AI Search Visibility (GEO) · Service
AI Visibility Intelligence
Ongoing, sampled measurement of how AI assistants mention, cite and describe your brand and competitors across markets and languages, with a disclosed method and no composite score dressed up as revenue.
An AI visibility audit tells you where you stand once; AI visibility intelligence tells you how it moves. SEO Evaluate runs a monitored prompt set on a fixed cadence across the assistants your buyers use, records every answer, and reports brand mentions, actual citations of your pages, answer position, competitor presence and factual accuracy per market and language. Every figure comes with its denominator, sample size, provider and run dates, so a change in the numbers can be traced to a change in answers rather than to a change in method. This consolidates the measurement side of our AI search visibility work into one monitoring offer; the one-off diagnostic remains the AI visibility audit.
How we work
/01 · Definition
What is AI visibility intelligence?
AI visibility intelligence is the ongoing, sampled measurement of how AI assistants respond to the questions your buyers ask: whether they name your brand, cite your pages, place you first or in passing, name competitors, and describe you accurately, per market and language. It runs a versioned prompt set on a fixed cadence across named providers, logs every response, and reports with explicit denominators and sample sizes. It is monitoring with disclosed method, not a proprietary score; it cannot see a platform's internal trust signals and does not equate visibility with revenue.
The hard part is comparability. Assistants answer differently from run to run, change models without notice and personalise by account and location. The protocol fixes what can be fixed: the prompts, the providers, the cadence, the number of repeats, the session conditions and the coding rules. What cannot be fixed is reported as variance, not hidden.
Any composite indicator we publish is an original, disclosed heuristic computed from the underlying counts, so you can recompute it or ignore it.
/02 · Who it is for
Who this is for
Brands that completed an AI visibility audit
You have a baseline and a roadmap; now you need to know whether the work moves the answers.
Multi-market businesses
Visibility differs by language and country; one global number hides the market where you are losing.
Teams reporting to leadership
You need figures with denominators and dates that survive scrutiny, not screenshots.
Categories with active competitors in AI answers
When assistants recommend alternatives, you want to know which ones, how often and on which questions.
/03 · Problems we solve
Problems this replaces
Screenshot evidence
One favourable answer in one session is not visibility. Repeated, dated sampling is.
Scores without denominators
A visibility score with no sample size, provider list or run dates cannot be compared month to month.
Mentions mistaken for citations
Being named is not being linked. The two are reported separately because they need different fixes.
Method drift
Prompt sets and providers change silently, and movement is misread as progress. Every method change is logged and reported apart from answer changes.
What is recorded for every answer
Each sampled response is coded on the same fields: provider and model label where visible; run time; the exact prompt; whether the brand or any alias is mentioned; whether one of your URLs is cited; position, using the definitions the platform's answer format supports (first named, listed, in passing, absent); every competitor named; every cited domain; and accuracy against a fact sheet you approve.
Sentiment is coded only when it is unambiguous; otherwise it is recorded as unknown rather than guessed. Failed or empty requests are counted and excluded from rate denominators, and duplicate answers within a run are de-duplicated before counting.
/05 · Comparison
Denominators and what each rate means
| Measure | Numerator | Denominator | Not counted |
|---|---|---|---|
| Mention rate | Runs naming the brand or an approved alias | Successful runs for the prompt set | Failed, empty or duplicate runs |
| Citation rate | Runs citing one of your URLs as a source | Successful runs on providers that show sources | Providers without visible sources |
| First-position rate | Runs naming you first among brands | Successful runs that name at least one brand | Runs naming no brand |
| Competitor share | Runs naming each competitor | Successful runs for the prompt set | Competitors outside the agreed list |
| Accuracy | Statements coded correct | Statements about the brand | Runs with no statement about the brand |
/06 · Method and process
How the monitoring runs
- 01
Set-up
Markets, languages, providers, competitors, cadence and the fact sheet are agreed. The prompt set is built from real buyer questions and versioned.
- 02
Sampling
Prompts run on the agreed cadence with repeats and clean sessions; every response is logged with provider, time and prompt.
- 03
Coding and validation
Responses are coded on the fixed fields; entity and alias matching is checked by a reviewer on a sample each cycle.
- 04
Reporting
The monthly report shows rates with denominators, competitor comparisons, the accuracy register and the source map, with method changes listed separately.
- 05
Decisions
Movements are tied to candidate causes and handed to the GEO, content or technical workstream that owns the fix.
/07 · Strategic considerations
Limitations we state in every report
Sampled, not census
A prompt set is a sample of buyer questions; rates describe the sample and its confidence, not every conversation.
Provider behaviour is a black box
We observe outputs. We do not see ranking signals, internal trust scores or why a model chose a source.
Visibility is not revenue
A higher mention rate is an input to demand, not a sales figure. Business outcomes are measured in your analytics, separately.
Platform and model changes
Providers change models and retrieval without notice; the change log records what we observed and when.
/08 · Frequently asked questions
AI visibility intelligence: frequently asked questions
How is this different from the AI visibility audit?
The audit is a one-off diagnostic with a cause analysis and roadmap. Intelligence is the ongoing measurement that shows whether the roadmap moves the answers, on a fixed cadence with the same method each cycle.
Which assistants do you monitor?
The providers your buyers use in your markets, agreed at set-up and named in every report. Providers without visible sources are reported for mentions but not for citations.
Do you provide a single visibility score?
We can compute an indicator from the underlying rates, and we publish exactly how it is calculated so you can recompute it. It is an original heuristic, not an industry standard, and it never replaces the rates and sample sizes.
Can you measure sentiment?
Only where it is unambiguous. Mixed or unclear tone is recorded as unknown rather than forced into positive or negative.
Does better visibility mean more sales?
Not by itself. Visibility is an input; conversions depend on your offer and your pages. We report business outcomes separately from visibility and never equate the two.
/09 · Supporting resources
Related services and tools
- AI search visibility (GEO)The parent programme: content, entity and technical work that moves the measures.
- AI visibility auditThe one-off diagnostic and roadmap.
- AI brand visibility & competitor analysis (free tool)A self-service version of the coding method on your own exported responses.
- Analytics and dataWhere business outcomes are measured.
Related services
Services worth considering together
Generative Engine Optimization (GEO) & AI Search Visibility
Be the source ChatGPT, Perplexity, AI Overviews and Gemini cite when buyers ask: measured, structured, honest about limits.
Learn moreAI Visibility Audit
A fixed-scope diagnostic of how ChatGPT, Google AI Overviews, Perplexity, Gemini, Copilot and Claude describe your brand today, and why.
Learn moreGEO Content & Entity Architecture
One clear record of who you are and what you offer, written into your pages, schema and profiles.
Learn moreAnalytics & Data Strategy
Measurement you can defend: GA4, Tag Manager, server-side tagging and Consent Mode v2 as one plan.
Learn moreSEO Services & Consultancy
One SEO programme across technical, content, authority, international and AI search, delivered natively in four languages.
Learn moreContent Marketing & SEO Content
Editorial strategy, topic clusters, briefs and native production in four languages, with every piece measured.
Learn more
- Meta
- TikTok
Next step
Request an AI visibility assessment
Tell us where you are today; we will scope the work, set priorities and outline the first 90 days together. No commitment, a written assessment.