Labs

Heuristics Evaluator

lab · running

We give it a public web page or a screenshot, and it returns a graded report measured against more than 605 heuristics drawn from Nielsen, Morville, Travis, and Revang.

It works in four passes we can watch arrive, reading the page, running the accessibility checks, matching what it finds against the corpus, and writing it up, in about forty five seconds. Every finding says how we know it, either measured on the page in front of us or matched from the corpus as something worth a look. We keep those two apart on purpose, because a pattern we have not observed on your page is a question rather than a fault.

Checking the Labs AI…

This tool runs its evaluation in the browser against our workshop, so it needs JavaScript switched on.

What we look at A web pageGive us a public address and we read the page itself, counting every element we check. A screenshotUpload an image and we review what is visible in it. PNG, JPEG, or WEBP, up to 10MB.

Page address

How deep we go Standard passFour passes, about forty five seconds. We read it, run the accessibility checks, match the corpus, and write it up. Expanded passFramework by framework, with visual diagnostics on the image, an edge map of where weight lands, a balance reading, and a density heatmap. Every finding names its framework and a specific fix, sorted by an impact score so the list reads as a remediation order.

The standard pass is open to everyone. For an expanded pass, reach out, or buy a Pup Cup for the kittens.

← Back to Labs