AI search intelligence
Find the buyer questions where AI recommends your competitors instead of you.
We measure recommendation, citation and positioning across ChatGPT, Gemini and Perplexity — then show which losses are commercially important, what the evidence around them looks like, and what to do next.
Founding pilot pricing while we validate. 14-day pilot, founder-run, cancel any time.
Example analysis · one buyer question
Illustrative example“What are the best AI visibility tools for a small SEO agency?”
- ChatGPT (OpenAI API, web search)
- Gemini (Google Search grounding)
- Perplexity
- Recommendations · citations · competitors · evidence
AI visibility is not one number
An answer puts your brand in one of three positions, and they need completely different responses. A single percentage hides which one you are in.
Mentioned
You are in the text. Worth the least — the same sentence can name you and send the buyer somewhere else.
Cited
One of your pages was used as a source. The state you can move most directly.
Recommended
The one that tracks revenue. The answer puts you forward as the choice, not as an also-ran in a list.
Mention≠Recommendation
If a buyer asks for the best tools in a category and you appear only in a caveat while a competitor is put forward, counting both as “mentions” hides the commercial outcome. We keep mention status and recommendation status apart, because they imply different work.
1 · Measure the questions that precede a purchase
A frozen set of 10–20 buyer questions, sent to all three engines repeatedly. Every raw answer is kept, so a better detection rule can re-score history instead of re-purchasing it.
What a measured question looks like
Illustrative example“Which AI visibility platform is best for tracking competitor recommendations?”
2 · Separate what matters from what is fixable
Two different questions, scored separately. A gap can be commercially critical and still have no action priority — and burying that is how agencies end up promising work they cannot deliver.
“Best [category] tools for agencies”
Competitor wins“Most affordable [category] platform”
Mentioned only“[Category] with SOC 2 compliance”
Competitor winsCommercial importance≠Fixability
Collapsing the two either buries the hard problems or pretends everything is addressable. Keeping them apart is why a report can say “this is your biggest loss and we do not yet know how to move it” — a more useful sentence than a confident wrong one.
3 · Diagnose without guessing
Four evidence categories, deliberately never merged. Most tools collapse them and end up recommending that a client publish something untrue.
Winning-answer themes
What the answers that beat you actually emphasise — pricing model, use-case fit, independent proof. Observed patterns, not hidden ranking factors.
Owned-site evidence
A bounded read of your public pages. The finding is always “no matching evidence in the inspected set”, never “your website does not have this”.
Source addressability
A frequently cited domain is not automatically somewhere you can get listed. Frequency is evidence of relevance, never evidence of access.
ProductTruth
Verified true, verified false, or unknown — checked before any content change is recommended. Unknown stays unknown.
Owned evidence≠ProductTruth
A capability missing from a page may still exist in the product, and a claim written on a page may be out of date. Silently converting webpage evidence into product fact is how an “optimisation” becomes inaccurate marketing that the brand — not the tool — carries.
4 · Turn a finding into one defensible action
Every action names its target, the change, the evidence behind it, and the question that will be remeasured. Nothing edits your site automatically.
Recommended action
Illustrative exampleTarget
The one owned page already matched to this buyer question.
Recommended change
Make a verified capability explicit on that page, in the terms the winning answers use.
Evidence
Winning-answer themes observed across 6 compatible runs, plus the inspected owned page, plus a ProductTruth state of verified true.
What we will remeasure
The same question, same engines, same protocol.
Confidence
Bounded by how much compatible evidence exists — and it reduces certainty, never inflates it.
Proof, not promises
5 · Show whether the measured state actually changed
Re-run the same question under a compatible protocol. If the cohort, the providers or the context moved enough to make the comparison invalid, we refuse the number and say why.
Before / after · same question, compatible protocol
Illustrative example“Which [category] platform has no monthly subscription?”
Baseline
Follow-up
Before / after≠Causation
“Recommendation coverage improved from 0 of 3 to 2 of 3 engines after the recorded change” is valid. “The page change caused two engines to recommend the brand” is not, and no observational before/after can support it.
What we have actually measured so far
We have no customers yet, so we will not show you any. What we can show is the run we did on our own brand to find out whether the engine surfaces a real commercial decision or just another chart.
OwnQR · stated positioning, by engine
Founder-owned dataThe useful output was not “AI visibility: 42”. It was: the core positioning is understood by one engine and not consistently by the other two, and the claim the brand says it is built on is being won by a competitor everywhere.
For agencies
One relationship covers several client brands, agencies already run recurring reporting, and a finding can become paid work. That is why we started here.
Agency workflow
Planned — not builtPer client
10–20 buyer questions built from their category, ICP and named competitors.
Approval gate
The client approves the question list before anything is measured. No provider cost is incurred until they have seen it.
Measurement
All three engines, reported per engine, failures excluded from denominators.
Deliverable
A founder-reviewed report file you can forward to the client, plus a walkthrough call.
Remeasurement
The same questions again after the work, under a compatible protocol.
Founding pilot pricing
Validation prices while we find out what this is worth. Delivered as a founder-reviewed report, not a dashboard — and priced accordingly.
Founding Brand Pilot
- 1 brand, 10–20 buyer questions
- ChatGPT, Gemini and Perplexity
- Recommendation status per question, per engine
- Citation evidence and competitors named instead of you
- Founder-reviewed deliverable
Founding Agency Pilot
- Up to 3 client brands
- Everything in the Brand Pilot, per brand
- Directional competitive visibility
- Prioritised findings with the evidence attached
- Founder-reviewed deliverable + strategy walkthrough
14-day pilot from the date of purchase — the same date Stripe bills from. Cancel any time. There is no free tier and no trial.
Where to go from here
Evaluating tools
What an AI visibility tool has to get right, and how seven of them compare — including where we lose.
Understanding the measurement
AI search visibility explains the model; AI search analytics defines the metrics underneath it.
Tracking it over time
AI search monitoring for the workflow, AI brand monitoring for how you are represented.
Seeing the output
What a useful report contains, or start with a free visibility check.
Common questions
Which AI assistants do you measure?
ChatGPT (via the OpenAI API with web search), Google Gemini (with Google Search grounding) and Perplexity. Results are reported per engine, never blended into a single "AI" number.
Do you measure the consumer ChatGPT app?
No. We measure the OpenAI API with web search enabled. That is an OpenAI measurement surface, not a logged-in consumer ChatGPT session, so results can differ from what an individual user sees in ChatGPT. Every report discloses the measurement surface.
How is this different from counting brand mentions?
We start from the questions a buyer asks before choosing a vendor, and report which of those questions a competitor wins. Mention volume is shown, but it is not what we treat as the value.
What do I get in a pilot?
A founder-reviewed audit of 10 to 20 buyer questions per brand across three engines, with the recommendation status, the citation evidence, the competitors named instead of you, and a prioritised list of what to address. Plus a walkthrough call.
What we actually measure
Every result comes from the official OpenAI, Google Gemini
and Perplexity APIs with web search or grounding enabled. In our data those
surfaces are recorded as openai_api_web_search,
gemini_api_google_search and perplexity_api_sonar, and reports name
them explicitly.
That is not the same thing as a logged-in person using the ChatGPT app. These are official interfaces with web search on, but they are measurement surfaces rather than consumer sessions: no personalisation, no memory, no app-only features. Results can differ from what any individual user sees, and we do not claim the two are identical. We would rather put that on the pricing page than in the small print.
Find out which buyer questions you are losing.
Send a domain and a category. We run five buyer questions across all three engines and tell you what they actually answered — including when the answer is that we found nothing worth acting on.
Founding pilot pricing while we validate. 14-day pilot, founder-run, cancel any time. Or request a free check first.