Skip to main content
The state of ai impact assessment
Does Your Opt-Out Stop the Requests? A Network-Level ChecklistLaws and Regulations
5 min readFor DSAR and Consent Operators

Does Your Opt-Out Stop the Requests? A Network-Level Checklist

By Cookie GatePublished

A visitor clicks “opt out,” the Consent Notice confirms the choice, and ad-tech requests keep flowing. The choice only works if it changes what your site sends. A confirmation message can’t prove that it did.

That gap sat at the center of the $2.975 million fine CalPrivacy announced against FedEx on November 10, 2026. The privacy policy said “We do not ‘sell’ or ‘share’ your Personal Information,” while Meta, LinkedIn, Google, Adobe, and other ad-tech cookies and pixels ran on fedex.com. This checklist is for DSAR and consent operators who want to verify the full opt-out path, from signal or click to network request, and keep evidence that it worked.

Prerequisites

Before you start, have these in place:

  • Access to a staging environment and production, plus the ability to change tags in each.
  • A browser that supports Global Privacy Control (GPC), such as Firefox, Brave, or DuckDuckGo, and developer tools with a Network tab.
  • A current export of your privacy policy and request-form text.
  • A shared location for test evidence, including screenshots, HAR exports, and dated notes.
  • A named owner for each item below. A check with no owner won’t run twice.

The Checklist

1. Build a tracker inventory from observed behavior, not from memory.
Scan your sites and apps, then list each tracker and assess whether it sells or shares personal information. The FedEx order requires scans at least quarterly (para. 62).
Good looks like: a dated inventory listing each vendor, the data it receives, and the page or Tag Manager container that loads it. Your team can explain every line.

2. Reconcile the inventory against your privacy policy.
Compare what the inventory shows with what your disclosures say about selling and sharing. Under the CCPA, “sharing” includes making personal information available for cross-context Behavioural Advertising, whether or not money changes hands. Cookie syncing, where vendors match their identifiers to recognize the same person elsewhere, can be evidence of that sharing.
Good looks like: no statement in the policy that the inventory contradicts, and a recorded sign-off from whoever owns the disclosure.

3. Confirm the GPC signal is detected.
Turn GPC on, open the console, and run navigator.globalPrivacyControl. If it returns true, the signal is active. Load the site and confirm the Consent Management Platform (CMP) registers it. California’s CCPA regulations require businesses to honor qualifying opt-out preference signals.
Good looks like: the CMP recognizes the signal on first page load, before non-essential tags fire.

4. Confirm GPC actually stops sale or sharing requests.
Detection isn’t enforcement. With GPC on, watch the Network tab for requests to ad platforms, then inspect their destinations and payloads. FedEx’s site wasn’t set up to detect the signal at all, so data kept flowing after people opted out this way (para. 62b).
Good looks like: no requests transmit personal information to inventoried sale/share vendors in a GPC-on session. Save a HAR file as evidence.

5. Test the banner opt-out the same way.
Opt out through the Consent Notice, then watch network traffic across several page loads and a navigation or two. FedEx’s Cookie Preferences window let visitors switch off “Analytical” and “Tracking” cookies, yet some visitor data kept reaching some ad-tech companies.
Good looks like: the same result as item 4, with no sale/share data transmitted. Confirm the choice persists after a reload and a new session.

6. Check that the control explains what it does.
If a cookie control serves as your opt-out, it must explain how its settings relate to selling and sharing (para. 66). A toggle labeled “Tracking” doesn’t say that.
Good looks like: plain-language text in the control that names selling and sharing, so a visitor knows the toggle is their opt-out.

7. Verify the footer link.
Look for a “Do Not Sell or Share My Personal Information” or “Your Privacy Choices” link (para. 65). FedEx had only a “Cookie Consent” link.
Good looks like: the required link is visible on every template, including checkout and logged-in pages, and it opens the actual choice mechanism.

8. Read your request form’s fine print.
FedEx’s CCPA form stated: “This form does not opt you out of cookies or tracking technologies.” A form that disclaims tracking sends people nowhere useful.
Good looks like: as a practical design goal, the form either handles sale/share opt-outs or directs people to the relevant control.

9. Gate new vendors before their tags go live.
Review third parties before onboarding to confirm the right contract is in place (paras. 62a, 64; Cal. Code Regs. tit. 11, § 7053).
Good looks like: a Tag Manager publish process that requires an approved vendor record, and no tag ships without one.

10. Retest on a schedule and on every change.
The order requires opt-out testing in staging and production at least quarterly, and whenever a new tracker is added (para. 63).
Good looks like: a calendar entry plus a change-management trigger. Each run produces dated evidence from both environments, with the tester named.

11. Retain the evidence.
Without retained evidence, it may be harder to show that a test happened. Keep HAR exports, screenshots, console output, and inventory versions.
Good looks like: a folder or ticket per test cycle, retrievable in minutes, retained for the life of any regulatory obligation you carry.

12. Don’t carve out your B2B pages.
The CCPA’s temporary B2B exemption expired on January 1, 2023. A California resident browsing your site for work has the same rights, and work visits can still feed personal ad profiles on the same device.
Good looks like: B2B templates, gated content, and partner portals sit in the same inventory and test scope as consumer pages.

Common Mistakes

  • Testing the interface, not the traffic. A confirmation message proves a script ran, not that requests stopped. An InfoTrust audit of 450 US websites found that 79% kept firing tracking tags after visitors opted out.
  • Testing only staging. Production often carries tags staging doesn’t. Test both.
  • One-time audits. Tags change faster than policies. A clean test in March says little about a tag added in April.
  • Treating the policy as the control. FedEx’s policy language didn’t match reality; the policy was the promise, not the mechanism.
  • Skipping edge paths. Logged-in areas, microsites, and apps often miss the footer link and signal handling.

Next Steps

Start with items 3 through 5. They take an afternoon and tell you fastest whether your opt-out changes what the site sends. Then assign owners, schedule the quarterly cycle, and add vendor review to your Tag Manager publishing workflow.

Remember the scale of the commitment the order imposes: implementation within 210 days, and obligations lasting three years. Treat your own program the same way. Opt-out compliance is a recurring operational test, and evidence from each run is what makes it defensible.

a promotional banner asking how ready are you for PCI DSS 4.0? With a call-to-action to get the checklist now.

You Might Also Like