Recognition
The system must resolve your name to a known entity — not a keyword.
Observable mechanics
How AI answers choose which businesses to mention
Nobody outside the labs knows the exact formula. But the observable mechanics are public: AI systems mention businesses they can recognize as entities, corroborated by consistent facts across independent sources.
The mechanics, in four lines
Corroboration
The same facts, stated identically, across independent sources.
Structure
Machine-readable data removes guesswork about what each fact is.
Honest limits
Exact weightings are proprietary. This is mechanics, not the formula.
Recognition comes before recommendation
Before an AI system can recommend a business, it has to recognize the business as a distinct entity — a thing in its model of the world, not a string of characters. Common names collide: HAVENGAI gets confused with HeyGen and Haven AI, and we documented exactly how that happens. Distinct, consistent identity is what separates an entity from a near-miss.
Corroboration is how machines trust
One website saying you exist is a claim. Five independent sources agreeing on your name, address, phone number, and URL is evidence. AI systems — like search engines before them — weigh agreement across sources. That is why the same facts must be stated identically everywhere: one variant spelling is a crack in the evidence. How the corroboration web works is covered in social profiles as entity corroboration.
Structure removes guesswork
Schema.org markup — Organization, LocalBusiness, FAQPage — tells machines exactly what each fact is, instead of making them infer it from prose. A phone number inside telephone schema is a fact; a phone number in a paragraph is a guess. Machines prefer facts.
What nobody can sell you
The exact weighting inside any AI system is proprietary, undocumented, and changes constantly. Anyone selling you the formula — or a guaranteed mention — is selling fiction. There is no legitimate channel that sells AI-answer placement. What you can do is supply the raw material every formula needs: a recognizable, corroborated, well-structured entity.
The two clocks
New information reaches AI systems on two clocks. Browsing models can find a crawlable, well-structured site within days. Training-data models absorb the web in bulk, months at a time — a new business gets found by the first clock long before it is remembered by the second. Both clocks reward the same thing: real, consistent, structured evidence.
Your next move
Give the machines something real to remember.
Direct answers
What serious buyers ask.
Why does AI mention my competitor but not me?
The observable reasons: they may already be a recognized, corroborated entity and you may not be yet. The exact internal cause is proprietary — anyone certain about it is guessing.
Can I pay to appear in AI answers?
No. There is no legitimate channel that sells AI-answer placement.
How long does it take to become mentionable?
Browsing models: days after you’re crawlable and corroborated. Training-data models: months. Both reward the same raw material.
What’s the first concrete step?
A facts page stating checkable claims, identical name-address-phone everywhere, and Organization schema on your homepage. The small-business AI visibility checklist walks through all of it.