic traffic wasn’t an initial business priority”​. Instead, Amanda created white papers, interactive tools, and webinars — content used in demand generation campaigns that directly produced thousands of new leads each quarter​. Her approach illustrates that content can be a “service” to multiple marketing goals (PR, sales enablement, demand gen) and not just a vehicle for SEO​. Amanda’s insight is a reminder that in 2025, winning organically requires a holistic content mindset, not a narrow fixation on keywords and rankings.

Rand Fishkin (Co-founder of SparkToro; former CEO of Moz): Rand has been at the front of analyzing search trends, especially the rise of zero-click searches. His research has put hard numbers to what many suspected: Google is keeping more clicks for itself each year. In 2024, Rand’s analysis (using clickstream data) found that almost 60% of searches now end without a click, and only 36% of searches result in a click to a non-Google website​. He frequently calls attention to Google’s incentive to “funnel search traffic to its own properties”, whether that’s YouTube, Maps, or newer services​. The implication of Rand’s work is clear – relying solely on Google for audience acquisition is increasingly risky. He argues marketers should invest in strategies that don’t depend on Google’s benevolence: for example, building direct audiences (through email, communities, or social), and engaging people on platforms beyond search. Rand has even popularized the idea of “zero-click content” on social media – content designed to reach and engage users within platforms like Twitter or LinkedIn, without requiring a click-out link​. This flips the usual SEO mindset on its head: rather than pull an audience to your site via Google, push content to where audiences already are. His perspective, rooted in data, highlights the need for marketers to diversify traffic sources and not be overly dependent on the goodwill of a search monopoly.

Wil Reynolds (Founder of Seer Interactive): Wil is known for his candid take on the limits of traditional SEO. Confronted with the reality of declining organic traffic (as noted, Seer saw a steep drop in 2024), Wil advocates for a shift in how we measure and pursue “SEO” success. He notes that people “are still buying stuff, they are just harder to track than ever”​. In a world of disappearing keyword data, “dark social” traffic, and AI answers, traditional SEO KPIs (like raw organic sessions or keyword rankings) don’t tell the full story​. Wil’s advice to SEO professionals is to broaden their focus to business outcomes and new metrics. For example, he suggests tracking on-page engagement (impressions, interactions) in places where you can’t get click data, and integrating SEO with other channels’ metrics​. He also stresses validating external predictions with one’s own data – when Gartner says 25% traffic drop, don’t just dismiss it, check your analytics and prepare​. Wil’s overarching insight is that SEOs must become true marketers and analysts, not just tacticians. The channel is changing (e.g. AI search giving “zero data”), so we have to adapt our strategies and reports to show value in new ways. His perspective reinforces that SEO isn’t “dead,” but it must integrate with a larger strategy and prove its worth in terms executives care about (like leads or revenue), especially as direct traffic attribution becomes murkier.

Kevin Indig (Growth advisor, former director of SEO at Shopify): Kevin writes a lot about shifts in organic search. On Growth Memo, he often writes about the growing importance of brand and user experience in organic growth. Indig observes that many SEO practitioners cling to comfortable, outdated tactics that now yield diminishing returns. In contrast, the real wins come from modernizing the playbook. ​As AI disrupts content creation and search, Kevin argues that the “winning formula is clear: genuine connection, unwavering authenticity, and relentless surround sound forged on deep audience understanding.” My take: Right now most of that is unachievable through GenAI.​

In other words, succeeding in organic marketing today means building a brand that audiences trust and creating content that resonates deeply (often in a more personal, human voice) rather than churning out keyword-stuffed pages to fight for clicks in a “sea of sameness”. He advocates a “surround sound” approach – make sure your brand appears in all the places your audience is looking (blogs, podcasts, YouTube, Reddit, etc.), not just your own website. SEO in isolation is not enough; it’s one part of a broader audience engagement strategy. Brand-building activities drive SEO: for example, startups that invest in PR, community, and unique content often earn the kind of high-quality backlinks and mentions that Google/SearchGPT reward​. SEO outcomes now often flow from non-SEO actions (like great product experiences that generate buzz, or thought leadership that builds brand searches). Following his advice means SEOs should expand their skillsets – e.g. understand user research, storytelling, and multi-channel marketing – to remain effective.

Common theme: the old, siloed way of doing SEO is fading. Integrate SEO with content strategy, focus on brand and audience, and measure success in terms of engagement and business impact. None of them says, “Ignore SEO completely.” Instead, they say SEO must be reimagined and incorporated into a bigger organic strategy.

This is a key mindset for 2025+.

Want to Be More Than SEO?

I do.

Given the trends above, many brands have transitioned past traditional SEO into more holistic organic marketing strategies. Instead of treating SEO as the be-all-end-all, they focus on broader goals like brand awareness, demand generation, and community building – with SEO being one channel among many. Here are a couple of illustrative examples and case studies of businesses finding success by looking past the SEO silo:

Demand Generation

As mentioned by Amanda Natividad (who led the effort), Fitbit’s enterprise division deliberately deprioritized SEO in its early content strategy​. They identified that immediately chasing blog traffic wasn’t as valuable as creating content to feed their sales pipeline. So, the team focused on demand gen content: in-depth white papers, interactive tools, webinars, and research pieces aimed at educating potential customers and capturing leads​. This content was promoted through email, events, and partnerships – channels where the target audience already engaged – rather than waiting for them to search. The result was a steady flow of thousands of new leads each quarter, even in the absence of significant organic search traffic at first​. Over time, some of that content did get picked up by search (as the site gained authority), but by then the business had already reaped rewards. The takeaway is that a new website or product can grow through content marketing without relying on Google at all. By aligning content to business development (instead of just keywords), Fitbit built a sustainable pipeline. SEO became a bonus rather than the sole objective. Many B2B companies are now taking a similar approach – using content as fuel for webinars, account-based marketing, and social media engagement, essentially treating content as a multi-purpose asset. This is a departure from the earlier era where a startup might invest solely in SEO blogs and wait a year for them to rank.

Original Content as a Moat

Another path beyond SEO is investing in brand-building content – the kind of material that people seek out by name or share organically because it’s uniquely valuable. We can look at digital media companies for inspiration. For instance, Industry Dive (a B2B media publisher) saw its publications’ web traffic grow 18% year-over-year in 2024, at a time when many sites were losing search traffic​. The reason? They produce specialized, authoritative journalism in various industries, which Google’s algorithms reward for expertise. Quality journalism and thought leadership content tend to attract direct visits, email subscribers, and word-of-mouth sharing – and they rank well because they meet Google’s E-E-A-T standards. It’s a virtuous cycle that doesn’t rely on SEO tricks, only on quality and authority. Similarly, many SaaS companies have created media-style content hubs (blogs, podcasts, video series) that establish them as trusted voices. Take ProfitWell (a SaaS firm acquired in 2022): they built a media arm called Recur, featuring video series and reports on SaaS metrics. This content wasn’t designed primarily for search; it was designed to earn the attention and trust of their target audience (subscription businesses). Over time, that brand affinity translated into organic growth: people would search specifically for ProfitWell’s studies or subscribe to their updates. The general pattern is that by creating content with distinct value – whether proprietary data, research, or entertainment – these brands generate demand that isn’t dependent on generic search traffic. Even if their SEO traffic fluctuates, they have an audience that will seek them out.

Holistic “Surround Sound” Strategy

Some businesses have successfully transformed their marketing by implementing a surround sound approach (as Kevin Indig described) – ensuring their presence across various platforms and communities so that potential customers encounter them organically everywhere. For example, consider a hypothetical mid-size software company that formerly invested only in blogging for SEO. If they pivot to a surround-sound strategy, they might do things like: launch a community forum for their niche, sponsor a niche podcast, contribute guest columns to industry publications, engage on Reddit and LinkedIn discussions, and produce YouTube tutorials. Over time, these activities generate a flood of brand mentions and independent content about the company. One real-world case: HubSpot in the 2010s started as an SEO-heavy content machine (with thousands of blog posts) but eventually expanded into a broader media company – launching the HubSpot Academy, podcasts, and acquiring The Hustle (a popular newsletter). Now, a huge portion of their organic traffic comes from branded searches or direct visits, driven by their reputation and ecosystem, not random Google queries. Yes, I’m aware of what happened in Feb 2025, but the point is still relevant and true.

The lesson is that companies can reduce their reliance on unpredictable Google algorithms by cultivating direct audience relationships and a strong brand presence across channels. In doing so, they often find their organic (non-paid) traffic remains robust even if pure SEO traffic lags. These holistic strategies treat “organic growth” as more than SEO – it’s the sum of search, social, community, referral, and direct traffic that comes from engaging content and brand strength.

It’s worth noting that transitioning beyond SEO doesn’t mean abandoning SEO altogether. In many of these cases, the companies still benefit from search traffic – it’s just that search is one outcome of doing lots of things right rather than the single focus. By building brand and demand through other means, they incidentally improve their SEO (through more mentions, links, and user trust). This contrasts with a narrow SEO-only strategy that might achieve rankings without broader brand love – an increasingly fragile strategy today.

Why SEO Remains Marginalized in Marketing Strategy

Given SEO’s proven potential (it can drive huge traffic and revenue in certain cases), one might ask why it often seems marginalized or under-resourced in many companies’ broader marketing strategies. Indeed, SEO is sometimes treated as a sideline activity, getting a sliver of the budget compared to channels like paid advertising or social media. There are several reasons for this phenomenon, supported by both research and candid observations from within the industry:

Attribution and Visibility Challenges

Marketing executives live and die by metrics, and SEO is notoriously difficult to attribute and forecast. Unlike PPC where you can spend $X and see immediate clicks, SEO is slow and the results are diffuse (often branded as “organic” or “direct” traffic, which may undervalue the SEO effort). A common refrain is that SEO’s impact is hard to prove. It’s not that SEO doesn’t work, but establishing the ROI to a CEO or CMO – especially in the short term – is challenging. This leads to SEO getting a “shoestring budget compared to PPC or TV ads,” and having to fight to justify its existence​. In many organizations, SEO is still seen as somewhat mysterious and not wholly controllable, so it doesn’t get the upfront investment that clearly trackable channels do.

“Last-Click” Bias and Organizational Silos

Regarding attribution, companies often credit the final touchpoint that led to a sale. SEO might start a customer journey (via a blog discovery), but by the time the customer converts, another channel (like a direct visit or an email) gets the credit. This can make SEO’s contribution appear smaller in analytics. SEO teams are sometimes siloed away from core marketing or product teams. They might sit in an isolated corner, focused on technical tweaks or content farms, rather than being integrated into campaign planning. This isolation can cause SEO to be an afterthought in significant marketing initiatives rather than a strategic pillar. Over time, this marginalization becomes a self-fulfilling prophecy: the less integrated SEO is, the less others see its value.

Historical Baggage and Misconceptions

SEO has at (many) times, had a reputation problem. Past practices (like spammy link building or “SEO content” that sacrifices quality for keywords) created skepticism among other marketers and executives. Some still (unfairly) view SEO as a set of hacks or snake oil rather than a modern, legitimate strategy. Thus, SEO professionals often face internal skepticism, needing to educate colleagues that SEO today is about quality content and user experience. But not everyone keeps up with SEO’s evolution, so in some companies, it’s still “undervalued and disrespected” as a channel, as one survey found​. When a discipline is misunderstood, it’s likely to be marginalized.

Focus on Immediate Results

Businesses under pressure for quick growth may favor channels that produce immediate leads and sales (e.g., paid search, social ads, outbound sales). SEO, by contrast, is a long game – it can take 6-12 months to show significant results and often 2-3 years to reach peak performance for a campaign, according to data analyses of ROI timing. Many firms (especially startups) simply can’t wait that long or are unwilling to invest in something that doesn’t pay off this quarter. So, even if SEO could yield a great payoff in year 3, the economic rationale in year 0 might lead a company to allocate budget elsewhere that promises a nearer-term win. This short-termism in marketing planning often leaves SEO underfunded and underutilized.

Google’s Erosion of Confidence

Ironically, as Google has taken more control over search traffic, it’s not just SEOs who have noticed – executives see it too. When they hear that “only 3-4 out of ten Google searches result in a website click now,” they might conclude that SEO is a shrinking opportunity. Some businesses question why they should pour resources into a channel where the platform owner (Google) keeps tightening the taps. There is a sense that the game is rigged in favor of Google’s revenue, which makes some executives hesitant to bet big on SEO. Instead, they diversify to other channels (even if more expensive) where they feel they have more control over their fate.

Despite these factors, it’s important to acknowledge that SEO still holds significant potential value – and that marginalizing it ultimately can be a mistake. The issue is often one of integration and perspective. SEO works best not as a siloed tactic but as part of an integrated marketing strategy (e.g., aligning SEO content with PR efforts or using SEO data to inform product strategy). When treated in isolation, it’s easier for organizations to undervalue it. But when SEO insights permeate content planning or SEO-driven content is repurposed across marketing channels, its impact becomes more visible and appreciated. For instance, if a company’s blog post ranks well and gets cited in the press and shared on social media, SEO is suddenly contributing to multiple goals (brand credibility, referral traffic, etc.). In 2025, the companies that get the most value from SEO don’t think of it as “SEO” – they see it as an organic presence, part of a tapestry alongside other efforts. Unfortunately, many businesses have not yet made that philosophical shift, so SEO continues to be marginalized – viewed as a niche speciality rather than a core strategic function.

Being Agile

What should SEOs do next? The answer is to move beyond the narrow confines of “SEO tactics” and embrace a broader organic strategy. Here are some clear paths forward.

Invest in Brand and Community

Brand is the excellent amplifier that makes all marketing more effective, including SEO. As Kevin Indig noted, brand-building activities (like PR, thought leadership, and community engagement) now have direct SEO benefits​, and they also provide independent value by creating loyal audiences. Professionals should shift some focus to nurturing a community around their niche – whether through forums, social media groups, newsletters, or events. When you have a loyal audience who trusts you, Google’s algorithm changes matter far less because that audience will seek you out. Think of it this way: in the media industry’s shift to digital, the publications that survived were those with strong brands (e.g., The New York Times leveraging its brand in digital subscriptions), whereas weaker brands vanished. Building brand equity in your space is a long-term moat that protects and boosts organic growth.

Be Cool With Multi-Channel “Organic” Marketing

If the playing field on Google is challenging, remember that search is just one channel. Audiences discover content through YouTube, podcasts, newsletters, LinkedIn, Twitter, TikTok, and more. Savvy marketers in 2025 allocate effort to these organic channels as well. For example, instead of writing 10 similar blog posts hoping to rank, you might write two great posts and produce 3 video explainers, one webinar, and four LinkedIn articles. This diversified content strategy means if Google doesn’t send traffic, YouTube, LinkedIn, or an email campaign might. It’s akin to how advertisers adapted when traditional TV lost audiences to streaming and social – they followed the eyeballs. Digital ad spending surpassed TV ad spending globally years ago as marketers reallocated budgets.​. Similarly, organic marketers should follow where their audience’s attention flows. If young professionals in your niche spend more time on Reddit threads or Slack communities than searching on Google, you need to show up there with valuable content or engagement. The goal is to create an “everywhere” presence (without being spammy) so your brand becomes hard to miss in your domain. This surround-sound approach increases the chances of word-of-mouth growth and also indirectly benefits SEO through more mentions and links.

Focus on Content Quality and Originality

In an age where AI can generate any article, what cuts through the noise is original, insightful content. This is where you should double down. Conduct proprietary research, share unique data or case studies, develop a distinct voice or perspective – do what generic AI or template-driven competitors cannot. The earlier analogy to media is instructive: When print media declined, many publications had to reinvent themselves by providing deeper analysis, interactive digital content, or subscriber-only quality journalism to survive. Quantity alone could not win. This means pivoting from a keyword-first to an audience-first mindset for SEO and content professionals. Instead of asking “What keywords can I rank for?”, ask “What does my audience genuinely care about and how can I tackle that better than anyone?”. By creating exceptional content, you accomplish two things: you appeal to your audience (building brand loyalty and direct traffic) and satisfy search engines’ increasingly demanding criteria for expertise and trustworthiness. Remember the stat that editorial sites saw growth due to Google’s E-E-A-T favoring original journalism​ – even Google effectively tells us that true expertise wins in the long run. So, professionals should hone their subject matter expertise, collaborate with industry experts, and create content that people would read even if search engines didn’t exist. That content will become the foundation of sustainable organic growth.

SEO Data as Market Intelligence

Moving broader than SEO doesn’t mean abandoning the valuable data and skills that SEO provides. Keyword research, search trends, and user query data are a goldmine for understanding customer needs and language. In fact, rather than using this data to only optimize titles and on-page content, we can use it to broadly inform product development, PR angles, and content strategy. For example, if you see rising search queries around a problem, that insight could drive a new feature in your product or a webinar topic or a pitch to the media about an emerging trend. Tory Gray noted that search data can be “powerful market intelligence” that informs better business decisions beyond SEO​. By repositioning SEO research as consumer insight research, we can up our roles in the business. This approach mirrors how PR and advertising evolved – they moved from siloed functions to being guided by data on what audiences care about. The modern SEO professional should collaborate with marketing and product teams to share search insights, thus embedding SEO’s value across the organization.

Adapt and Learn from Others

Finally, let’s consider parallels from related fields. When traditional PR saw diminishing returns in press releases and print coverage, savvy PR professionals embraced social media, influencer partnerships, and content creation as part of “digital PR”​. They didn’t stick only to old tactics; they learned new ones to remain effective. The same goes for advertising agencies that re-trained creative staff to produce for YouTube and Facebook, not just TV. We can take a cue from this playbook: be willing to re-skill and broaden expertise. This might mean learning more about UX and conversion optimization (to ensure the traffic you get converts better), or mastering new content formats (audio, video, interactive) to engage users who prefer those formats. It could also mean studying community management or partnership marketing. The underlying principle is to evolve from an “SEO specialist” into an “organic growth strategist.” The value of deep SEO knowledge doesn’t disappear – it becomes one tool in a more extensive toolkit. Just as a seasoned print journalist who learned to excel at podcasting would have an edge (content skills plus new media savvy), an SEO expert who understands the broader digital landscape can distill and create powerful integrated strategies.

In concrete terms, what might this evolution look like? For someone in a professional SEO role today, it might start with shifting how you present your plans and results. Instead of reporting “we improved domain authority and rankings for 50 keywords,” tie your work to bigger-picture goals: e.g., “our content helped generate 500 new email sign-ups and was cited by two industry publications this quarter.” Make SEO a means to an end (the end being business growth, brand visibility, and customer trust), not an end in itself. Advocate for SEO considerations in other departments – for instance, work with UX teams to ensure site speed and mobile friendliness (which help SEO and overall user satisfaction), or with customer support to publish answers to top asked questions (which can become great searchable FAQs). Embed SEO into cross-functional initiatives; you’ll improve those initiatives and reduce the risk of SEO being sidelined.

Here’s Where I’m Going in 2025

As someone who has studied and practiced SEO for almost two decades, arriving in 2025 feels like a turning point. The old tactics still matter, but they are no longer enough. The search landscape has fundamentally shifted – driven by Google tightening its grip on user attention, GenAI altering how people seek information, and an industry that was never really good at defining itself or the value it provided. It’s clear why the ROI of doing “SEO-as-usual” has diminished and why many veterans are frankly disillusioned. But this doesn’t mean that organic marketing is a lost cause, far from it.

What we’re seeing is a transition, not an end. Just as print media’s decline didn’t mean people stopped reading – they just read on different platforms – the decline of easy SEO wins doesn’t mean people stopped seeking information – they just find it differently. Our mission in the field is to adapt to how people and platforms have changed. That means carrying forward the core strengths we developed in SEO (data-driven content decisions, technical savviness, empathy for user queries) and applying them in new, creative ways.

SEO will always have value – there will always be a place for optimizing content to be discovered in search engines. However, its relative role in the mix is now more minor and needs to be supported by a broader strategy. The companies and professionals thriving today treat SEO as a comprehensive approach to organic growth. They are as comfortable discussing podcast strategy or community engagement as discussing title tags. They have moved on from siloed thinking and embraced marketing more holistically.

It’s worth noting that moving beyond does not mean dismissing those who still focus on it, nor is it about blindly jumping on new trends. It’s about pragmatism. The data-backed evidence tells us that user behavior and platforms are shifting; the economics tell us that the returns on SEO alone have dropped; the signs tell us to broaden our perspective. We should heed these signs, but with clear eyes and not hype (but full hearts). I’m not going to “give up on SEO,” but I’m going to move forward.

The state of SEO in 2025 is one of transformation. Those who continue seeing SEO as onsite content, on-page, meta tags, and backlinks will continue to see diminishing returns and frustration. Those who see SEO as part of a bigger puzzle – who focus on delivering value to audiences across channels – will find that organic growth is very much alive; it’s just wearing a different face. I’m an SEO vet ready to move on to this next chapter.

I’m not abandoning search. I’m enlarging the scope to optimize for attention and trust wherever I can earn it. I will do that with a more resilient, creative, and audience-centric approach to organic marketing.