Blog · Foundations

Who owns AEO inside a company

Published 11 September 2026 · Updated 12 October 2026 · 7 min read

Who owns AEO inside a company
Photograph: Igor Omilaev / unsplash

The short answer: One named person should be accountable, and in most companies that is whoever already owns organic search. They cannot do the work alone, because the controls that decide whether an engine can reach you, and repeat you correctly, sit with three different teams: infrastructure holds crawl access, engineering holds rendering and markup, and brand or PR holds everything said about you elsewhere. Accountability is singular. Execution is not.

That split is the whole problem, and it is why the question keeps getting asked.

What one 2026 survey found

One self-selected industry survey, 599 professionals in SEO, content and growth roles, published 19 June 2026, found that 52.8 per cent of respondents put AI search under their existing SEO team. Another 19.4 per cent reported a cross-functional effort across SEO, content and engineering or product. Only 9.2 per cent reported a dedicated specialist or team, and 2.7 per cent outsourced to an agency. Among respondents at small companies, 13.1 per cent reported no clear owner at all.

Those are self-reported answers from people who chose to take a survey about AEO, and the sample skews to consultants, freelancers and companies of one to ten people. It is not a census of the market, and its own authors call the findings directional. Read it as a snapshot of one practitioner population: half of them put the work where the skills are, almost none of them have given it a dedicated head, and a sixth of the smallest companies have nobody holding it. The last group is the dangerous one, because an unowned programme is not a slow programme. It is a programme where nobody notices the day a robots.txt edit removes the site from one engine's answers.

The three switches, and who holds each one

Ownership becomes concrete the moment you list the controls instead of the aspirations. The named controls below are Google's, because Google documents them. Other engines publish much less about what a site owner can set, so treat the pattern as a way to find your own gaps and not as a rule you can assume holds elsewhere.

Crawl access sits with infrastructure. Google's own documentation is blunt about this: robots.txt directives for Googlebot are the control for how a site is crawled for Search, and AI Overviews and AI Mode draw on that same crawl. The Google-Extended token, which a lot of legal teams reached for in 2023 and 2024, governs whether crawled content trains Gemini models and grounds Gemini Apps and Vertex AI, and Google states plainly that it does not affect inclusion in Google Search. Two teams can therefore both be wrong at once: legal thinks it has opted out and has not, while someone else blocks Googlebot to stop AI Overviews and quietly leaves Search entirely.

Rendering and markup sit with engineering. Whether the page arrives with content in the HTML, whether a data-nosnippet wrapper is swallowing the paragraph that answers the question, whether max-snippet is clamped to a length too short to be useful. Google's AI features page lists nosnippet, data-nosnippet, max-snippet and noindex as the levers that limit what gets shown. None of them live in a marketing tool. Each is one pull request away from changing what Google is allowed to show.

Corroboration sits with brand, PR and sometimes sales. No single team owns what third parties say, which is exactly why it drifts. This is the part that takes longest and the part a dashboard cannot fix, and it connects directly to entity resolution: if your company name resolves to two different descriptions across your own properties, an engine has less to agree with, and in buyer-facing answers a hedge reads much like absence.

Why the SEO team is usually the right accountable owner

They already hold the crawl relationship, the structured data backlog and the standing arguments with infrastructure about bot rules. They know how to run a change, wait, and re-measure. Google's July 2026 optimisation guide reframes the whole category as continuous with search fundamentals rather than a separate discipline, and explicitly lists things you do not need: a new machine readable file such as llms.txt, chunking your pages into small pieces, rewriting in a special style for AI systems, or chasing inauthentic mentions. That guidance is Google's, and the page scopes it to Google Search. It is not a statement about how ChatGPT, Perplexity, Copilot or Claude behave, and none of them publish an equivalent list.

Where that framing misleads people is in the next step. The work is continuous with search, yes. The reporting is not. A rank is a position you can defend; an answer is a sentence that can come back differently each time someone asks, which is the argument in AI search is not a ranking game. If you give the programme to a team whose targets are all positions and sessions, they will optimise for the thing they are measured on, and the mention with the correct claim attached will stay nobody's job.

So pair the owner with a scorecard they do not already have. Named or not named, claim correct or wrong, which sources the engine leaned on. Three columns, run repeatedly.

The failure modes, in the order we see them

The silent block. A WAF rule, a bot-management vendor default, or a rate limit added during an incident. Usually nothing in the marketing stack reports it. In client work we have seen this survive for months, because the team that owns the rule and the team that feels the loss never share a ticket queue.

The two descriptions problem. The homepage says one thing about what you sell, the schema says a second, the funding announcement says a third. Each is owned by a different person and all three are defensible in isolation.

The orphaned opt-out. Someone on legal blocks an AI crawler to protect content, which is a legitimate business call, and nobody tells the team whose numbers will move. This is also the case we can say least about from outside: we cannot see what a model already holds about a brand, so we cannot tell you in advance what a given opt-out will cost you.

The approval tax. In a large organisation, the fix is found in week one and shipped in week eleven. Nothing is wrong with the analysis. The path is long, which is a different problem and gets its own treatment in AEO inside an enterprise.

A check you can run this week

This takes about an hour and it tells you whether you have an owner or a hope.

  1. Open your own robots.txt in a browser. Note every AI and search user agent named there,

    2. and what each is allowed.

  2. Find out who can merge a change to that file. Name the person, not the team.
  3. Ask that person when it last changed and why. If nobody knows, you have found your gap.
  4. Pull the three sentences that describe your company from your homepage, your

    5. Organization schema and your most recent press coverage. Put them side by side.

  5. Ask five buyer questions on two engines and record, for each, whether you were named and

    6. whether the claim was true.

If steps two and three have no answer, the fix is not a tool. It is a line in somebody's objectives. The self-check extends this to ten tests you can run without any tooling, and the broader framing of what the discipline covers sits in what answer engine optimisation is.

What we are still measuring

Whether a dedicated owner outperforms a well-wired SEO owner is unresolved, and nothing here settles it. The survey records what a self-selected set of practitioners chose, not what worked, and we have found no public dataset tracking outcomes against reporting structure. The three switches are drawn from Google's documentation, so the Google claims are sourced and the equivalents at other engines are inference, not evidence. Our own failure modes come from client engagements rather than a counted sample: we have not kept a structured record of org chart against result, so we cannot put a rate on any of them, and the absence of a public number is not evidence that the effect is small.

FAQ

Should AEO sit under SEO or under brand? Accountability under whoever owns organic search, with a named deputy in brand for the corroboration layer. Splitting accountability across both tends to mean neither escalates when something breaks.

Do we need to hire an AEO specialist? Not to start. In the June 2026 survey cited below, 9.2 per cent of respondents reported a dedicated specialist or team. A specialist earns their seat when the manual measurement loop gets too expensive to repeat, not before.

Who owns the technical fixes? Engineering ships them, the accountable owner writes them up with evidence and sequences them. The practical blocker is usually priority, not capability, so the owner needs a route into the sprint planning conversation rather than a document.

What does an audit change about ownership? It gives the arguments a shared artefact. The deliverable is the report, scoped to the client, with the evidence attached to each finding so an infrastructure lead can check the claim rather than take it on trust. The sample audit report shows the shape of one.

Sources

All four sources were opened and checked on 12 October 2026.

Related reading

What is answer engine optimisationAI search is not a ranking game, and what replaces itAEO inside an enterprise, where approval takes longerEntity resolution: making engines sure which company you areAEO vs GEO vs SEO: what the three terms actually meanAn AI search glossary that defines the terms properly