Last updated: 15 July 2026
The short version
A widely shared post argued that OpenAI and Anthropic "don't care about GEO or AEO", because their search hires carry the title SEO Lead, not AI Search Lead. The conclusion many readers took: generative engine optimisation is marketing, so ignore it.
That conclusion is wrong, and the job posting used as evidence quietly proves it.
We sell GEO for a living, and we still agree that the acronyms are marketing. What is not marketing is the work, and the way you now have to measure it. Here is the honest version, with the numbers.
Schmitdy is the AI-search team at AI Heroes. We track how brands show up inside ChatGPT, Claude, Perplexity, Google AI Overviews, and Copilot, we read those engines every working day, and we put half of our fee on the outcome. So when someone says the category is fake, it lands on our desk.
Did Anthropic really leave GEO out of the job?
No. Read the posting. Anthropic's SEO Lead role, live in 2026, lists two responsibilities in plain language: "Monitor and analyze performance in AI-powered search environments, including AI Overviews and other answer engines," and "Optimize content structure and markup for AI Overviews and other LLM-powered search experiences." Under the strong-candidate section it asks for "Understanding of Answer Engine Optimization (AEO) and Generative Engine Optimization (GEO) strategies for LLM-powered search."
So the company held up as proof that GEO is a fantasy wrote AEO and GEO into the brief, by name. What Anthropic did not do is build a separate priesthood for it. They folded the work into one senior SEO hire. That is a real signal, and it is close to the opposite of "they don't care."
Then why don't the labs make a fuss about GEO?
Because OpenAI and Anthropic are the answer engines. You do not optimise to be cited by a system you own and operate. When ChatGPT writes an answer, OpenAI is not trying to get quoted in it. Everyone else is.
The rest of us are trying to be the source an answer is built from, inside engines we do not control. Taking your playbook from the two companies that run the engines is like taking racing tips from the crew that pours the track. Their incentives are not your incentives, and their in-house titles are a poor template for your growth plan.
Is GEO just SEO with a new name?
Mostly the same craft, with different targets and a different scoreboard. The fundamentals a good SEO already respects, fast crawlable pages, clear structure, genuine authority, and content that answers the question, still decide most of the game. Nobody needs to relearn their job.
What changes is what you optimise toward, which signals carry weight, and how you know it worked. Here is the split we actually work to:
| The question | Classic SEO | AI search work |
|---|---|---|
| The goal | Rank in the ten blue links | Be the source the answer is generated from |
| Primary authority signal | Backlinks | Brand mentions across independent sources |
| Winning content | Comprehensive pages that build up | Self-contained answers a model can lift |
| Where you measure | Search Console and rank trackers | Across engines you do not own, prompt by prompt |
| What winning looks like | A click | Often a citation with no click at all |
Where do SEO and AI search actually diverge? (with numbers)
Ranking is not the same as being cited. In a March 2026 study of 10 sites and roughly 150,000 pages, the top ten organic pages captured 55% of organic sessions but only 29% of sessions from LLMs, and 49 of the top 100 organic pages received no AI traffic at all. Ranking on page one no longer guarantees you get quoted.
Brand mentions now outweigh backlinks. Ahrefs studied 75,000 brands across 76 million AI Overviews and found branded web mentions correlated 0.664 with being cited, against 0.218 for backlinks. YouTube mentions scored 0.737. The single most trusted signal in classic SEO matters roughly three times less than plain brand presence when an engine decides who to quote.
Content shape decides your odds. In the same traffic study, trends and analysis pieces earned AI citations 78% of the time, data-led posts 61%, and how-to content just 12%. A page can rank beautifully and still be the wrong shape to be pulled into an answer.
Optimising for extraction is measurable. The peer-reviewed study that coined the term generative engine optimisation (Princeton and Georgia Tech, 2024) found that adding sources, statistics, and clear framing to a page lifted its share of an AI answer by 30 to 40%, in a controlled test across roughly 10,000 queries. That is a real, repeatable intervention, not vibes.
You get cited without the click. 14% of the AI-receiving pages in that study got no organic clicks at all in the window. You can be the source a buyer acts on and never see the visit in your analytics. That breaks the one number SEO taught everyone to trust.
What does being "the source an answer is built from" look like?
Picture a buyer who asks ChatGPT, "who are the best AI automation agencies for UK accountants?" There is no ten blue links page. There is one paragraph, naming three or four firms, assembled from whatever the model trusts about each of them.
Classic SEO gets you ranked for that phrase. It does nothing to make sure you are one of the names in the paragraph. Being in the paragraph comes from being talked about in the places the model leans on, having a page that states plainly what you do for that exact buyer, and being consistent enough across the web that the model treats you as a known entity. That is the work. It is unglamorous, and it is not the same as chasing a rank.
If it is not a new discipline, what actually changes for you?
The playbook rhymes with SEO. The scoreboard does not. Four things sit on your plate now that did not in 2019:
- Measure across engines you do not own. There is no single console. You watch ChatGPT, Claude, Perplexity, Google AI Overviews, and Copilot separately, prompt by prompt, because each cites differently.
- Earn mentions, not just links. The engines lean on where you are talked about: Reddit, YouTube, industry press, review sites. Broad, credible mentions beat a big link profile with no conversation around it.
- Write for extraction. Put the clear, attributable answer at the top of the page, not three scrolls down. Make it easy to lift one clean sentence.
- Accept invisible wins. Some of your best content will move buyers through an AI answer and never register a click. Report for that, or you will kill your most effective pages by accident.
What should a marketing team do on Monday?
You do not need a new budget line or a new acronym on the org chart. You need four habits added to the SEO you already run.
- Pick ten buyer questions and watch them. Ask ChatGPT, Claude, Perplexity, and Google AI Mode the questions your buyers actually ask, once a week, and record who gets named. That is your baseline. Without it you are guessing.
- Go and get mentioned. Answer real questions on Reddit, get onto a podcast or a YouTube channel your buyers watch, earn a line in an industry roundup. Diversity of mention moves the needle more than another guest post for a link.
- Rewrite your top pages to answer first. Lead each key page with one plain, quotable sentence that says what you do and who for. Bury it and a model will skip you for a competitor who said it cleanly.
- Split your reporting. Stop judging AI-influenced pages on clicks alone. Track citations and assisted conversions next to sessions, so an invisible win still shows up as a win.
None of that is exotic. It is the same instinct good SEO always had, pointed at a scoreboard that finally looks different.
So who is right?
Both, partly. The critics are right that GEO and AEO are marketing words, and that anyone selling "AI search optimisation" as a mystical new discipline, detached from SEO fundamentals, is selling you fog. We would not buy it either.
Google says as much in its own words. Its 2026 developer guidance calls optimising for AI search "optimizing for the search experience, and thus still SEO." When the largest search engine on earth agrees the label is redundant, we are not going to fight for the label.
They are wrong that the work is identical. The targets moved, the authority signals reweighted, and the scoreboard changed shape. Call it GEO, call it modern SEO, call it Kevin. The label is marketing. The measurable difference in what gets you chosen is not.
Anthropic settled this quietly in its own job posting: one SEO hire, AEO and GEO written into the brief. Not a new religion. Not a hoax either. A workstream inside modern search, with its own scoreboard.
What we actually do about it
We do not sell you the acronym. We make you the source an answer is built from, earn the mentions that get you cited, shape pages so a model can quote them, and measure the whole thing across five engines every working day so you can watch it move. Half our fee rides on that outcome. If the category were fake, that would be a very short business.
