# Brand perception scores

> How GetIntel's knowledge and recommendation scores are calculated from each engine's answers, with the points behind every result.

Source: https://getintel.ai/docs/academy/metrics/brand-perception-scores/

[Brand perception](/docs/features/brand-perception/) produces two scores out of 100, each averaged across ChatGPT, Perplexity, Gemini and Google AI Overviews. It runs weekly, on Mondays.

## Knowledge: does AI know you?

Each engine is asked directly what your product is. Its answer earns points:

| Result | Points |
| --- | --- |
| Accurately describes you | 100 |
| Partially accurate | 60 |
| Vague / unsure | 30 |
| Describes you wrong | 10 |
| Doesn't know you | 0 |

Knowledge = the average across engines.

## Recommendation: would AI recommend you?

Two turns per engine, blended **40% unprompted, 60% when asked**.

**Unprompted**, a category question that never names you:

| Result | Points |
| --- | --- |
| Named you first | 100 |
| Named you | 65 |
| Named nobody | 40 |
| Named competitors, not you | 20 |

**When asked**, a follow-up that names you:

| Result | Points |
| --- | --- |
| Recommends you | 100 |
| Lists you as an option | 60 |
| No clear signal | 30 |
| Points to competitors | 20 |
| Speaks negatively | 0 |

**Worked example.** One engine names you (65) unprompted, then lists you as an option (60) when asked: 65 × 0.4 + 60 × 0.6 = **62**. Recommendation = the average of these per-engine scores.

## What moves them

- **Knowledge** improves when there is clear, consistent information about what you do across your site and the places AI reads, such as review sites and directories.
- **Recommendation** improves with the same levers as [win rate](/docs/academy/metrics/win-rate/): comparisons and reviews that show why you are the better choice.