Problem · AI search
We can’t measure our AI visibility. A fixed prompt set, measured the same way.
Management asks how you’re doing in ChatGPT and AI Overviews, and there is no number to show. AI answers vary, so measurement needs a fixed method, not screenshots.
Is this you?Signs we usually see
- No report on AI mentions or citations
- Screenshots instead of numbers
- Different people get different answers
- No link between AI visibility and leads
01 / Quick checks
Three checks you can run today
Each takes minutes. Together they tell you whether this page is about your problem.
- Write 20 promptsThe questions your buyers ask, by stage and type.
- Run them twiceSame prompts, same engines, a week apart. Note what changes.
- Check referralsIn analytics, look for referral traffic from AI assistants.
Prompt sampling and denominator register
Define a fixed buyer-intent sample and a repeatable collection method before reporting a visibility percentage.
| Situation to check | Proposed scope | Boundary |
|---|---|---|
| Write 20 prompts | The questions your buyers ask, by stage and type. Mention rate, citation rate and recommendation rate answer different questions and need separate counts. | Changing the prompt set between runs can create apparent growth without any improvement on comparable questions. |
| Run them twice | Same prompts, same engines, a week apart. Note what changes. | Keep an unchanged comparison sample so this check can distinguish the proposed cause from unrelated variation. |
| No fixed prompt set | Different questions each time, so no trend. | Treat this as a hypothesis until the affected URL or account evidence supports it. |
| Ready to commission a fix | Publish the denominator, failures, dates and engine modes beside each trend, with sampled answers available for inspection. | [CONFIRM: access, scope, owner, fee and review date before work begins.] |
Prompt sampling and denominator register
Illustrative blank specification. This is a diagnostic structure, not a client result.
- Affected population
- Define a fixed buyer-intent sample and a repeatable collection method before reporting a visibility percentage.
- Decisive observation
- Mention rate, citation rate and recommendation rate answer different questions and need separate counts.
- Alternative explanation
- Changing the prompt set between runs can create apparent growth without any improvement on comparable questions.
- Acceptance evidence
- Publish the denominator, failures, dates and engine modes beside each trend, with sampled answers available for inspection.
- Supporting check
- In analytics, look for referral traffic from AI assistants.
- Approval
- [CONFIRM: accountable owner and authorised next step.]
What the first handover should make clear
- Define a fixed buyer-intent sample and a repeatable collection method before reporting a visibility percentage.
- Mention rate, citation rate and recommendation rate answer different questions and need separate counts.
- Publish the denominator, failures, dates and engine modes beside each trend, with sampled answers available for inspection.
- The original observations, unresolved questions and any client access needed for the next scope.
02 / Likely causes
What usually causes it
In the order we usually check them. Most cases have more than one.
- No fixed prompt setDifferent questions each time, so no trend.
- Answers vary run to runOne run per prompt isn’t enough.
- No source trackingMentions counted, but not where they come from.
- AI referrals hiddenSome AI traffic arrives without a referrer.
- No competitor baselineA number without context.
03 / What we do
How we fix it
Four steps, in order. You see the evidence and the plan before anything changes on your site.
- Step 01Define the setPrompts agreed with you, by market and language.
- Step 02BaselineEach prompt run several times on each engine.
- Step 03Report monthlyMentions, citations, sources and competitors, dated.
- Step 04Connect to resultsAI referrals and leads tracked where possible.
04 / Services
Services that fix it
Starting prices in USD. TopClick is not VAT-registered. Scope and price in writing before work starts.
TopClick is not VAT-registered
AI visibility monitoring ↗Mentions, citations and share of voice in AI answers, tracked against competitors.From $[X] /monthRequest
AI visibility audit ↗A one-off baseline of where you stand in AI answers, with a prioritised plan.From $[X] /projectRequest
SEO reporting ↗Work done, search signals and business results tracked as separate measures.From $[X] /monthRequest
SEO & AI search training ↗Workshops that let your marketing and dev teams run the work in-house.From $[X] /sessionRequest
05 / Questions
Questions about this problem
01Why not just use a tool?
Tools help. What matters is a fixed prompt set, repeated runs and someone reading the answers, we use tools and check by hand.
02Can AI visibility be tied to revenue?
Partly: AI referrals, branded search growth and leads that mention AI. We report what can be measured and say what can’t.
03Where do we start?
With the free snapshot: we check your site for this problem, confirm the likely cause and send a written first scope with a price.
Start here
Check your site for this
- 1Send your site — we confirm the cause and where we would start
- 2We check search and AI visibility
- 3PDF by email: five website problems and AI visibility, within two business days of receiving your website URL
Prefer to talk?Book a 20-min callhello@topclick.expert
Request receivedWe'll reply within one business day. Want to talk sooner? Book a call in the next tab.See what happens next →
20-min intro callVIDEO CALL · FREEA strategist looks at your site with you and suggests where to start. No pitch.
Times in your time zone · your local time
Enter a valid email address
We couldn’t send this. Please try again or email hello@topclick.expert.
RequestedWe’ll confirm the time by email within [X] business hours.How to prepare for the call →