How we analyze
Each week, this site captures the screen of each AI product right after signing in, on a desktop screen 1280 by 800 pixels, before scrolling. Each screenshot then gets a short design analysis.
What we look at
- Layout: where the main areas sit, such as the sidebar and the message box, and how much of the screen they take.
- What stands out most: which items carry the most visual weight, based on their size and contrast.
- Text and color: whether text is light or dark enough against its background to meet common accessibility guidelines.
- Where you start: the message box, suggested prompts, and whether the product says it can make mistakes.
How we measure
Colors, contrast and sizes are measured from the screenshot pixels with a script, using the contrast formula from the Web Content Accessibility Guidelines (WCAG 2.2). Other observations, such as what stands out most, are made by an AI model following a fixed checklist. A person reviews every analysis before it is published.
What we don't claim
An analysis only describes what is visible in one screenshot. It does not judge whether a product is easy to use, predict what people will do, or certify that a product meets accessibility standards. Measurements taken from a screenshot are close estimates, not official test results. Mobile screenshots are shown but not analyzed.
Week-to-week changes
When the previous week's analysis exists, the two weeks are compared. Differences that could come from the account, the time of day or a rotating greeting are not reported as redesigns.
Special cases
Kimi is captured while signed out, because no account is used for it. Muse always opens on a conversation, so only the interface around it is analyzed.
Privacy
Screenshots come from real accounts. Names, conversation titles and other account details are left out of every analysis, and are covered in the screenshots where possible.