Visibility Metrics
These metrics sound similar and get confused constantly. Here’s what separates them.
The Short Version
| Metric | Answers the question |
|---|---|
| Mention rate | How often does AI talk about us at all? |
| Citation rate | How often does AI pull from our website? |
| Share of voice | Of all the brands AI names here, how much of that airtime is ours? |
| Visibility score | One number combining everything, for tracking over time. |
Mention and citation are your anchor metrics. They’re direct, they’re measured the same way every report, and they’re the ones to quote when someone asks how you’re doing. Share of voice adds competitive context. The visibility score is a roll-up for spotting trends, not a number to optimize in isolation.
Mention Rate
The share of AI responses that name your brand.
A mention counts when AI refers to you by name in its answer - whether that’s recommending you, listing you among options, or just using you as an example. It doesn’t matter whether the AI linked to you.
This is the most basic question: are you in the conversation at all? If your mention rate is low, nothing else can improve much, because rankings and recommendations only happen once you’re named.
Citation Rate
The share of responses where AI references your website as a source.
Citation is about your content rather than your name. The AI went to your site, drew on what it found, and pointed at it. That usually means your pages are structured clearly, answer the question directly, and are credible enough to lean on.
Mention and citation come apart in revealing ways. Mentioned often but rarely cited means AI knows your name but isn’t reading your content - a content problem. Cited but rarely mentioned means your content is useful but your brand isn’t landing as a recommendation - a positioning problem.
Share of Voice
Your slice of the attention in the responses where anyone gets named.
Where mention rate compares you against every response we collect, share of voice compares you against the other brands in those responses. If AI names five companies and you’re one of them, you hold a portion of that airtime. It’s the difference between “how visible are we” and “how visible are we relative to the field.”
We report two versions, because they answer different questions:
- Mention share of voice - your share of all the brand names AI brings up. Airtime.
- Citation share of voice - your share of all the sources AI actually cites. Credibility.
Compare the two. Citation share of voice running ahead of mention share of voice means AI leans on your content more than it says your name - you’re doing the work and a competitor may be getting the credit. The reverse means your name travels further than your content does.
Share of voice is measured against the competitors you track. Add a major competitor and your share will drop; remove one and it rises - without anything changing in the real world. Keep your competitor list stable if you want to compare share of voice across reports.
Ranked and Recommended
Two higher-intent signals that only apply once you’ve been mentioned:
- Ranked rate - how often you appear in a “top X” or “best of” list, and where you land in it.
- Recommended rate - how often AI actively puts you forward as a solution, rather than just naming you.
These carry the most buying intent, but they appear in fewer responses than mentions do, so they move around more between reports. Treat them as signals of quality, and mention and citation as your steady baseline.
Visibility Score
A single composite number rolling up everything above. It’s the most complicated metric we report, so here’s exactly what feeds it.
1. Each response earns points from four signals
We score every AI response individually, then average across all of them. A single response can collect:
- Being mentioned - by far the largest component, roughly half of what a response can earn. Presence is the thing the score rewards most.
- Being ranked - the next largest. Appearing in a “top X” or “best of” list is worth more than a passing citation.
- Being cited - a smaller share.
- Being recommended - a smaller share, equal to citation.
A response where AI names you, ranks you, cites you, and recommends you earns full marks. One that merely names you earns about half. One that ignores you entirely earns nothing.
2. Ranked and recommended depend on being mentioned
You can’t be ranked in a list that doesn’t name you, so those two signals only count in responses where you’re mentioned.
This is why mention rate sets the ceiling on your score. If AI rarely names you, the majority of the available points are out of reach no matter how good your content is.
Citation is the one signal that doesn’t require a mention. AI can draw on your page as a source without naming your brand in the answer - so your content can be earning credit even in responses where you’re invisible.
3. Platforms are weighted by how much they’re used
Each platform’s average is then weighted by its share of real-world AI usage. A strong showing on a platform millions of people use every day moves your score more than the same showing on a niche one, so your number reflects reach rather than treating every platform as equal.
We keep these weights current as the market shifts. If your buyers skew toward one platform, we can tune the weighting for your brand - and each report stores the weights it used, so past reports stay comparable even when the weighting changes.
4. Only platforms with data count
If a platform returns nothing for a report, it’s left out of the calculation rather than counted as a zero. A collection issue on one platform won’t crater your score - the remaining platforms are simply re-balanced between themselves.
Use the visibility score for direction over time - is the trend up or down. Because it blends four signals across four platforms, a single number can’t tell you why it moved. For that, look at mention rate, citation rate, and share of voice, which each point at a specific problem.
Platform-by-Platform Breakdown
Your report shows each metric separately per AI platform, which reveals patterns a single number hides:
- Strong everywhere: consistent presence across platforms.
- Strong on some, weak on others: platforms favor different sources. Your actions will call out platform-specific gaps.
- Blind spots: near-zero presence on a major platform is a critical gap. Being invisible on the platform most of your buyers actually use means missing most of the market, even if you’re strong everywhere else.
Ranking Performance
When you do appear in “top X” or “best” lists:
- Average Position: your typical ranking. Lower is better - #1 is ideal, #5 means you’re often an afterthought.
- Best Position: your highest ranking achieved. Hitting #1 proves AI can position you as a leader.
Average position matters more than best position. Consistently ranking #2-3 beats occasionally hitting #1 but usually appearing at #7.
Visibility by Journey Stage
We track these metrics across buyer journey stages:
- Problem Recognition: Are you mentioned when buyers discover their problem?
- Solution Research: Are you included when buyers explore options?
- Vendor Evaluation: Are you recommended when buyers compare vendors?
- Competitive Intelligence: Do you hold up when buyers dig into head-to-head comparisons?
Most companies should prioritize vendor evaluation - that’s where buying decisions happen. If you’re creating a category, solution research may matter more.
Interpreting Changes
Your numbers fluctuate as AI models update. Small changes (+/-5%) are normal. Larger shifts warrant investigation.
Went up? Check whether you published content that’s now being cited, earned a mention somewhere major, or a competitor slipped.
Went down? Check whether a competitor pushed hard, new negative coverage appeared, models updated - or whether you changed your tracked competitor list, which moves share of voice on its own.
Common Questions
What’s a “good” score?
It depends on your market. In crowded categories, 15-20% mention rate might be excellent. In niche markets, you might expect 40%+. Share of voice is the better yardstick, because it’s already relative to your competitors rather than an absolute number.
Which metric should I report to my team?
Mention rate and citation rate for how you’re doing, share of voice for how you’re doing against competitors. Keep the visibility score for the trend line.
Why does my visibility differ between AI platforms?
Each has different training data, recency bias, and source preferences. Perplexity leans on recent web content; ChatGPT relies more on training data. Those differences tell you where to focus.
My rates improved but the score barely moved. Why?
Usually the gains landed on a platform with a smaller share of AI usage, so they carry less weight in the composite. The other common cause is improving citations while mention rate stayed flat, since mention is the largest component and gates the ranked and recommended points.
Why is my share of voice missing for some slices?
If no brand - yours or a competitor’s - was named in that slice of responses, there’s no airtime to divide, so we show nothing rather than a misleading zero.
How do I improve a specific metric?
We generate targeted actions based on which are weak. Low citation rate triggers source and content work; low ranked rate triggers comparison content. Approve them and the Engine does the work.