Press slash to focus this field, or Escape to return home.
AEO: what the vendors document, and what they don't
AEOAnswer Engine Optimisation is the work of being the source an assistant uses when it answers. It is a real discipline. It is also sold with a confidence that the documentation does not support, so this page separates the two.
Start with the most inconvenient sentence in the field, from Google's own guidance on AI features:
"There are no additional requirements to appear in AI Overviews or AI Mode, nor other special optimizations necessary."
It goes further:
"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."
If you have been sold an AEO package on the promise of special markup for AI, that is the vendor of the largest answer engine on earth telling you it is not required.
So why is there a discipline at all
Because that sentence is about Google, and Google is one engine. Its eligibility rule is stated just as plainly: a page must be "indexed and eligible to be shown in Google Search with a snippet." AI Overviews are drawn from the Search index, which is built by a crawler that renders JavaScript. For Google, ranking is the gate and the fundamentals are the whole job.
ChatGPT, Claude and Perplexity each run their own crawler and keep their own index. All three publish far less than Google about how they choose. Being in Google's index does not put you in theirs, and that gap is what AEO has to cover.
Three crawlers per vendor
Every major vendor now runs the same three-way split, and the tokens are controlled separately in robots.txt. Only the middle column decides whether you can appear in an answer.
| Vendor | Training | Search index | Live user fetch |
|---|---|---|---|
| OpenAI | GPTBot | OAI-SearchBot | ChatGPT-User |
| Anthropic | ClaudeBot | Claude-SearchBot | Claude-User |
| Perplexity | none stated | PerplexityBot | Perplexity-User |
This is the distinction most robots.txt files miss. OpenAI states it directly: "a webmaster can allow OAI-SearchBot in order to appear in search results while disallowing GPTBot." You can decline to feed the training set and stay findable. A file written to keep AI out usually blocks all three, and the third column is the one that costs you customers.
Perplexity documents its own wrinkle. PerplexityBot is "designed to surface and link websites in search results on Perplexity. It is not used to crawl content for AI foundation models." And of the user fetcher: "this fetcher generally ignores robots.txt rules", because it acts on a person's request.
What none of them document
All four vendors leave the same question unanswered: whether their crawlers execute JavaScript. I read OpenAI's bot documentation end to end to be sure.
The industry's confident claim that they do not render traces back to a single study, from Vercel and MERJ in December 2024, with no published replication since. It is good evidence. It is also a single measurement, and the field now cites it with the confidence it would give a vendor's own documentation. The full trace is here, including the one vendor sentence that does exist and the scope people miss when they quote it.
Nobody publishes ranking factors for AI answers either. There is no documented list of what makes an assistant pick one supplier over another. Anyone selling you the formula has reverse-engineered a guess, and the guess may even be a good one, but it should be labelled.
What that leaves you
Less than is usually sold, and it is all verifiable.
Allow the search crawler for each engine you care about. Ship HTML that carries its meaning before any JavaScript runs, because no vendor promises to run yours. Rank in Google: for the largest answer engine that is the entire eligibility test. Keep your entity data consistent everywhere it appears. Assistants reconcile what they find, and inconsistency is what makes them describe you vaguely. Get corroborated off your own domain, because a claim that appears only on the page making it is the weakest evidence there is.
Structured data still earns its place, for entity clarity and for rich results in Search. Those are the payoffs Google documents.
The questions people ask
Each of these is its own page, with its sources named on it.
- How do I get my business mentioned by ChatGPT?
- Why doesn't ChatGPT know my brand? How do you choose an AEO builder in Dubai?
- How do I check whether AI mentions my brand?
- How do I show up in AI search and AI Overviews?
- How do I know if the LLMs trust my service?
- How is AEO different from SEO?
The measurements behind all of this
Where I have tested something, it is written up with the date and the method.