How answer engines decide who to cite

When you ask an AI a question, it usually does not answer from memory alone. For anything current or specific, it runs a search, pulls a set of pages, and writes its reply from what it just read. The sources that make it into that reply are the ones that were easy to find, easy to trust, and easy to quote.

Easy to find means you were crawlable and you actually ranked for the underlying search the model ran. The engines still lean on conventional retrieval under the hood, so being invisible to search usually means being invisible to the answer.

Easy to trust is murkier but recognisable. Models favour sources that agree with each other. If three reputable pages describe you the same way, a model will repeat that description with confidence. If your own site says one thing and everyone else says another, it hedges or leaves you out.

Easy to quote is the part most brands ignore. A page that states a claim plainly, with a number or a clear sentence, gives the model something to lift. A page that buries the point under throat-clearing gives it nothing, so it moves on to a competitor who said it straight.

You cannot see the model's weights, and you do not need to. Make yourself findable, consistent, and quotable, and you line up with how these systems actually pick.