Consent Banner Testing on Mobile and Desktop: Two Different Audits
A banner that behaves on desktop can cover the screen, trap scroll, or hide its own reject button on mobile. Why consent testing needs two passes, and what each one catches.
Most consent testing happens where the tester sits: a desktop browser, a wide screen, a mouse. Most visitors arrive somewhere else: a phone, one handed, on a small viewport where the same banner occupies half the screen and its buttons stack in an order the designer never saw.
This guide treats mobile and desktop consent as the two separate audits they are: what to check on each, and the defects that only exist on one side of the divide.
Mobile: space is the whole problem
On a phone, a banner competes with the page for a few inches of glass. The failures follow: the banner covers the content entirely with no way to scroll past, the reject option falls below the fold of the banner itself, or the close control is a target too small to hit reliably with a thumb.
Test at the smallest common viewport, in portrait, with the browser chrome visible, because that is the real estate the banner gets. A banner that only behaves in a maximised test window behaves for nobody.
Mobile: touch, zoom, and accidental consent
Touch changes the ethics of the click. Buttons placed where a scrolling thumb lands harvest accidental consent, and a banner that shifts position as the page loads can slip accept under a tap aimed at content. Accidental consent is not consent, and layouts that engineer it are visible in the data as implausibly instant acceptances.
Zoom behaviour matters too. A banner that cannot be read at 200 percent zoom, or that traps the page zoomed inside itself, fails visitors who enlarge text as a matter of course. Consent must be readable to be meaningful.
Desktop: density hides different defects
Desktop banners fail by subtlety: the reject option present but styled into invisibility, settings layers that open off screen on smaller laptop windows, and keyboard traps in preference centres that a mouse using tester never meets.
Run the desktop pass with keyboard only and at a 13 inch laptop size as well as a large monitor. Preference centres deserve their own pass: every toggle reachable, labelled, and operable, with the save action confirmed in text.
One defect log, tagged by surface
Keep findings in one log tagged mobile, desktop, or both, with the viewport or device noted. Patterns emerge quickly: most stores discover their consent defects are overwhelmingly mobile, which is another way of saying they are overwhelmingly where their visitors are.
Retest both surfaces after every banner change and every major site release. The banner is a guest in the page layout, and layout changes evict its assumptions on both surfaces without asking permission.
Related reading
If this topic is useful, these earlier guides go deeper on connected questions: Consent Banners in App Like Stores That Never Reload a Page, A Consent Banner Accessibility Audit You Can Run in an Hour, and Consent Banner Reject and Accept Parity: What Equal Prominence Really Means.
Frequently asked questions
Why test consent separately on mobile and desktop?
Because the failure modes differ. Mobile fails through space: coverage, hidden reject options, accidental taps. Desktop fails through subtlety: invisible styling, off screen settings, keyboard traps. One pass cannot see both sets.
What is the first mobile check?
At the smallest common viewport, confirm the banner can be read, scrolled past or dismissed, and that accept and reject are both reachable without hunting. If a visitor cannot decline on a phone, nothing else about the banner matters.
Can accidental taps count as consent?
No. Consent produced by a shifting layout or a button placed under a scrolling thumb is not a freely given choice, and instant acceptance patterns in the data are a warning sign worth investigating, not celebrating.
How often should both surfaces be retested?
After every banner change and every significant site release, because the banner depends on the page layout around it. Both surfaces, both times, with the findings logged by surface.