What is a good AI visibility score?
There isn’t a universal “good” AI visibility score. Your number only makes sense in the context of completed questions, what was counted as mentions vs citations, the six‑page technical sample, and the specific provider, locale and time of observation.
There is no universal good AI visibility score. Interpreting any score depends on what was actually measured: how many of the ten questions finished, whether your brand was mentioned or your exact domain was cited, your six‑page technical readiness sample, and the provider, locale and timestamp of the observations. The score is our disclosed evidence score, not an industry standard or a search‑engine metric (How the audit works).
What the score measures
The combined score has three visible components with fixed weights (How the audit works):
- Technical readiness: 40% (six public pages checked for crawlability, identity, structured data and answer clarity within strict bounds)
- Site citations: 40% (out of ten completed answers, count of answers that include a native citation to your exact hostname)
- Brand mentions: 20% (out of ten completed answers, count of answers that literally mention your expected brand)
The arithmetic is disclosed and rounded to a whole number: readiness × 0.4 + cited answers × 4 + mentioned answers × 2 (How the audit works). An answer can contribute to both mention and citation, but there is no extra “either” bonus.
Why no single threshold is “good”
1) Denominator and completion status
A final number only appears when all ten reviewed questions return valid grounded answers and technical readiness is known. Until then, you see a possible score range. Unknown questions retain their full zero‑to‑six‑point potential, and unknowns are never silently replaced with zeros (How the audit works). A partial run can’t be compared fairly with a complete one.
2) What counts as a win: mention vs exact‑domain citation
A literal brand mention and a native citation to your exact hostname are separate observations. Prose URLs, search‑result‑only links, other subdomains or lookalike hosts do not count as citations. Completed answers without either observation are confirmed negatives for this sample, not failures of the tool (How the audit works). Whether a score feels “good” depends on which of these you prioritize.
3) Provider, locale and time
Results describe only these reviewed questions, the API model and the recorded time; API answers can differ from consumer apps. ChatGPT and Perplexity are not tested; Gemini provides an unscored preview. An optional site‑restricted Google discovery snapshot runs in US English on desktop and does not change the combined score (How the audit works). Provider choice, locale and timestamp all shape interpretation.
4) Selected questions and sample bias
The ten questions are neutral, site‑based prompts prepared from your sampled public pages; they are not People Also Ask or measured keyword demand. If the entry page doesn’t ground useful questions, the tool asks for a better page rather than inventing them (How the audit works). Your score therefore reflects visibility on this curated sample, not market share.
5) Technical readiness is a bounded sample
We inspect the HTTPS entry page and up to five internal pages, within strict size and time limits. We don’t sign in or execute scripts; JavaScript‑rendered content and pages outside the sample can be missed. A missing signal in six pages is not proof it is absent from the entire domain. Unknown observations stay visible, and entire unavailable categories are excluded with weights normalized (How the audit works).
A hypothetical worked example
Hypothetical numbers for arithmetic only (not a measured result):
- Technical readiness: 80/100
- Across ten completed answers: 4 include a native citation to your exact domain; 3 include a literal brand mention
Score = 80 × 0.4 + 4 × 4 + 3 × 2 = 32 + 16 + 6 = 54, displayed as 54/100. If only six answers had completed at that timestamp, the tool would show a range because the four unknown questions still carry zero‑to‑six potential each (How the audit works).
How to judge your own result
- Compare like with like. Hold provider, locale and timing constant when trending. API answers can differ from consumer apps, and Gemini’s preview is unscored (How the audit works).
- Check denominators. Only compare complete ten‑question runs; partials have ranges and unknowns are not zeros (How the audit works).
- Read the per‑question evidence. Confirm whether a “win” was a brand mention, an exact‑domain citation, or both (How the audit works). For interpreting their impact, see AI citations, brand mentions and visits: how to read the evidence.
- Validate real traffic separately. Google notes that links surfaced in AI Overviews and AI Mode are part of overall Search Console Web traffic; analyze changes there and in your analytics (AI features and your website). For workflows, see How to see AI referral traffic in GA4?
- Use the ten‑question method consistently. If you need a repeatable procedure, start with How to measure AI search visibility with a ten‑question audit.
Quick checklist
- Confirm 10/10 answers completed and technical readiness measured.
- Note the provider/model and the timestamp of the run.
- Record mention and exact‑domain citation counts separately.
- Review any unknowns; don’t treat them as zeros.
- Inspect the six sampled pages and fix crawl, identity, and structured‑data issues.
- Re‑run after changes and compare runs with the same setup.
- Cross‑check Search Console Web traffic for landing pages cited or mentioned (AI features and your website).
- Read the evidence per question and plan page‑level improvements; see the readiness vs visibility distinction in AI search readiness and visibility: what an audit can actually tell you.
FAQs
Does a higher score guarantee inclusion in Google’s AI Overviews?
No. The score is not a Google metric, and it measures only this sample’s observations at a point in time. Appearing in AI features follows the same technical eligibility as Search overall, and clicks from AI features are counted within overall Search Console Web data (How the audit works) (AI features and your website).
Can I benchmark against another brand’s number?
Only with caution. Unless both runs used comparable questions, a similar technical sample, the same provider, locale and timing, differences may reflect setup rather than true visibility variation (How the audit works).
Why is my range wide?
Unknown measurements remain unknown until completed; each unanswered question retains its full potential contribution and is not treated as zero. Finish the ten questions to collapse the range into a single number (How the audit works).
Related reading: For structuring pages that answer clearly, see How to structure content for AI search? and explore whether broader AI changes may impact your traffic in Is Google AI reducing website traffic?
Related guides
Sources and further reading
Provider documentation can change. Check the current guidance before changing crawler or search settings.