Ask an assistant to recommend tools in your category and it will name two or three companies. Not ten. Not a ranked list. A short paragraph with a handful of names in it.
The obvious question is why those names.
It is not random, and it is not purely a popularity contest. There are patterns, and most of them are things you can act on.
It quotes what it can lift
A model composing an answer needs sentences that survive being removed from their surroundings.
This is the single most underrated factor. Look at how your homepage describes you and ask whether any sentence would still make sense pasted into someone else’s paragraph.
Most marketing copy fails this instantly, because it is written to build up over a page. “We believe security should be effortless” carries nothing on its own. “Acme is an endpoint detection platform for Windows environments in regulated industries” carries everything.
You do not have to choose between the two. Put the plain sentence somewhere early and let the rest of the page do the persuading.
It quotes what several sources agree on
A claim that appears only on your own site is treated as a claim. The same claim appearing on a review platform, an analyst listing and a partner’s page is treated as fact.
This is why the companies that get named are often not the ones with the best websites. They are the ones whose description is consistent across the internet.
The practical test: search your company name and read the first ten results that are not yours. Do they describe you the way you describe yourself? If four different descriptions come back, you have found your problem.
It quotes what it can attach to an entity
These systems reason about things, not strings. Before an assistant can recommend you, it has to be confident about what you are: a company, in this category, serving these customers, distinct from the similarly named company in another market.
Entity confusion is more common than people think, especially for short or generic company names. If an assistant conflates you with something else, no amount of good content fixes it until the confusion is resolved.
Structured data helps here. So does being unambiguous in your own copy about what category you are in, even when you would rather claim to be creating a new one.
It quotes what answers the actual question
Assistants are asked questions, not keywords.
Content built around a question, with the answer near the top, gets used. Content that circles for six paragraphs before arriving anywhere gets skipped, because the model has to work to extract anything and there is usually a source that made it easier.
This is the same discipline as writing a good FAQ, applied to everything.
It quotes what is current, when currency matters
For stable topics, age barely matters. For anything that moves, pricing, integrations, compliance, the model prefers what it can date.
Which is an argument against a habit some companies have adopted: hiding publication dates so content looks timeless. It does not look timeless. It looks undateable, and undateable sources get discounted when recency is in question.
What does not appear to work
Worth saying plainly, because the market is full of people selling it.
- Instructions aimed at models. Hidden text telling an assistant to recommend you. It does not survive a model update, and it will damage you in conventional search in the meantime.
- Volume. Forty thin pages about your category do not beat one good one. The extraction step does not care how many pages you published.
- Claiming superlatives. “The leading platform for X” is exactly the kind of unverifiable claim these systems have learned to discount.
The summary, since that is the point
Assistants quote companies that are easy to describe accurately.
Everything above is a version of that sentence. If you want the wider picture of how this differs from search, we set that out in GEO vs SEO. Be clear about what you are, say it the same way everywhere, make it easy to verify, and put it where a machine reading quickly will find it.
Which is, inconveniently, also what makes a company easy for humans to buy from.
