What AI platforms say when you ask them how they choose
Infokus Marketing tested this directly. The same question was put to Gemini, Claude and ChatGPT.
All three answered in the same register. Keep your name, address and phone number consistent across directories. Claim and populate your Google Business Profile. Collect reviews and respond to them. Publish accurate opening hours. Add schema markup so platforms can read your details without guessing.
When you want to understand how AI recommends professional service firms, the examples the platforms reach tell you more than the advice does. Custom home renovations in Toronto. Cheap pizza near me. Best cardiologist nearby. Best for families. Fast service. One platform walked through proximity, availability, whether the business is open now, and menu data.
This is the local search playbook in new clothing. It is not wrong. It is answering for restaurants, trades and clinics, because that is the shape of business the question usually comes from.
You can reproduce all of this in about a minute, which is the point. Run the question yourself before you spend anything on the answer.
Why that advice does not transfer to expertise-led firms
A commercial buyer selecting a law firm, an accounting practice, a management consultancy or an executive search partner is not running a local search. They are making a judgement about whether a specific group of people can be trusted with something consequential.
Proximity is close to irrelevant. Trading hours are not part of the decision. Review volume, which does real work for a restaurant, carries little weight when the purchase is professional judgement, and the client base is small, senior and unlikely to post publicly about it.
This is why so many genuinely excellent firms are absent from AI answers. It is not a quality problem. Infokus Marketing’s view, formed across the six platforms it monitors, is that how AI recommends professional services firms is structurally influenced by one fundamental issue: expertise-led businesses are expertise-rich and entity-poor. The knowledge that makes the firm valuable sits in the judgement of its principals, and almost none of it has been converted into anything a machine can read, verify and cite.
A fifteen-year accounting practice with a thin website, an inconsistent Google Business Profile and no published thinking is invisible to AI. Not because the work is poor. Because there is nothing to cite.
What actually decides whether an expertise-led firm gets named
Four things, in Infokus Marketing’s assessment.
- Expertise in an extractable form. The firm’s knowledge has to exist somewhere as structured, readable content that answers a question directly. Not a services list. Not a capability statement. An actual answer to something a buyer would ask.
- A named practitioner behind it. Platforms weigh content attached to a real person with verifiable credentials. Anonymous corporate authorship is a weaker signal than a named expert with a traceable professional history.
- Independent corroboration. Whether other credible sources describe the firm the same way. For expertise-led businesses, these are industry publications, professional associations, awards bodies and LinkedIn, not consumer review platforms.
- Positioning specific enough to match. A platform will not name a firm it cannot categorise confidently. A practice describing itself as offering bespoke commercial solutions has given the machine nothing to match against. One that says it advises family-owned manufacturers through ownership transition in Victoria has given it everything.
Most of what AI knows about your firm comes from pages you do not own
This is where the standard advice misleads most expensively, because it points almost entirely at your own website.
Analysis by AirOps of 21,311 brand mentions across ChatGPT, Claude and Perplexity found that around 85% originated on domains the brand does not own, with only 13.2% coming from the brand’s own site. Brands earning visibility for high-value commercial queries were roughly 6.5 times more likely to surface through third-party content than through their own pages.
Read that against how most firms respond to an AI visibility problem. They rebuild the website. The website matters, and it is necessary. It is also, on this evidence, the minority of the signal.
The work that moves the needle is unglamorous. Association profiles that are current and complete. Directory entries that agree with each other. Industry publications that quote your people by name. Awards and panels and commentary that put your firm in someone else’s sentence.
Why AI platforms name so few firms
Worth understanding, because it explains why absence is not a small disadvantage.
Platforms are deliberately selective. During Infokus Marketing’s testing, ChatGPT described its own behaviour plainly: assistants recommend very few businesses rather than many because a low-confidence recommendation costs them more than an omission. It also identified the cold-start problem, in which firms without corroborating signals struggle to enter the set at all.
The commercial consequence is blunt. When a buyer asks an AI platform who they should speak to, the firms named are added to the shortlist. Firms not named are not considered, because there is no second page to scroll to. Three names, and the decision starts there.
What matters less than you have been told
This section will be uncomfortable for anyone who has invested heavily in the standard checklist, so it is worth being precise. None of the following is worthless. All of it matters less than the advice implies, when the buyer is purchasing professional judgement.
- Review volume. Valuable for transactional businesses. Rarely decisive when a firm has forty clients, all of them senior, none of them reviewing you publicly.
- Proximity and service-area signals. Close to irrelevant for advisory work that happens over video and travels. Geography still matters as an entity signal, which is a different job from proximity ranking.
- Contact detail consistency beyond accuracy. Get it right once. It is a hygiene factor, not a growth lever. Repeated auditing of it is displacement activity.
- Schema markup on its own. Necessary, and Infokus Marketing specifies it in every engagement. But if 85% of mentions originate off-site, structured data alone cannot carry the outcome. It makes you legible. It does not make you cited.
- Keyword density. Peer-reviewed research presented at KDD 2024 by Aggarwal and colleagues at Princeton and IIT Delhi found keyword stuffing performed around 10% worse than baseline in a live generative engine. Adding citations, credible quotations and statistics performed best.
That last finding leads somewhere more interesting. In the same study, when all sources were optimised at once, citing sources increased visibility for a fifth-ranked site by 115.1%, while the top-ranked site’s visibility fell by 30.3%. The authors explain it directly: traditional search rewards backlinks and domain presence, which small firms cannot buy, whereas generative engines work from the content itself. They describe the effect as democratising, letting independent businesses compete with far larger ones.
If your firm sits outside the top rankings, that is not a handicap in AI search. On the current evidence, it is an advantage, and it will not stay available indefinitely.
Six questions that show where your firm stands
Two minutes, honestly answered.
- Does your firm’s expertise exist anywhere as published content that answers a specific question a buyer would ask?
- Is a named person with verifiable credentials attached to that content?
- Does any independent professional source describe what your firm does, in words you did not write?
- Could a stranger read your positioning and say confidently which situations you are the right call for?
- Do your website, LinkedIn, association listings and directory entries agree with each other?
- Has anyone actually tested what the platforms currently say when asked your buyers’ questions?
A no on questions one, two or three is a structural gap, and it is where the work starts. A no on four is a positioning problem, and no amount of content will fix it. A no on six means every other answer is a guess.
Is your firm in the answer, or not?
The test takes thirty seconds. Open ChatGPT or Perplexity, type the question your best prospective client would type, and read what comes back.
If you are not named, that channel is currently closed to you. If a competitor is named, they are building a position that compounds with every week they hold it.
Infokus Marketing includes an AI visibility context in the Free Marketing Audit: No pitch, and you leave with something useful whether or not anything follows.
