How Design Check works
A quick look at what powers each part of the review.
Design Check looks at four areas. Some are measured exactly, some are read by an AI brand director, and most use both. The score stays behind the scenes. You see a plain verdict and the exact spots to fix.
| Area | How we check it | Depth |
|---|---|---|
| Colors | We sample the actual colors in the design and measure each one against the brand palette and approved gradients by perceptual distance. Anything with no close brand match is flagged, and the score is capped when none of the brand's signature colors are present. Runs: Color sampling, exact palette and gradient matching, signature-color check, plus an AI palette read. | Strongest. Mostly exact measurement, backed by AI. |
| Type | An AI brand director identifies the typefaces in each text element and compares them to the brand's fonts. It's most reliable at catching type that's clearly off-brand — a serif, script, or slab where a clean sans belongs — and less certain when telling the brand fonts apart from a very similar sans-serif. Runs: AI typeface identification, per element. | Good at clear mismatches. AI judgment, no exact font math yet. |
| Copy | We read the words in the design, then run them through the brand voice linter for banned terms and hype phrases, alongside an AI read of the messaging. Runs: Text reading (OCR), exact banned-term and voice rules, plus an AI messaging read. | High. Hard rules plus AI. |
| Imagery | An AI brand director judges the tone of any photography or illustration against the brand. When the design has no imagery, there is nothing to check and we say so. There is no dedicated imagery model yet and no separate image-present detector, so this is the lightest of the four. Runs: AI tone read only. | Lightest. The most room to grow. |
Where there is nothing of a kind to check (no imagery on a type-only design, no text on a photo), Design Check says so plainly rather than calling it a pass it did not earn.