AI search visibility (GEO) is a genuinely new discipline with no long track record, no agreed standards, and — as of 2026 — a large number of providers making claims that outpace what anyone can currently prove. Evaluate a prospective partner on what they can show you live, how honestly they describe measurement, and whether their compliance knowledge is specific to your profession — not on the size of the percentage in their marketing.
Ask to see your own results, live
Any credible provider should be able to run real questions about your practice through ChatGPT, Gemini, and Perplexity in front of you, on the spot, and show you the actual answers — not a canned demo or a generic score. If a provider cannot do this in the first conversation, it is a reasonable signal that their process is thinner than their pitch.
Be skeptical of guaranteed rankings
AI assistants are probabilistic systems that change behaviour without notice or public changelog. No provider can guarantee a specific ranking, citation, or number of new clients, for the same reason no honest SEO agency guarantees a #1 Google ranking. A provider who guarantees outcomes is either not being straightforward with you or does not understand the systems well enough to know why that promise cannot be kept.
Ask exactly how "success" will be measured
The honest answer is some version of: your share of AI recommendations for a defined set of real questions, measured before and after, with evidence. Be cautious of vague claims tied to "traffic" or "leads" with no clear mechanism connecting an AI mention to a tracked outcome — that attribution problem is real, and a credible partner should say so plainly rather than paper over it.
Check whether compliance is a genuine specialty or an afterthought
For regulated professions, this matters more than almost anything else on this list. A generalist marketing agency applying the same playbook to a law firm as to a restaurant is not equipped to know what a legal-advertising rule or a medical-advertising ordinance actually restricts. Ask directly what rules govern your profession's public-facing content, and see whether the answer is specific and confident or vague and reassuring.
A short list of questions worth asking directly
| Question | What a strong answer sounds like |
|---|---|
| "Can you show me what AI says about us right now?" | A live demonstration, on the spot, using your actual name |
| "How will you measure success?" | A specific, falsifiable metric — not vague promises |
| "What happens if it doesn't work?" | An honest answer, not a deflection |
| "What rules govern advertising in our profession?" | A specific, confident answer |
| "Do you test in the languages our clients actually use?" | Yes, with evidence they've done it before |
This is a young category — treat it accordingly
Generative Engine Optimization has existed as a named discipline only since a 2023 research paper introduced the term.1 Best practice is still being established, and a healthy amount of what circulates as advice is closer to folklore than to evidence. That is not a reason to wait — the underlying shift in how people search is real and already underway — but it is a reason to favour partners who are candid about what is known versus assumed, over those who project unearned certainty.