How AI assistants choose the sources they cite

Every cited answer starts with a quiet decision: out of everything on the web, these few pages get named. If your business depends on being one of them, it helps to know what the platforms themselves document about how that decision works — and where guesswork begins.

The answer is assembled from search results

AI assistants do not know your market from memory alone. OpenAI's documentation describes its web search tool as giving models access to current information from the internet, so they can answer "with sourced citations"¹. In the simplest setup, the documentation says, "the non-reasoning model sends the user's query to the web search tool, which returns the response based on top results"¹. The answer you see is built on whatever pages that retrieval step surfaced — which is why a page that never gets retrieved can never get cited.

Citations are a shortlist, not a bibliography

Being retrieved is not the same as being cited. In OpenAI's web search, answers carry inline citations by default, and the documentation adds that "the number of sources is often greater than the number of citations"¹. In other words, the system reads a wider set of pages than it names: the cited few are the ones that actually carried the answer. That is the shortlist you are competing for.

Each platform decides differently

There is no single algorithm to optimize for. Google documents that its two AI search features can run on different models and techniques, so their answers and links will differ³ — and that is within one company's products. On the ChatGPT side, a dedicated crawler, OAI-SearchBot, is used to surface websites in ChatGPT's search features². Different retrieval systems, different crawlers, different answers.

What the platforms agree on is reassuringly boring. Google states that appearing in AI Overviews or AI Mode has no extra requirements, and that ordinary SEO best practices still apply³. OpenAI recommends allowing OAI-SearchBot in your robots.txt file to help ensure your site appears in search results².

What this means for your site

Three practical consequences:

1. Be retrievable. If a crawler is blocked or your content is invisible to it, nothing downstream can happen.
2. Be the page that answers the question. Retrieval starts from the buyer's query, not from your brand name. A page that directly answers a buyer's question has a reason to be on the shortlist; a page about your product vocabulary does not.
3. Be named where answers already look. The cited pages are not always yours — comparisons, directories, documentation. If those pages mention a competitor and not you, the answer follows them.

OLAB Assist works on exactly this chain: it records the AI answers to your questions, opens the pages those answers cite, and flags the ones where you are missing — so you can see, per question, which link in the chain is failing.

See your own shortlist

Start on the free plan, one site for 7 days with no card needed, or try any paid plan free for 7 days. Add the questions your buyers ask and read the actual citations behind the answers.