One score per AI model
Every AI model in your audit gets one score from 0 to 100. It blends three numbers:
| Part | Weight | What it measures |
|---|---|---|
| Cited | 50% | How often the model links to your site as a source in its answer. |
| Visibility | 30% | How often you show up at all: linked, named, or among the pages the model read. |
| Share of Voice | 20% | Your share of all the sites the model cited across your questions. |
Each part is a percentage of the questions the model answered. If a model gave no answers (an outage, for example), it gets no score rather than a zero.
Example
Say Claude answered your five questions. It linked to your site in 2 answers (Cited 40%), named you without a link in 1 more (Visibility 60%), and your site was 10% of all the sites it cited (Share of Voice 10%). The score is 40 × 0.5 + 60 × 0.3 + 10 × 0.2 = 40.
Rolling up: engine and overall
- Engine score. Each engine (ChatGPT, Gemini, Claude, Google AI) averages the models inside it.
- Overall score. Every engine counts the same.
You can change both weights in the report under How the overall score adds up. Set a weight to 0 to leave an engine or model out. Weights go from 0 to 10.
What the bands mean
| Score | Band |
|---|---|
| 0-19 | Not showing up |
| 20-39 | Weak |
| 40-59 | Fair |
| 60-79 | Strong |
| 80-100 | Leading |
API vs as seen
Cited can ask a model two ways:
- API: the model's developer API with web search turned on.
- As seen: what a real person sees in the ChatGPT or Gemini app.
Where an engine has as-seen results, those count toward the engine score and the API results are shown beside them for comparison.
Repeats
AI answers vary from one ask to the next. In a paid audit each question is asked three times per model, and the question's row shows the majority result and how many of the three runs cited you.
The site-check score
The site check has its own 0-100 score, separate from the AI scores. It measures whether your site gives AI what it needs: bot access, llms.txt, sitemap, structured data and more. See Site check.
Using your scores
- Look at the answers, not just the number. Every question shows each AI's answer and the sites cited instead of you. That list is often the most useful thing in the report.
- Compare like with like. Scores move when you change questions or models. To track progress, re-run the same questions and check Since the last audit in the report.
- Give fixes time. AI engines pick up site changes over days to weeks. Re-check two to four weeks after you ship a batch of fixes.
For a list of every term, see Metric definitions.