The short answer
Use an AI visibility score for orientation only after reading its formula. Use citation rate when you need to know how often your domain received a visible citation across a defined set of answer opportunities. Neither metric replaces the underlying prompts, answers, dates, products, and cited URLs.
“AI visibility score” has no single industry-wide formula. Different publishers can combine mentions, citations, prominence, sentiment, accuracy, or competitor comparisons using different weights. Two tools can therefore give different scores without either calculation being broken.
What an AI visibility score measures
A composite score compresses several observations into one number. That makes dashboards easier to scan, but it can hide the state that changed.
- Read the formula and weights.
- Check whether mentions and citations are separate inputs.
- Check whether branded and brand-free prompts are mixed.
- Check whether products with different citation volumes are combined.
- Check whether representation accuracy or sentiment affects the number.
- Confirm that the raw observations remain accessible.
A score is most useful for routing attention: it can tell you where to investigate. It is weaker as proof that a specific page change produced a specific outcome.
What citation rate measures
Define citation rate before reporting it. Really Good GEO uses this page-level form when the question is “How often did the observed answers visibly cite our domain?”
If a frozen test produced 20 eligible responses and 5 visibly cited the measured domain, the observed citation rate is 25% for that test.
Another publisher may define citation rate as a share of all citations awarded, rather than a share of responses. Both calculations can be valid, but they answer different questions. Publish the formula, numerator, denominator, product, prompt set, dates, and run count.
A worked comparison
| Observation | Test A | Test B |
|---|---|---|
| Eligible responses | 20 | 20 |
| Brand mentioned | 12 | 9 |
| Own domain visibly cited | 4 | 7 |
| Mention rate | 60% | 45% |
| Own-domain citation rate | 20% | 35% |
A composite score could rank either test higher depending on its weights. The separate rates show the real tradeoff: Test A produced more mentions, while Test B produced more own-domain citations.
Which metric should you use?
| Decision | Primary measure |
|---|---|
| Scan many brands or topics for investigation | Transparent composite score plus components |
| Determine whether the brand appears | Mention rate |
| Determine whether the domain receives visible attribution | Own-domain citation rate |
| Evaluate whether the answer represents the brand correctly | Accuracy review using a disclosed rubric |
| Evaluate a controlled page change | Repeated component observations, baseline, comparison, and limitations |
For test stability, see why one ChatGPT result is not a ranking and how many runs an AI citation test needs.
Diagnose the page, then observe the outcome.
A readiness audit and a citation test answer different questions. Keep both visible.
Audit a page free