The Reject Button Audit: Placement, Wording, and Colour Traps That Fail Review
A reject control fails review on three measurable things: it takes more steps than accept, it is visually demoted to a link or a low contrast grey while accept is a filled button, or its wording obscures what it does. All three are checkable from a screenshot, which is why regulators check them first.
Steps, prominence, wording: the three tests
Count the steps. Accept all sits on the first layer, one click. If reject all lives in a settings layer, that is two clicks minimum plus a save, and the asymmetry is the finding regardless of how reasonable the settings layer looks. Several European regulators now state plainly that a first layer reject option is required where a first layer accept option exists.
Prominence is comparative, not absolute. A reject button the same size, weight, and contrast family as accept passes even in muted colours. A reject rendered as an underlined text link beside a large filled accept button fails, because the design is steering, and steering is the thing the rule exists to stop. Wording completes the picture: reject all and refuse consent say what happens. Maybe later, and close icons that silently accept, do not.
The colour and contrast traps
The classic trap is the grey reject: text in a mid grey on a light panel, technically legible, practically invisible next to a brand coloured accept. Contrast checkers catch part of it. Measure the reject text against accessibility contrast ratios, and then go further and compare the two buttons as a pair, because the legal question is relative prominence, not minimum legibility.
Order steers too. On a left to right reading path, the first filled button collects the reflexive click. Swapping which action sits in the primary visual slot between page loads, or A B testing reject placement for lower refusal rates, is optimisation against the visitor's choice, and an audit trail showing that test will read badly in any proceeding. Banner Fatigue: Repeat Prompts Train Visitors to Consent to Anything covers the mobile version of these traps.
Running the audit and keeping the evidence
Screenshot the banner fresh on desktop and mobile, annotate steps, sizes, and contrast values for both actions, and file the audit with a date. Repeat it after every banner redesign, vendor template update, or consent platform version change, because template updates have quietly demoted reject controls on sites whose owners believed nothing had changed.
Where the audit fails, the fix is almost always an afternoon of configuration rather than a rebuild: promote reject to a first layer button, match its visual weight to accept, and plain word it. The expensive version of this lesson arrives as a regulator letter or a platform compliance warning, and it asks for the same afternoon, plus an explanation.
Keep reading: Banner Fatigue: Repeat Prompts Train Visitors to Consent to Anything A/B Testing Tools Are Data Collection Too: Consent for Experimentation Platforms Consent Banner A/B Tests: What You Can Test Without Breaking the Law.
Frequently asked questions
Does reject all have to be on the first layer?
Where accept all is offered on the first layer, several European regulators, including the French CNIL and the German authorities, expect reject to be reachable with the same ease. A settings only reject path is the most common banner finding and the easiest to fix.
Can reject be a text link if accept is a button?
That pairing is treated as visual demotion: one action is presented as the designed path and the other as a footnote. Match the reject control's size, weight, and interactivity to the accept control. Colour can differ; prominence cannot.
What wording should a reject button use?
Plain verbs that state the outcome: reject all, refuse consent, decline. Avoid wording that hides the action, such as maybe later, or a close icon whose effect is unclear. The visitor should be able to predict exactly what the control does before touching it.
How often should we audit the reject button?
After every banner change, including vendor template updates you did not initiate, and at least quarterly alongside the consent record test. Template updates are the silent cause: a platform release restyles the banner and the reject control loses prominence without anyone deciding it should.