How do AI answer engines decide which businesses to recommend?
The engines do not publish a formula. What is documented is the pipeline: their crawlers find and index pages, the engine retrieves pages relevant to the question, and it names the businesses those pages clearly and consistently support. Being readable, specific and consistent across sources is what you control.
Is there a published ranking formula?
No. None of OpenAI, Anthropic, Google or Perplexity publishes how it chooses which businesses to name, and anyone who claims to know the weights is guessing. What each documents is the part before the choice: how pages get into the system.
OpenAI describes OAI-SearchBot as the crawler for ChatGPT search (OpenAI). Perplexity says PerplexityBot exists to surface and link websites in its search results (Perplexity). Anthropic lists Claude-SearchBot as crawling to improve search result quality (Anthropic). A page that is blocked is not in the running.
What happens when someone asks for a recommendation?
In outline, three things. The engine turns the question into searches. It retrieves pages that look relevant. Then it writes an answer from what those pages say, naming the businesses the pages support. Some questions are answered from the model's training instead, with no search at all.
For Google, the AI features draw on pages that are indexed and eligible to appear in Search with a snippet, and Google states there are no additional technical requirements (Google). It also says indexing and serving are not guaranteed.
What makes one business get named over another?
The engines do not explain it, so this is what the evidence supports and what is inference.
Documented: the page must be reachable and indexed, and Google says it favors helpful, reliable, people-first content, with trust the most important part of E-E-A-T (Google).
Reasonable inference: a business is easier to name when pages state clearly what it does, where and for whom, when several independent sources say the same thing, and when the page matches the question's wording.
How can you check how it plays out for you?
The only way to know is to ask. Run the questions your buyers ask on each engine, several times, and record which businesses are named and which pages are cited. The answers vary between runs, so one result is an anecdote and a set is evidence.
Why do different engines give different lists?
Each engine has its own crawler, its own index and its own way of turning a question into searches. OpenAI, Perplexity and Anthropic each document separate bots for search and for training, which means your site can be open to one and closed to another without you meaning it. Add to that the fact that answers are generated fresh each time, and you get lists that differ between engines and between runs. The practical lesson is to measure each engine separately and to repeat questions. A business that appears in two of four engines has a different problem from one that appears in none.
Frequently asked questions
Do engines use the same sources?
No. Each has its own crawler, index and way of searching, which is why you can be named in one and missing in another.
Does a higher Google ranking guarantee an AI mention?
No. Google says indexing and serving in AI features are not guaranteed, and other engines use their own systems.
Can I pay to be recommended?
We know of no documented way to pay an answer engine for a recommendation in its answers. Treat anyone selling that with caution.
Why do I get named one day and not the next?
Answers are generated each time and vary. That is the "Unstable" finding in Sightline.