How AI decides which businesses to name.

What actually happens between a question and an answer?

Both systems expand one human question into several machine queries, retrieve candidates from an index, then synthesise an answer and cite some of what they found. Google describes a "query fan-out" technique that it says lets these features "display a wider and more diverse set of helpful links." OpenAI's ChatGPT search "typically rewrites your query into one or more targeted queries" before sending them to search providers. Different vocabulary, same shape: expand, retrieve, synthesise.

That shared shape matters more than it looks. It means the retrieval layer is still an index a page has to be in, and the query a page needs to answer is rarely the exact sentence the user typed. It is one of several expanded versions of it.

Why do OpenAI's three crawlers get confused with each other?

OpenAI runs three separate bots and they do three different jobs; blocking the wrong one has no effect and blocking the right one removes a business from ChatGPT search entirely. Most site owners manage one robots.txt rule and assume it covers "AI." It does not.

• OAI-SearchBot is "used to surface websites in search results in ChatGPT's search features." OpenAI says sites that block it "will not be shown in ChatGPT search answers." This is the one that governs whether you can be cited.
• GPTBot is "used to make our generative AI foundation models more useful and safe." OpenAI states that "disallowing GPTBot indicates a site's content should not be used in training." Blocking it opts a site out of training data. It does not touch search citation.
• ChatGPT-User handles certain user-initiated actions and is "not used to determine whether content may appear in Search."

The practical consequence: a business that blocks GPTBot to stay out of model training keeps its ChatGPT search eligibility intact, as long as OAI-SearchBot still has access. Conflating the two is the single most common and most fixable mistake in this area.

Blocking GPTBot opts you out of training. Blocking OAI-SearchBot opts you out of being found.

What does "being retrievable" mean if it isn't a separate AI discipline?

Google is direct about this: "There are no additional requirements to appear in AI Overviews or AI Mode, nor other special optimizations necessary," and "you don't need to create new machine readable files, AI text files, or markup to appear in these features." Its stated guidance is to apply the same foundational SEO practices used for regular Search.

What changes is not the requirement, it is the target. Because a question gets fanned out or rewritten into several queries, a page only written to answer the literal, exact phrase a person typed is optimised for one query out of several. Being retrievable means the page also holds the plain-language variants and adjacent phrasings a query expansion is likely to produce, and holds one self-contained passage clear enough to be lifted whole into an answer. That is a content and structure discipline, not a new machine-readable file.

Google is also explicit that meeting requirements is not a guarantee: "Just because a page meets all requirements, best practices, and complies with the policies, doesn't mean that Google will crawl, index, or serve its content." OpenAI's help documentation says the same about ranking generally—there is "no way to guarantee top placement." Neither platform promises inclusion. Both only describe the mechanism.

What should a business actually do with this?

Treat "AI visibility" as the same SEO discipline with two additions: crawler-specific permissions and query-expansion awareness, not a parallel set of rules. Nobody controls Google or an AI platform, and no source here claims otherwise. What is checkable is whether the right bots can reach the page, whether the page answers the questions people actually ask rather than the one exact phrase, and whether there is a passage clear enough to be cited on its own.