Research
Clicking Reject Often Leaves the Tracking Cookies in Place - Here's the Data
GDPR Privacy Monitor Research · 2026-04-11 · 6 min read
Every cookie banner makes the same implicit promise: click "Reject" and the tracking stops. The banner disappears, and you browse in peace. That is the bargain the consent model is built on -- the idea that users have a meaningful, enforceable choice.
We tested that promise on 28,891 EU websites. On 22,445 of them the test gave a clear result, and on nearly half of those (48.3%) the promise is broken: non-essential cookies are still there after the user explicitly clicks reject. Advertising pixels keep firing. And on 1,743 sites, cookies that the reject action initially clears come back the next time the page loads -- a pattern we call consent respawn.
This is not a marginal finding about a few misconfigured sites. It points to a systemic issue in the mechanism that hundreds of millions of EU residents rely on for their privacy rights.
What "Reject" Is Supposed to Do -- Legally
The legal framework is clear. Under Article 5(3) of the ePrivacy Directive, storing or accessing information on a user's terminal equipment requires prior consent, with a narrow exception for cookies strictly necessary for a service the user has explicitly requested. GDPR Article 7(3) adds that withdrawing consent must be as easy as giving it, and the data controller must act on that withdrawal.
The EDPB's Guidelines 05/2020 on consent under GDPR spell out what this means for reject buttons: if a website offers an "Accept All" option, it must also offer an equally prominent and equally accessible option to refuse. And that refusal must be effective -- it must actually prevent the processing that consent would have authorized.
The CJEU reinforced this in Planet49 (C-673/17): consent requires a clear affirmative act, and pre-ticked checkboxes do not constitute valid consent. The logical corollary is that a refusal -- an explicit negative act -- must be respected as definitively as an acceptance. If clicking "Accept" turns tracking on, clicking "Reject" must turn it off. Not partially. Not temporarily. Completely and persistently.
That is what the law says. Our data shows what actually happens.
How We Test the Reject Flow
Our reject-flow test is designed to be as close as possible to the experience of a real human user clicking reject, with the addition of comprehensive instrumentation. Here is the process step by step:
Step 1: Fresh browser session. The scanner launches a clean Chromium instance -- no cookies, no local storage, no cached resources, no browsing history. This is a genuine first-time visitor with no prior consent state. Step 2: Navigate and record pre-consent state. The scanner loads the target URL and waits for the page to stabilize. Before any interaction, it records every cookie present in the browser, every network request made, and every third-party domain contacted. This is the pre-consent baseline -- what the site does before the user has any opportunity to make a choice. Step 3: Detect the consent banner. Using our detection library covering 45 known CMPs and generic heuristics, the scanner identifies the consent mechanism and locates the reject button. This step uses a layered approach: first checking for known CMP script signatures and their associated DOM structures, then falling back to generic detection of fixed-position elements with consent-related text and button labels like "Reject All," "Decline," "Refuse," or equivalent translations. Step 4: Click reject. The scanner clicks the reject button and waits for the page to settle. "Settle" means waiting for pending network requests to complete, DOM mutations to stop, and any animations or transitions to finish. We use generous timeouts to avoid false positives from slow CMP implementations. Step 5: Post-reject snapshot. The scanner records the complete state again: all cookies, all network activity, all third-party domains. This snapshot is compared against the pre-consent baseline to determine what changed. Step 6: Reload and check for respawn. The scanner reloads the page and waits for it to stabilize again. It then takes a final snapshot. This step is critical: it reveals whether the reject action is persistent (respected across page loads) or whether cookies and tracking re-activate when the page loads fresh. If cookies that were removed in Step 5 reappear in Step 6, that is consent respawn.Every step produces timestamped, archival evidence: full cookie inventories with names, values, domains, expiry dates, and flags; complete network request logs with URLs, methods, response codes, and timing; and full-page screenshots. This evidence chain allows any individual finding to be verified and challenged.
The Results: Nearly Half Keep the Cookies
We found and clicked a reject button on 28,891 websites. On 6,446 of them (22.3%) the result was inconclusive, for example because the page could not be reloaded and verified, so we leave them out. Of the remaining 22,445 conclusive tests:
- 5,650 (25.2%) passed. After clicking reject, non-essential cookies were removed, tracking requests stopped, and the rejection persisted across page reloads.
- 16,795 (74.8%) failed. Non-essential cookies stayed, or requests to tracking services continued, after the user explicitly clicked reject. This count includes requests that may be Google Consent Mode cookieless pings (see the caveat below).
The failures are not uniform. They fall into distinct categories that overlap -- a single site can exhibit multiple failure modes simultaneously.
Persistent cookies: 10,848 sites
On 10,848 sites (48.3% of the conclusive tests), non-essential cookies remained in the browser after the reject flow completed. These are cookies classified as analytics, marketing, or tracking-related that should have been removed or prevented from being set after a reject action.
The most common persistent cookies belong to Google Analytics (`_ga`, `_gid`, `_gat`), Meta/Facebook (`_fbp`, `fr`), and various advertising platforms. In many cases, the CMP successfully removes some cookies but fails to catch all of them -- suggesting incomplete integration between the consent management platform and the site's full roster of third-party scripts.
The presence of these cookies after reject is not merely an aesthetic problem. Each one is a persistent identifier that ties the user's current session to past and future visits. The `_ga` cookie, for instance, is a unique client identifier that Google Analytics uses to build a longitudinal profile of the user's behavior across sessions. If it persists after reject, the user's browsing continues to be tracked and aggregated regardless of their expressed refusal.
Persistent trackers: 15,662 sites
On 15,662 sites (69.8% of the conclusive tests), tracking services remained active after reject -- meaning the browser continued making requests to known tracking domains even after the user clicked the reject button.
This is a distinct failure mode from cookie persistence. A tracker can be active without setting a cookie: pixel requests, beacon calls, and script loads all transmit information (at minimum, the user's IP address, the referring page, and timing data) to third-party servers regardless of whether a cookie is stored. This is sometimes called "cookieless tracking" and is increasingly common as browsers tighten restrictions on third-party cookies. One caveat: on 2,341 sites the only signal was a request to Google Analytics or Google Ads, which can be a Google Consent Mode "cookieless ping" that some readers would not count as tracking.
The gap between the cookie figure (10,848) and the tracker figure (15,662) is significant. On 5,947 sites, the reject action successfully cleared cookies but failed to prevent tracking requests from continuing. The CMP handled cookie cleanup but did not block the scripts that generate tracking traffic in the first place. This points to a common architectural problem: the CMP is configured to manage cookies (delete them on reject, prevent them on load) but not to manage script execution.
Consent respawn: 1,743 sites, 5,226 cookies
Consent respawn is the most troubling pattern we identified. On 1,743 websites, cookies that were successfully removed by the reject action reappeared after the page was reloaded. Across those sites, 5,226 individual cookies exhibited this respawn behavior.
Consent respawn is not a bug in one CMP or one configuration. It is a structural problem in how consent is implemented across the web stack.
How a cookie comes back after deletion. When a user clicks reject, a properly configured CMP does two things: it deletes existing non-essential cookies (by setting their expiry to a past date), and it updates a consent record (usually a cookie itself, or a localStorage entry) that tells the CMP's blocking logic to prevent non-essential scripts from running on future page loads. If both of these work, the user is clean: no tracking cookies, no tracking scripts.Consent respawn happens when the deletion succeeds but the blocking fails. The most common scenario:
1. The CMP deletes the `_ga` cookie when the user clicks reject.
2. The CMP sets a consent-state cookie recording that the user has refused analytics.
3. On the next page load, the CMP checks its consent state and knows the user refused.
4. But: the Google Analytics script tag is embedded directly in the page template (outside the tag manager) or is loaded by a tag manager rule that does not check consent state.
5. The GA script loads, sees no `_ga` cookie, and creates a new one.
6. The user's reject decision has been overridden.
Variations on this pattern include: third-party scripts loaded via hardcoded `<script>` tags that the CMP does not control, server-side `Set-Cookie` headers that are sent regardless of client-side consent state, and advertising scripts that re-fire on every page load irrespective of the consent API's return values.
The legal implications are severe. Consent respawn means the website is not merely failing to implement rejection -- it is actively undoing the user's explicit refusal. Under GDPR Article 7(3), the controller must ensure that withdrawal of consent is effective. A system that deletes cookies on reject but allows them to be immediately recreated does not meet this standard. Whether the cause is technical negligence or deliberate design, the user's rights are equally violated.
The CMP Problem: Whose Fault Is It?
An obvious question: is this the CMP's fault or the website operator's?
The answer, in most cases, is both -- but primarily the website operator's. Major CMPs like Cookiebot, OneTrust, and Usercentrics provide the mechanisms to block scripts before consent and remove cookies on rejection. But those mechanisms only work if every non-essential script on the site is routed through the CMP's blocking logic. In practice, this requires disciplined tag management: every script must be categorized, every tag manager rule must check consent state, and no tracking code should be embedded directly in the page template outside the CMP's control.
The reality on most websites is messier. Marketing teams add tracking pixels directly to the site. Developers embed analytics scripts during testing and forget to route them through the CMP. Third-party integrations load their own sub-resources that the CMP does not know about. The result is a consent mechanism that controls some tracking but not all of it -- a partial barrier that looks complete from the CMP's dashboard but leaks from the browser's perspective.
This is also why the 11.9% Consent Mode mismatch rate matters. On these sites, the CMP reports to Google's API that consent was denied, but the browser is still making tracking requests. The CMP believes it has done its job. Google's systems believe consent was denied. But the actual network behavior tells a different story. These mismatches are invisible to anyone looking at dashboards or consent logs. They are only visible when you observe what the browser actually does -- which is exactly what our scanner does and what regulators increasingly audit.
What a Working Reject Flow Looks Like
For contrast, here is what a properly implemented reject flow produces:
1. Pre-consent: The page loads. No non-essential cookies are set. No tracking requests are made. Third-party scripts are blocked by the CMP's default state.
2. Reject clicked: The CMP records the refusal and signals all controlled scripts. The banner dismisses.
3. Post-reject state: Only essential cookies remain (session identifiers, the consent-state cookie itself). No tracking domains are contacted. The page is fully functional without non-essential scripts.
4. After reload: The CMP's consent-state cookie is read. It says "rejected." Non-essential scripts remain blocked. No tracking cookies appear. The state is identical to post-reject.
This is achievable with any major CMP. It requires that every non-essential script is managed through the CMP's blocking mechanism, that no tracking code is loaded outside the tag manager, and that server-side responses respect client-side consent state. It is not technically difficult. It is operationally disciplined.
The 25.2% of conclusively tested sites that pass our reject-flow test prove it can be done. The 48.3% that still keep non-essential cookies after reject show that many organizations have not done the work to make it happen.
How to Test Your Own Site
You do not need automated scanning tools to perform a basic reject-flow test. Here is a manual process anyone can follow:
1. Open your site in a private/incognito window. This gives you a clean browser state with no stored cookies or consent records. 2. Open developer tools before the page loads. In Chrome: F12, then the "Application" tab for cookies and "Network" tab for requests. Make sure the network log is recording from the start. 3. Navigate to your site. Watch what appears in the cookies panel and network log before you interact with anything. Note any cookies from analytics or advertising domains. Note any requests to third-party tracking services. 4. Click the reject button. Check the cookies panel again. Non-essential cookies should disappear. Check the network log: tracking requests should stop. 5. Reload the page. Check cookies again. If cookies that disappeared in step 4 are back, you have consent respawn. Check the network log: if tracking requests resume on reload, your blocking is not persistent. 6. Navigate to another page on your site. Repeat the checks. Consent rejection must persist across page navigation, not just on the initial page.If any of these steps reveal persistent tracking, the root cause is almost always one of three things: scripts loaded outside the tag manager, tag manager rules that do not check consent state, or third-party scripts that ignore the consent signal. The fix is to route all non-essential scripts through your CMP's blocking mechanism and verify that none bypass it.
Recommendations for Website Owners
Based on our analysis of 28,891 reject flows, here are five concrete steps to fix the most common failures:
1. Audit every script on your site. Use your browser's developer tools or a tool like Google Tag Manager's preview mode to identify every script that loads on your pages. For each one, determine whether it is essential (strictly necessary for the service the user requested) or non-essential (analytics, marketing, advertising, social media). Every non-essential script must be managed by your CMP. 2. Test the reject flow yourself, regularly. Do not rely on your CMP's dashboard. Open an incognito window and go through the process described above. Do this after every change to your site's scripts, tag manager configuration, or CMP settings. Automate it if possible. 3. Eliminate hardcoded tracking scripts. If any analytics or advertising code is embedded directly in your HTML templates (not loaded through your tag manager), move it into the tag manager and assign it to the appropriate consent category. Hardcoded scripts are the single most common cause of reject-flow failure. 4. Verify server-side behavior. If your server sets cookies via HTTP headers (Set-Cookie), those cookies are set before any client-side consent logic runs. Ensure your server does not set non-essential cookies in its response headers. This is particularly common with server-side analytics, A/B testing platforms, and CDN-level tracking. 5. Monitor for consent respawn specifically. After implementing reject, test not just the immediate post-reject state but the state after reload, after navigation, and after a delay. Consent respawn is invisible in single-snapshot testing -- you only catch it by observing the site across multiple page loads.Related Guides
If your own test shows cookies or tracking after Reject, these guides explain the most common causes and fixes:
- How to test your cookie consent banner: a 10-minute manual check and an automated scan
- Cookiebot: visitors click Deny, but tracking continues
- Complianz: Reject clicked, but the GA4 cookies stay
- CookieYes: Reject clicked, but the Meta Pixel keeps firing
The Larger Question
The reject-flow failure rate raises a question that goes beyond individual website compliance: does the consent model work?
The theoretical framework is sound. Users should be able to accept or refuse tracking, and their choice should be respected. But when nearly half of conclusively tested reject buttons leave non-essential cookies in place, the framework is failing in practice. Users who go out of their way to click reject -- who actively exercise their privacy rights -- are not meaningfully protected by doing so. On much of the EU web, the reject button is a placebo.
This is not primarily a legal problem (the law is clear) or a technical problem (working implementations exist). It is an enforcement and accountability problem. Until reject-flow testing becomes a standard part of regulatory audits, and until non-functional reject buttons carry the same enforcement consequences as missing consent banners, the economic incentive to fix them will remain too low.
In the meantime, the burden falls on website operators who take compliance seriously to go beyond installing a CMP and actually verify that it works. The data shows that installation alone is not enough.
Test your reject button. Run a free scan at gdprprivacymonitor.eu to see whether your reject flow actually stops tracking -- or whether your visitors' refusal is being quietly ignored.