Last updated: 24 October 2026
An AI visibility tool's score is only as real as the methodology behind it, and that methodology is rarely published in full. A nine-platform audit found that three vendors even contradict their own stated definitions, while two tools can score the same brand 45 and 22 with no actual change in visibility — so the number on the dashboard needs verifying, not trusting, before a UK business signs a contract.
Key Takeaways
- A nine-platform methodology audit found three vendors publish internally contradictory definitions for their headline visibility metric, according to AuthorityTech, 2026.
- Two AI visibility tools can score the same brand at 45 and 22 with zero real change in visibility, because each tracks different engines, prompts and weightings, according to Pixis, 2026.
- Only 49% of marketing and finance leaders can clearly explain their measurement approach to the board, according to a 2026 survey of 500 senior decision-makers cited by Metricus, 2026.
- Dashboards measuring the same brand in the same week can differ by 20 to 30 points because of different sampling frequencies, prompt libraries and model versions, according to Influencers Time, 2026.
- Over 40 GEO platforms now measure "AI visibility", and most track metrics that don't correlate with traffic or conversions, according to Mintec, 2026.
What is an AI visibility score, and is there an industry-standard methodology?
An AI visibility score is a vendor-calculated number, usually expressed as a percentage or index, that claims to represent how often a brand appears in answers generated by AI assistants such as ChatGPT, Gemini, Perplexity, Claude, Google AI Overviews and Microsoft Copilot. There is no industry-standard methodology for producing it — no equivalent of a British Standard or an ISO specification governs how the figure is calculated, and no regulator such as the Advertising Standards Authority has issued guidance on what a "visibility score" must mean.
This absence of a shared definition is the root cause of most scoring confusion. An independent audit of nine platforms found three vendors publish multiple metrics that are easy to compare incorrectly, and three publish definitions that only agree once you check every surface carefully, according to AuthorityTech, 2026. The same audit traced a worked example on one well-known platform where a headline "Visibility Score" produces either 50% or 83.3% for the identical dataset, depending on which of two published denominators is used, according to AuthorityTech, 2026.
Treat "AI visibility score" the way you'd treat "search ranking" before Google standardised its own reporting through Search Console — a useful concept, but meaningless until you know exactly what's being counted and over what base.
What underlying data sources does the tool use, and can the vendor disclose this?
A visibility score is only as trustworthy as the data feeding it, and the three realistic sources are live AI model queries, scraped search engine results, or statistical estimates extrapolated from a smaller sample. Ask the vendor, in writing, which of the three it uses for your specific plan tier — many tools reserve real-time querying for premium plans and quietly fall back to estimates for lower tiers.
An independent methodology review of one leading AI visibility platform identified six structural flaws, including a severely limited data source, a skewed user panel, and amplified extrapolation errors, according to Metricus, 2026. That matters because an extrapolation error on a small sample doesn't shrink as the dashboard scales — it compounds, producing a score that looks precise to one decimal place while resting on thin evidence.
Before paying, insist the vendor confirms in writing:
- Whether scores come from live queries against ChatGPT, Gemini, Perplexity, Claude, Google AI Overviews and Copilot, or from a cached/estimated proxy
- How many actual prompt-response pairs underpin last month's reported score
- Whether synthetic (AI-generated) prompts or real user query logs were used to build the prompt library
Can the score be independently verified by manually testing prompts yourself?
Independent verification means running the same prompts the vendor claims to track, directly in ChatGPT, Gemini, Perplexity and Claude, and comparing what you see against the vendor's reported figure. This is the single most reliable check available to a UK buyer, because it bypasses the vendor's black box entirely and uses the free consumer interfaces of each engine.
Request the vendor's actual prompt list — not a category description like "informational queries" but the literal strings being sent to each model. Run 15–20 of them yourself across a working day, screenshot the responses, and tally how often your brand appears compared with named competitors. If your manual spot-check sits wildly outside the vendor's reported score, that's a red flag worth raising before the trial ends, not after the invoice arrives.
Bear in mind that AI assistants are non-deterministic: the same prompt sent twice can produce two different answers, which is precisely why a single manual check is a sense-check, not a replacement for ongoing measurement. Aether AI builds this expectation into its own citation tracking across six engines — ChatGPT, Perplexity, Google AI Overviews, Claude, Gemini and Copilot — by sampling repeatedly rather than reporting a single snapshot as fact.
Does the vendor disclose sample size, query frequency or a confidence interval — or is it a black box?
A credible AI visibility tool discloses the number of prompt runs, the frequency of re-testing, and ideally a confidence interval behind every score it publishes — without these three figures, the number is a black box. Dashboards tracking the same brand in the same week have been shown to differ by 20 to 30 points purely because of different sampling frequencies, prompt libraries and model versions, according to Influencers Time, 2026.
This sits alongside a wider trust gap at board level. A 2026 survey of 500 senior decision-makers found only 49% of marketing and finance leaders can clearly explain their measurement approach to the board, according to Metricus, 2026. If your own marketing lead can't explain the sample size behind a vendor's score, the board certainly won't be able to defend the budget line that funds it.
A reasonable minimum to request: 30+ prompt runs per tracked keyword, re-tested at least weekly, with raw response logs available on request — not just the aggregated score.
How does the tool handle UK-specific visibility — UK English, UK competitors and regional intent?
UK-specific AI visibility means the tool tests prompts phrased the way British buyers actually search — using "nearest branch in Manchester" rather than a US-style query, tracking UK competitors such as a London-based rival rather than a US equivalent, and recognising UK English spelling and terminology. Many GEO platforms were built US-first, and their default prompt libraries, model regions and competitor sets reflect that.
Ask directly whether the tool's prompts are written in UK English, whether its default competitor set includes genuinely UK-based rivals, and whether it can isolate results for a specific region — for example comparing visibility in Leeds against visibility in Bristol for a business with branches in both. If the vendor can't answer precisely, assume the dashboard is reporting a global or US-weighted average dressed up as a UK figure.
"It's a land grab: venture-funded dashboards, SEO suites bolting on AI modules, and platforms that actually produce and publish content. Judge them by what happens after the report — visibility you can't act on is just a prettier way to worry." — Lauren Dawkins, Head of Content, Aether AI
What are the common vendor tricks that inflate a visibility score?
Score inflation usually comes from denominator manipulation, cherry-picked prompts, or synthetic panels built to flatter the result rather than reflect reality. The same nine-platform audit that found contradictory metric definitions also found platforms publishing multiple similarly-named metrics that are easy to compare incorrectly — a structure that makes month-on-month "improvement" easy to manufacture simply by switching which metric is reported, according to AuthorityTech, 2026.
Watch for these specific patterns:
| Red flag | What it looks like | Why it inflates the score |
|---|---|---|
| Denominator switching | "Visibility" quietly redefined between dashboard views | Same data, different base, looks like growth |
| Synthetic-only prompts | No real user query logs behind the prompt library | Scores a hypothetical market, not your actual buyers |
| Cherry-picked competitor set | Weak or irrelevant competitors chosen for comparison | Makes your brand look dominant by default |
| No raw data on request | Vendor refuses to share underlying prompt/response logs | Impossible to audit or reproduce |
| Single-snapshot reporting | One test run presented as "your score" | Ignores natural AI answer variability |
Over 40 GEO platforms now measure "AI visibility", and most track metrics that don't correlate with traffic or conversions at all, according to Mintec, 2026. A rising score that never translates into enquiries or sales is the clearest practical sign something upstream is wrong. That disconnect is severe: only 9% of brand marketers say they can confidently prove which AI chatbot mentions actually drove revenue, according to eMarketer survey data cited by Influencers Time, 2026.
What should a UK business pay for an AI visibility tool, and what should justify that cost?
A fair UK price benchmark depends on what's actually included — live multi-engine querying, competitor tracking, Google Search Console integration and content recommendations each add cost, and a bare leaderboard number should sit at the bottom of any pricing tier, not the top. Treat any plan priced like a premium product but delivering only a single dashboard score, with no raw data, prompt library or re-testing schedule, as poor value regardless of the headline figure.
Before committing, confirm:
- Which of the six major AI engines are actually queried (ChatGPT, Perplexity, Google AI Overviews, Claude, Gemini, Copilot), not just claimed on the marketing page
- Whether competitor and keyword tracking is included or a paid add-on
- Whether there's a free audit or trial period long enough to run your own manual verification (see the earlier section)
- What the cancellation and refund terms are if the reported score can't be reproduced
Aether AI publishes its pricing openly and offers a free AI-visibility audit at /audit, specifically so a prospective buyer can see a sample of real citation tracking across ChatGPT, Perplexity, Google AI Overviews, Claude, Gemini and Copilot before committing any budget.
Who should validate the score before budget sign-off, and how often should it be re-tested?
Responsibility for validating an AI visibility score should sit with whoever owns the marketing budget line — typically a marketing lead, in-house SEO manager, or the retained agency — and it should never be signed off on the vendor's dashboard alone. Given that only 49% of senior marketing and finance leaders can clearly explain their own measurement approach to the board, according to Metricus, 2026, the person signing off spend should be able to answer the same questions a board member would ask.
Re-testing should happen at least weekly, not monthly or quarterly, because AI models update frequently and answers drift. Dashboards that sample only occasionally can diverge by 20 to 30 points from one another in the same week, according to Influencers Time, 2026.
Your AI visibility score verification checklist
- Request the vendor's exact prompt list, not a category description
- Run 15–20 of those prompts manually in ChatGPT, Gemini, Perplexity and Claude
- Ask for the sample size and re-testing frequency behind last month's score
- Confirm whether prompts are UK English with genuinely UK-based competitors
- Check whether the vendor can show raw response logs, not just the aggregated dashboard figure
- Compare the score trend against actual traffic or enquiry data for at least one quarter
- Confirm trial length, cancellation terms and refund policy in writing before paying
- Re-test weekly once live, not quarterly, to catch model-driven drift early
FAQ
What is a good AI visibility score?
There is no universal "good" score because no two vendors calculate it the same way. A score of 60% on one platform and 25% on another could represent identical real-world visibility, so judge trend and reproducibility within one tool rather than comparing absolute numbers across tools.
Why do different AI visibility tools give different scores for the same brand?
Different tools track different engines, use different prompt libraries, apply different weightings, and often define the metric's denominator differently. Two tools have been shown to score the same brand at 45 and 22 with no actual change in visibility, according to Pixis, 2026.
Can I trust AI visibility leaderboards and rankings?
Treat leaderboards as directional at best, not authoritative. An independent audit of nine platforms found contradictory or easily-confused metric definitions across the majority of vendors checked, according to AuthorityTech, 2026, which undermines any ranking built by comparing scores across tools.
Is AI visibility score a vanity metric?
It can be, if it isn't tied to a reproducible method and doesn't correlate with business outcomes. Most of the 40-plus GEO platforms now measuring "AI visibility" track metrics that don't correlate with traffic or conversions, according to Mintec, 2026, so always cross-check the score against enquiries, leads or sales.
How many prompt runs are needed before a score is statistically reliable?
There's no universally agreed minimum, but a reasonable floor is 30 or more prompt runs per tracked keyword, re-tested at least weekly, given that same-week dashboards have diverged by 20 to 30 points due to differing sampling frequency, according to Influencers Time, 2026.
Should I track AI visibility alongside traditional SEO metrics?
Yes — AI visibility and traditional SEO measure overlapping but distinct things, and tracking both avoids over-reacting to a single noisy number. A rising AI visibility score with flat organic traffic, Search Console clicks or enquiries is a signal to investigate the tool's methodology rather than celebrate.
What should I ask a vendor before signing a contract?
Ask for the exact prompt list, the data source (live query versus estimate), sample size, re-testing frequency, UK-specific competitor handling, and whether raw logs are available on request. A vendor unwilling to answer any of these in writing is itself the clearest red flag.
Verifying visibility with Aether AI
Choosing between competing AI visibility dashboards is exactly the decision this article has worked through, and it's the same due diligence Aether AI expects buyers to apply to its own platform. Aether AI runs citation tracking across six AI engines — ChatGPT, Perplexity, Google AI Overviews, Claude, Gemini and Copilot — alongside keyword and competitor tracking and Google Search Console integration, built on the same engine that writes and publishes for the founding team's own four brands.
Aether AI published 281 articles in the last 30 days across the four brands it writes and publishes for, spanning security, facilities software, branding and the platform itself. If you want to apply the seven checks in this article to a real dashboard before paying anything, start with the free AI-visibility audit at aether-ai.co.uk/audit and see the underlying data for yourself.