Hands-On With Consent Banners: What the Headlines Miss About Real Privacy Choices

Archi3 min read

privacyconsenttrackingregulation

Contents
  1. The Banner Is Not the Boundary
  2. What I Did
  3. The Headlines Miss the Default
  4. The Post-Reject Traffic
  5. What Regulators Actually Check
  6. Three Settings I Changed
  7. The Cost of Getting It Wrong
  8. What to Watch Next

The Banner Is Not the Boundary#

Last month I spent two weeks testing consent banners on 50 U.S. news, retail, and health sites. I clicked "Reject All" where it existed, cleared cookies, and watched network traffic. The headlines say consent rates are low and regulators are starting to notice. My logs show something else: the reject button is often a decoy, and the tracking that matters happens after you click it.

What I Did#

I used a clean browser profile for each site. I recorded every network request for 60 seconds after the banner appeared, then again after I clicked reject. I also captured the banner's DOM to see which buttons were real and which were links to a settings page that defaulted to on.

On 22 of the 50 sites, the "Reject" button did not stop analytics and advertising scripts from loading. On 9, the reject button was a link that opened a preferences panel with all toggles already set to "on." On 4, the banner reappeared on every page load until I accepted.

The Headlines Miss the Default#

Most coverage focuses on the consent rate: the share of visitors who click accept. But the rate is a symptom. The cause is the default state. When the default is "on" and the reject path takes three clicks, the rate tells you more about design than about user preference.

I measured time-to-reject on 30 sites. The median was 11 seconds. The fastest was 4 seconds (a single button). The slowest was 47 seconds, which required scrolling through a vendor list of 312 partners. That is not a choice. That is a maze.

The Post-Reject Traffic#

On the 22 sites where scripts loaded after reject, the requests came from three sources: a tag manager that fired before the consent check, a server-side container that ignored the client flag, and a pixel embedded in the page HTML. In each case, the banner said the site respects your choice. The network log said otherwise.

One retail site loaded a session recording script after reject. The script captured keystrokes in the search box. Another site loaded a social pixel that sent the page URL and a hashed email if you were logged in. These are not edge cases. They are the default configuration of popular tag templates.

What Regulators Actually Check#

I pulled enforcement actions from the past 18 months. Most focus on the banner text and the presence of a reject button. Few look at what loads after the click. That gap is where the real risk sits. A company can pass a banner review and still send data to ad partners without consent.

My tests found that the most common failure is not a missing button. It is a tag that fires on page load before the consent management platform (CMP) initializes. The fix is technical: move the tag behind the CMP callback and test with a clean profile. But the audit rarely goes that deep.

Three Settings I Changed#

After the tests, I changed three things on my own sites and recommend them to anyone running a CMP.

  1. Block all tags by default. Allow only after the CMP callback fires. Test with a network log, not a screenshot.
  2. Make reject a single button with the same visual weight as accept. No settings panel, no scrolling vendor list.
  3. Log consent state server-side with a timestamp and the exact vendor list version. If a regulator asks, you can show what the user saw.

The Cost of Getting It Wrong#

The fines are not the biggest cost. The bigger cost is the audit trail you cannot produce. If a user rejects and your tag manager still fires, you have no defense. I saw this on 22 sites. The banners looked compliant. The logs told a different story.

What to Watch Next#

Regulators are starting to ask for network logs, not just banner screenshots. That changes the test. Companies that treated consent as a UI problem will need to treat it as an engineering problem. The ones that already do will have an easier time.

For now, the headline number is the consent rate. The number that matters is how many scripts fire after reject. On my sample, it was 44 percent of sites. That is the story the headlines missed.

More articles