← Part of the full guide: Why AI Doesn’t Recommend Your Brand
TL;DR: Seven questions to ask before signing with an AI visibility agency: show me a real report, which platforms do you measure, is there a baseline, do you track competitors, do you work off-site, what exactly do you guarantee, and what happens if nothing moves. This page explains what each question is testing, how to tell a good answer from a walk-away answer, and publishes our own answers to all seven.
Why you need a question set
GEO, AEO and AI SEO became sales categories very recently. A number of providers rebranded existing SEO work without building any AI-side measurement, and buyers cannot tell the difference from a website alone. A fixed question set separates them faster than reviewing portfolios.
One useful signal: ask ChatGPT how to choose an AI agency and it generates a list very like this one, every time. These criteria are not our private opinion — they are what a system reading the whole market converges on independently.
Question 1 — Can I see a real redacted report?
Testing for: whether they have shipped work or only built a sales page.
Good answer: a sample file, with before-and-after figures broken out per dimension, a list of work actually delivered, and a section acknowledging limitations.
Walk-away answer: screenshots of rising graphs with no method and no dates, or a claim that every report is fully confidential so nothing can be shown. A report containing only good news did not measure anything.
Question 2 — Which platforms do you measure, and how?
Testing for: whether they understand that each system retrieves from a different pool.
Good answer: named systems — at minimum ChatGPT, Perplexity, Gemini and Google AI Overviews — plus an explanation of how they control session state and inferred location so results are repeatable.
Walk-away answer: “we measure AI” with no systems named, or measuring ChatGPT alone and assuming the rest behave identically. In our own testing, the lists ChatGPT and Perplexity returned for the same question overlapped only partly.
Question 3 — Is there a baseline before work starts?
Testing for: whether they can prove their own contribution later.
Good answer: a baseline captured before anything is touched, using the same question set that will be used afterwards, with the first set of numbers handed to you to keep.
Walk-away answer: remediation starts immediately and measurement happens later. Without a baseline you cannot separate improvement you paid for from platform-side changes in how sources get selected. This question eliminates more providers than any other.
Question 4 — Do you track competitors or only my brand?
Testing for: whether they understand this is a contest for a limited number of seats.
Good answer: every answer is logged with which companies were named and in what order, so you can see who is holding the seat you want.
Walk-away answer: a report stating only how many times your brand appeared, with no record of who else was in the same answer. That number cannot tell you whether you are winning or losing.
Question 5 — On-site only, or off-site too?
Testing for: whether they know where the real blocker usually sits.
Good answer: both — structure and schema on your site, plus business listings, directories and the third-party sources that describe your brand.
Walk-away answer: a proposal describing only article production and schema markup. We proved the limitation on ourselves: 22 published pages across two languages and still not named once, because external corroboration was missing. The full record is in our self-test.
Question 6 — What do you guarantee, and what do you not?
Testing for: honesty about what is outside their control.
Good answer: guarantees covering scope, delivery date and a refund if work is not delivered — combined with an explicit statement that placement in AI answers is not guaranteed.
Walk-away answer: a guarantee that ChatGPT will recommend your brand. There is no paid placement in these answers, so that guarantee sells something which does not exist as a product.
Question 7 — What happens if nothing moves in six months?
Testing for: whether anything ties their fee to outcomes.
Good answer: a review point agreed in advance, a definition of what counts as no progress, and a stated next action — change approach, pause, or say plainly that the blocker sits outside their scope.
Walk-away answer: “this work takes time” with no criteria attached. That means there is no line at which they can be held accountable.
Our answers to all seven
| Question | CiteLogics answer |
|---|---|
| Real report | Available on request, and two client cases publish per-dimension before-and-after figures on the site |
| Platforms measured | Four systems — ChatGPT, Claude, Perplexity, Gemini — with at least 25 questions per business |
| Baseline | Captured before any work, using the identical question set used for later rounds |
| Competitor tracking | Every answer logs which competitors were named and in what order |
| Off-site work | Included — business listings and third-party sources, not only the website |
| Guarantees | Scope, 24-hour delivery on the audit, 7-day money back. No guarantee of AI placement |
| If nothing moves | We publish our own results including a zero, and disclose platform ceilings before you spend more |
The scoring rubric behind our reports is published at the AIVS methodology page, and how to evaluate any quote is covered at our pricing page.
CiteLogics — we answer all seven
The AI Visibility Audit runs your real buying questions across four platforms, captures a baseline first, records which competitors were named instead of you, and scores five AIVS dimensions against a publicly documented rubric.
- AI Visibility Audit — full AIVS Score report plus a 30-minute call. THB 2,000, delivered in 24 hours, 7-day money back.
- Big Win (Audit + WebFix) — audit plus full implementation across crawlability, structure and schema. THB 20,000.
Get your free AI visibility scan, or contact sales@citelogics.com.
Frequently asked questions
What if I ask all seven and still cannot decide?
Have the final two or three vendors measure your site using the same question set, then compare whose report explains its calculation more clearly. That surfaces the difference faster than reading proposals, because you see actual work before signing anything.
Can my existing SEO agency do this work?
Yes, if they have real AI-side measurement. The technical skills overlap heavily. The gap is measurement, because a Google ranking tells you nothing about whether you are named in AI answers. Focus questions two and three on them.
How long a contract should I sign?
Start with one clearly scoped piece of work before committing to a retainer. The first engagement tells you both the state of your site and how the provider works. If a retainer has no review point agreed in advance, ask for one before signing.
Could I do this myself instead of hiring?
The measurement is doable in-house — roughly half a day per round once you have a fixed question set. The technical remediation depends on how much control your team has over code and schema. Many teams measure themselves and hire only for the code-level fixes.
Should I immediately rule out anyone offering guarantees?
Ask what is being guaranteed first. Guaranteeing delivery against an agreed scope is normal and welcome. Guaranteeing that AI will recommend your brand is selling something nobody controls, and that alone is sufficient reason to rule them out.
Written by
Nipun Kasevayuth
Founder & CTO, CiteLogics
An engineer and AI enthusiast who reverse-engineers how AI models decide which brands to cite. He proved the method on his own websites first — then delivered the same results for brands like Coldtubb. He and the CiteLogics team have scanned 1,000+ websites for AI visibility.
Connect on LinkedIn →
