Four dimensions determine how an AI model represents a brand. Here's exactly what each one measures.
Scan your domain across multiple AI engines and see your Recognition, Depth, Accuracy, and Confidence scores in seconds.
Run your free AI Awareness Scan →An AI visibility score is measured across four dimensions — Recognition (does the AI know the brand exists), Depth (how much specific detail it holds), Accuracy (whether what it knows is correct), and Confidence (how certain its language is) — each scored 0-100 per AI engine. A separate Consistency Score measures agreement across engines. There's no universal "good" threshold yet, but consistently scoring above 60-70 on Recognition and Accuracy across most tested engines generally indicates solid foundational visibility.
"AI visibility" is often used loosely to mean "does AI know about my brand," but that's actually four separable questions, not one. Each gets its own score because a brand can score very differently on each — high Recognition with low Accuracy is a genuinely different problem than low Recognition with nothing yet to be accurate or inaccurate about.
This is the floor — a brand with zero Recognition isn't being misrepresented, it simply doesn't exist yet in that model's knowledge, and nothing else can be meaningfully scored until this is established.
Products, positioning, history, what makes the brand distinct — as opposed to a vague, generic acknowledgment that could describe almost any company in the category.
A model can have strong Recognition and substantial Depth while still being wrong about specific facts — outdated information, a factual error, or conflation with a different, similarly named entity.
Hedged, uncertain phrasing versus direct, assured statements. Tracked separately from Accuracy specifically because a model can be highly confident while stating something incorrect.
There's no tally of how many times an AI mentions a brand to increase. Current consumer AI models don't expose a countable citation metric — what's measurable is the model's internal representation, which is what these four dimensions capture instead.
Each of the four dimensions is scored separately per AI engine, not as one blended average — Meta AI, Google AI, Mistral, and Gemma are tested individually (free), with ChatGPT and Claude available as well if you provide your own API key. This matters because the same brand routinely scores very differently across engines, since each model was trained on different data with a different cutoff and different weighting of sources.
A brand showing strong Recognition and Accuracy on Google AI but near-zero Recognition on Mistral isn't a measurement error — it's an accurate reflection that these models simply don't know the same things. Treating one engine's result as representative of "AI" broadly is the most common misreading of this kind of score.
One scan tests Recognition, Depth, Accuracy, and Confidence across Meta AI, Google AI, Mistral, and Gemma simultaneously.
Run your free AI Awareness Scan →Separate from the four per-dimension scores, a Consistency Score measures how much the different AI engines agree with each other about a brand. High consistency means the engines broadly align on what they know; low consistency means a brand is represented very differently depending on which AI a person happens to ask.
| Metric | What it measures | Scored | Priority |
|---|---|---|---|
| Recognition | Does the model know the brand exists at all | Per engine | Check first |
| Depth | How much specific detail the model holds | Per engine | After Recognition |
| Accuracy | Whether what the model states is correct | Per engine | Urgent if low |
| Confidence | How certain the model's language is | Per engine | Follows others |
| Consistency | How much all engines agree with each other | Cross-engine | Check separately |
A brand could score moderately on Recognition across the board (consistent but unremarkable) or score very high on one engine and very low on another (inconsistent, even if the average looks fine). Averaging would hide this; the Consistency Score is built specifically to surface it.
This is a newer category of metric without the kind of long-established, widely agreed benchmarks that exist for something like traditional domain authority. That said, a few practical reference points hold up reasonably well:
Recognition below 30-40 on most engines generally means the brand has minimal foundational presence and other dimensions aren't yet meaningful to optimize.
Recognition above 60-70 with Accuracy in a similar range, consistently across most engines, indicates solid foundational visibility.
High Confidence with low Accuracy is the specific combination worth the most immediate attention — the model is stating something wrong with conviction, not hedging about something it's unsure of.
| Score pattern | What it means | Urgent? | First action |
|---|---|---|---|
| Recognition < 30 | Brand barely exists in model's knowledge | ✗ Build first | Establish public entity info |
| Recognition > 60 + Accuracy > 60 | Solid foundational visibility | ✓ Healthy | Maintain and deepen |
| High Confidence + Low Accuracy | Model states something wrong confidently | ✗ Most urgent | Correct misinformation in public sources |
| Good individual scores + Low Consistency | Engines disagree significantly | Medium | Audit per engine individually |
There's no established direct link demonstrated yet between this score and sales, traffic, or conversion outcomes. It measures how AI models represent a brand internally — a distinct, newer category from traditional web analytics.
The process is the same regardless of brand size or industry.
One scan tests your domain across Meta AI, Google AI, Mistral, and Gemma simultaneously, with ChatGPT and Claude available if you add your own API key.
Look at Recognition first — if it's low on a given engine, that's the priority before Depth, Accuracy, or Confidence are meaningful to address on that engine specifically.
A low Consistency Score, even with decent individual numbers, points to fragmented representation worth addressing engine by engine rather than as one number.
A single blended "AI visibility" number would hide more than it reveals. A brand with strong Recognition and weak Accuracy needs a completely different response than one with weak Recognition and nothing yet to be accurate about — fix the wrong thing first and the score won't move.
This is also why the score is broken out per engine rather than averaged. Optimizing for an average can mean a brand looks "fine" overall while being essentially invisible on one engine entirely — a gap the Consistency Score exists specifically to catch.
Run the same four-dimension scan described in this guide on your own domain — Recognition, Depth, Accuracy, Confidence, and Consistency, in one pass.
Run your free AI Awareness Scan → Tech stack checkerSee your Recognition, Depth, Accuracy, and Confidence scores, or check what technology any website is built on.
Run your free AI Awareness Scan → Tech stack checker →A single blended "AI visibility" number would hide more than it reveals, since a brand with strong Recognition and weak Accuracy needs a completely different response than one with weak Recognition and nothing yet to be accurate about — which is why this is measured as four separable dimensions plus a distinct cross-engine Consistency Score, not one aggregate figure.
Each dimension is scored 0-100 per AI engine based on that engine's actual generated response when queried about the brand, tested across Meta AI, Google AI, Mistral, and Gemma by default, with ChatGPT and Claude available via your own API key. The Consistency Score is calculated by comparing dimension scores across all tested engines for the same brand.
Need a specific analysis tool for your workflow? Describe it below and we'll build it.