What does an automated checker do with a cookie notice? Ours fooled it

By Albert Garcia Pujadas · Published on November 10, 2026.

In short: on November 3 I measured uxmachine.app, and the first thing the report asked me to change was a form. That form wasn't part of the website: it was the checkboxes of our own cookie notice. UXMachine didn't recognise the notice and treated it as part of the page. We fixed it that same day and wrote down what it does with any cookie notice.

The recommendation that wasn't

Every report has a section called "If you change only one thing". That day it said, in Spanish, the language the site was measured in: "Mark which fields in this form are required". The form asked for two things and didn't say which were needed. (The quotes from the report are my translation.)

I went looking for it on the home page and couldn't find it. The report's evidence said where it was: next to the "Rechazar todo" ("Reject all") button, two checkboxes. It was the settings panel of our cookie notice, the one where you choose which cookies to accept.

So the checker didn't see that it was a notice, counted its checkboxes as a form on the website and, out of everything it measured, that was the first thing it told me to change. An agent following the report to the letter could have started marking the cookie checkboxes as required.

In the report's detail, an identical form, asking for two things with no markings, showed up on eight of the ten pages.

The same report said it couldn't tell whether there was a cookie notice on any of the ten pages. I wrote about that in another note, from the angle of comparing measurements. What matters here is something else: a notice that isn't recognised doesn't just stay open. It also leaks into everything else that gets measured.

What we changed

That same day we changed two things:

UXMachine recognises notices built like ours, on any website. It isn't an exception for our domain.
What's inside a cookie notice, including its settings panel even when it's hidden, no longer counts as a form on the website.

I measured again a little over an hour later. The notice was recognised and rejected, and the report says so: "The cookie notice was rejected once, and that rejection was kept on the other pages measured on this site". The two-field form with no markings no longer showed up on any page. The first recommendation became a different one: "Make the main action stand out from the others".

Careful not to read that as the site getting better. The site didn't change: the way we measure it did. That's why the comparison between the two measurements comes out "not comparable".

If the notice has a button to reject, we reject. If it doesn't, we accept.
We keep that choice on the other pages of the same site, as a person browsing would.
If the notice can't be closed, we hide it to measure the page underneath and say so in the report.
If the notice blocks the whole page, that page isn't measured.
The home page is always measured without earlier cookies.
What's inside the notice isn't measured as if it were part of your website.

Why reject first? There's no right answer. We picked one, we apply it the same way on every website and we say it in the report. What matters is that two measurements of the same website are made the same way, because otherwise they can't be compared.

What this doesn't fix

When we hide a notice that won't close, we measure a page nobody sees that way.
There will be notices we still don't recognise. When that happens, the report says it couldn't tell whether there was a notice. If you read that in your report, be wary of what it says about forms and the first screen.
The accessibility of the notice itself, and what the page loads before you answer it, we only check on the home page.
This is what I saw on one website, ours. I don't know how often the same thing happens to other checkers.

If you check your website with an automated checker or an agent

Look in the report for what it did with the cookie notice. If it doesn't say, you don't know what was measured.
Before changing a form because you were told to, open the evidence and check that the form is yours.
If your website has a notice, always measure the same way: the same address, the same language, and the earlier measurement to compare with.

The data

First measurement of uxmachine.app, November 3, 2026, at 14:35 (Madrid time). Cookie notice: we couldn't tell whether there was one. First recommendation: "Marcar qué datos de este formulario son obligatorios" ("Mark which fields in this form are required"), on the home page, next to "Rechazar todo".
Second measurement of uxmachine.app, November 3, 2026, at 15:52. Cookie notice: rejected once, and that rejection was kept on the other pages. First recommendation: "Diferenciar la acción principal de las demás" ("Make the main action stand out from the others").
In each measurement, ten pages, viewed on a computer (1280×800) and on a phone (390×844).
Comparison: not comparable on all 17 rules, because the way we measure changed.
How we measure.

← All notes