Answer engine optimization: what Google actually says

"Answer engine optimization" and "generative engine optimization" are sold as new disciplines with new techniques. Google publishes documentation on this, and it is unusually blunt: there is no separate lever. That does not make the work pointless — it makes it ordinary, and the ordinary version is checkable.

The confusion that costs teams real visibility is not about content style. It is about crawler tokens: which switch governs which surface. Several of them do the opposite of what their name suggests.

What is actually established

Quoted from Google's and the model providers' own documentation.

There is no special optimization. Google: "There are no additional requirements to appear in AI Overviews or AI Mode, nor other special optimizations necessary."
Eligibility is ordinary search eligibility. "To be eligible to be shown as a supporting link in AI Overviews or AI Mode, a page must be indexed and eligible to be shown in Google Search with a snippet."
No AI-specific files or markup. "You don't need to create new machine readable files, AI text files, or markup to appear in these features. There's also no special schema.org structured data that you need to add." Google's optimization guide states it ignores llms.txt entirely — it neither harms nor helps.
Google names the tactics that do not work: breaking content into tiny pieces for AI, writing in a specific way just for generative AI search, and building inauthentic mentions across the web.
`nosnippet` is the real off switch for answer inclusion. It stops the text snippet *and* "will also prevent the content from being used as a direct input for AI Overviews and AI Mode." max-snippet limits how much may be used; data-nosnippet scopes it to specific elements.
Google-Extended does not remove you from AI answers. It governs whether crawled content trains future Gemini models, and Google states it "does not impact a site's inclusion in Google Search nor is it used as a ranking signal." It has no separate user-agent string — it exists only as a robots.txt token.
OpenAI's tokens are independent. OAI-SearchBot surfaces sites in ChatGPT search; GPTBot feeds foundation models; ChatGPT-User is user-triggered. OpenAI states a site can allow OAI-SearchBot to appear in search while disallowing GPTBot.
Perplexity's user fetches ignore robots.txt. PerplexityBot respects robots.txt and is not used for foundation-model training, but Perplexity-User is user-triggered and "generally ignores robots.txt rules."
Structured data still earns nothing by itself. "Google does not guarantee that your structured data will show up in search results, even if your page is marked up correctly."

How to run it

What is left, once the invented techniques are removed, is a short and unglamorous list.

Confirm the page is indexed and snippet-eligible. That is the stated entry condition, and it is the one most often broken by an inherited noindex or an over-broad robots rule.
Audit your snippet directives before writing any content. A nosnippet or a tight max-snippet anywhere on a template silently excludes those pages from AI Overviews and AI Mode.
Decide the crawler policy per surface, in writing: search inclusion, model training, and user-triggered fetches are three separate decisions with three separate tokens.
Write the page for a reader who needs the answer, and keep the answer on the page rather than behind an interaction. Google's own priorities list unique, non-commodity content, page experience, and technical requirements.
Use structured data where it maps to a real rich result, and make sure it matches visible content. Do not add it expecting an AI-answer effect.
For commerce and local, keep Merchant Center feeds and Google Business Profiles current — Google names these as inputs that surface products and services in AI responses.
Re-check after any migration or CMS change. Snippet directives and robots rules are exactly the settings that get restored to a default nobody chose.

How to measure it

Measure the surfaces the platforms report on, and be honest that the click is often not there to count.

AI-feature impressions from Search Console's generative AI performance report, by page.
Citations and Citation Share from Bing Webmaster Tools, which is the only query-level share metric published by a platform.
Indexed and snippet-eligible coverage as a hygiene metric — the entry condition, tracked as a number.
Visit quality rather than raw clicks. Google's own guidance on succeeding in AI search names this shift explicitly.
Branded search and direct traffic, where exposure without a click tends to land.

Where teams get this wrong

Buying an llms.txt implementation for Google visibility. Google says it ignores the file.
Blocking Google-Extended believing it removes you from AI Overviews. It governs Gemini training and explicitly does not affect Search inclusion.
Shipping a nosnippet for brand-safety reasons without realising it also removes the page from AI Overviews and AI Mode.
Chopping pages into fragments "so AI can parse them" — a tactic Google names as ineffective.
Assuming a robots.txt block stops Perplexity entirely; user-triggered fetches generally ignore it.
Buying mentions to look authoritative. Google names inauthentic mentions as a tactic that does not work.

Sources

Answer engine optimization: common questions

Is AEO different from SEO?

Per Google's documentation, not in terms of levers: eligibility for AI Overviews and AI Mode is ordinary indexing plus snippet eligibility, with no additional requirements and no special markup. What changes is measurement, because a cited answer often produces no click.

Should we publish an llms.txt file?

Google states it ignores llms.txt and that the file neither harms nor helps visibility or rankings in Google Search. Other engines may differ, but do not expect a Google effect.

How do we stop our content appearing in AI answers?

On Google, nosnippet prevents content being used as a direct input for AI Overviews and AI Mode, and there is a separate Search generative AI control. Blocking Google-Extended is not the switch — that governs Gemini training only.

Does structured data help us get cited?

Google states there is no special schema.org markup for generative AI features and that structured data is not required for them. It remains worth adding where it maps to a rich result and matches visible content.