Validate.QA Now Catches Visual Bugs
Cut-off text, covered buttons, broken images and low contrast, found on every page a survey or discovery run visits. No AI needed, built to stay quiet, and never a test failure.
A button that is cut off by the edge of the screen still passes most tests. So does a price that spills out of its card, a "Buy now" button with a transparent overlay on top of it, and a product photo that never loaded. The element is in the DOM, the test finds it, and the assertion is green. A person looks at the page and sees a bug in about a second.
Validate.QA now looks for these. On every page a survey or discovery run visits, it checks the live page for layout problems and lists what it finds, each with a screenshot of the element. The checks need no AI, and they are built to stay quiet on the things sites do on purpose.
What it catches
Five kinds of problem, all things a visitor would notice and an assertion would not:
How it works, without AI
The checks run inside the page itself, in the same browser that is already exploring your app. They look at the real layout of the page as the browser drew it. There is no model in this path, so there is no prompt, no token cost and no waiting on an API.
That is also why it is fast. On our demo page it takes about 150 to 200 ms per page.
Each finding comes with a cropped screenshot of the element, a CSS selector, its position on the page, a plain-English description and a severity of high, medium or low.
How false positives are kept low
A layout checker that reports every truncated label is a checker people turn off. Low noise was the design goal, so the checks skip patterns that sites use on purpose:
Findings are also capped per page, so one broken template can't bury everything else.
We tested this on a demo page with 6 planted bugs and 8 intentional patterns from the list above. It found all 6 bugs and flagged none of the 8 patterns. That is one page we built for the purpose, not a measure of how it will do on your site, so expect to dismiss the odd finding.
Where you see the issues
The optional Gemini AI review
Some visual problems can't be found by measuring: a layout that is technically fine and looks wrong. For those, there is a second pass with Google Gemini that ships in the same release. It starts working once an admin saves the team's Gemini API key in Admin AI Settings, and the same page has a switch to turn it off.
Topics: Visual Checks, Layout, Site Intelligence, MCP.
Read the full article · Get Started Free