Know what’s on your site before a plaintiff’s firm does
A free privacy audit shows which cookies, pixels and trackers are running — and which ones commonly appear in privacy claims.
In August 2025, a San Francisco jury found that Meta violated the California Invasion of Privacy Act by obtaining women’s menstrual cycle and pregnancy information through its tracking code inside a period-tracking app. Three other defendants in the same case, Flo Health, Google and Flurry, had already agreed to settle for a combined $59.5 million. As of September 2026 reporting, Meta’s damages had still not been set, and the statute allows $5,000 per violation without proof of financial loss. The verdict stands after the court rejected Meta’s efforts to overturn it.
Now Meta is about to change what its own tracking code does on your website. Meta has told advertisers that, starting November 12, 2026, a new Meta Pixel feature will begin rolling out that recovers conversions missing from campaign reporting today, because they never make it back to Meta to be measured. Instead, those conversions are measured on the customer’s device, using information customers already share with the business, such as an email address, together with temporary, non-identifying device data. Meta says both stay on the device, that it learns only that a purchase happened and which campaign should get credit, that no updates to existing integrations are required, that the feature applies to datasets already sending events, and that advertisers can opt out at any time in Events Manager.
For a media buyer, that sounds like better reporting at no cost. For a privacy officer, or for the agency that manages dozens of client ad accounts, it reads differently: a new default-on behavior in the most litigated tracking technology on the web, scheduled to begin in 25 days, described in a few sentences that leave the key legal questions unanswered.
This guide connects the announcement to the Meta Pixel litigation record, explains who carries the risk (agencies, marketers, privacy officers and their counsel), lists the questions to put to Meta in writing, and gives a plan that fits inside the 25-day window. Our analysis rests on Meta’s notice as it was sent to advertisers. We have not seen technical documentation beyond it, so treat the mechanics below as questions to verify, not conclusions, and expect to update your view when Meta publishes more.
What Meta Announced, and What It Left Out
Reduced to its core, the notice says five things.
- A new Pixel behavior. Starting November 12, 2026, the Meta Pixel gains a feature that recovers conversions that do not appear in campaign reporting today.
- A different measurement location. Those conversions are measured on the customer’s device instead of being sent to Meta for measurement.
- Inputs described as ordinary. The measurement uses information customers already share with the business, such as an email address, plus temporary, non-identifying device data.
- A limited disclosure to Meta. Meta says it learns only that a purchase happened and which campaign should receive credit.
- Default-on, with an opt-out. No integration updates are needed, the feature applies to datasets already sending events, and advertisers can opt out in Events Manager.
What the notice does not say matters more to a compliance team than what it does. It does not say how the feature treats visitors who declined consent or sent an opt-out signal. It does not say what the on-device code reads from the page or the browser, how long the temporary device data lives, or what identifier ties a recovered purchase to a campaign. It does not say whether advertisers in sensitive sectors are treated differently, and it does not say whether the feature can be narrowed rather than switched on or off entirely. Those gaps are the agenda for the rest of this article.
Part of a Pattern: Meta’s 2026 Default-On Measurement Changes
The November 12 change is not an isolated event. According to industry coverage of Meta’s announcements this year, advertisers have been absorbing a series of measurement features that arrive switched on and have to be switched off. Confirm each in your own Events Manager, because Meta renames and relocates these controls often.
| Date | Change | How it arrives | The privacy question |
|---|---|---|---|
| April 15, 2026 | An AI-enhanced Pixel that enriches events with product and page details, plus a one-click, Meta-hosted Conversions API | Existing users received a notification and a 30-day review window before the AI feature activated, with controls to disable it or narrow it by data category | What page content is now read and sent, and did your notice and consent cover it? |
| August 3, 2026 | An Automatic Events control in which AI detects website actions such as leads and purchases | Reported as enabled by default for new Pixels added from that date | Who decided which actions are measured, and are any of them sensitive? |
| November 12, 2026 | On-device recovery of conversions that never reach Meta | Rolls out to datasets already sending events; opt out in Events Manager | Does it respect visitors who declined consent or sent an opt-out signal? |
The pattern is the point. Each feature is individually defensible and each arrives through a settings page rather than a code change on your side. That is exactly how a website’s data flows drift away from what its privacy notice and consent banner describe, and drift is what plaintiffs and regulators look for.
Why the Meta Pixel Sits at the Center of Pixel Litigation
If you wonder why a measurement feature deserves this much attention, look at where web privacy claims actually land.
The numbers from the claims data
In Coalition’s State of Web Privacy report, which analyzed nearly 200 data privacy claims reported by policyholders, the Meta Pixel was cited in 43 percent of web privacy claims, far ahead of any other analytics technology. Yet it appeared on only 19 percent of the scanned websites that run any tracker. Plaintiffs choose technologies that send behavioral and identifying data to an advertising platform because those fit their theories best. We covered the full findings in our analysis of the Coalition State of Web Privacy report, including how a small number of plaintiff firms generate most of the demand letters and how one firm’s playbook focuses on users who were allegedly tracked after opting out.
The verdict that made trial real
The Flo Health case, Frasco v. Flo Health, involved Meta’s code inside a mobile app rather than a website pixel, but its significance for every advertiser is the same. A jury applied the California statute’s eavesdropping provision to tracking code that captured confidential health information, and the verdict survived post-trial motions. The three settling defendants paid a combined $59.5 million, with Flo Health’s share reported at $8 million. A final approval hearing for those settlements is set for October 29, 2026. Commentators note that the facts of the case, reproductive health data and a class of identified app users, were unusually sympathetic, but the lesson holds: a theory that pixel and SDK tracking is unlawful interception no longer collapses before a jury.
The healthcare settlement stream
Healthcare organizations have paid steadily for Meta Pixel deployments on patient portals. The Coalition report cites Advocate Aurora Health at $12.25 million in 2023 and Mass General Brigham at $18.4 million in 2022. In 2026, Atrium Health reached a settlement over pixels on its patient portal and website, reported at roughly $2 million, and press coverage describes a continuing run of hospital and health system settlements. The allegations follow a consistent template: tracking code on pages where patients log in or book care, information flowing to Meta and sometimes Google for advertising, and no valid consent first.
What a California change does and does not fix
California’s SB 690, signed on September 30, 2026, removes the private right of action under the pen register provision of the California statute for conduct on websites and apps, effective January 1, 2027. That eliminates one popular theory. It does not touch the wiretapping and recording provisions that were at issue in the Flo verdict, other states’ wiretap laws, federal law, common-law claims, or regulators. Anyone who expects pixel exposure to disappear on January 1 should read the bill as narrow.
The three facts every pixel case turns on
Strip away the statutes and nearly every pixel case reduces to three facts: what the tracking code collected, who received it, and whether the visitor was told and agreed before it happened. The November 12 feature changes the first two facts, quietly and by default. That is why it belongs on a privacy officer’s calendar even though it is described as a reporting improvement.
Five Questions the Notice Leaves Open
1. Does it respect consent and opt-out signals?
The notice says the feature recovers conversions that never make it back to Meta today. It does not say why they are missing. Some are surely lost to technical failures. Others are plausibly missing because the visitor declined consent, used a blocker or sent an opt-out preference signal such as Global Privacy Control. If the new measurement runs regardless of those choices, it changes the legal analysis, because a choice the visitor made is being worked around. If it respects them, the recovered volume will be smaller than the marketing suggests. Either answer is acceptable to a privacy team. An absent answer is not.
2. What does the code read on the device?
The description refers to information customers already share with the business, such as an email address. On a checkout or lead form, that means the script is reading something a visitor typed into a page. Wiretap and eavesdropping theories turn on whether code read the contents of a communication, and European ePrivacy rules treat reading information from a user’s device as a regulated act that generally needs consent unless strictly necessary. Neither framework asks only whether data later leaves the device, so “it stays on the device” is a statement about one step, not a legal safe harbor.
3. Is “non-identifying” a fact or a label?
Meta describes the device data as temporary and non-identifying and says it learns only that a purchase occurred and which campaign gets credit. A conversion credited to a specific ad interaction is event-level information tied to a person’s behavior, and regulators have been blunt about labels like this. In 2024, FTC staff wrote that data is only anonymous when it can never be associated back to a person, and that hashing and similar obfuscation do not meet that bar. Under the GDPR, pseudonymous data is still personal data. Until Meta documents exactly what identifiers are involved, treat “non-identifying” as marketing language that your own notice should not repeat.
4. Who is responsible for telling the visitor?
Under European law, the Court of Justice held in the 2019 Fashion ID decision that a website operator that embeds a social network’s code can be a joint controller with the network for the collection and transmission of visitor data. The practical meaning is that you cannot hand the disclosure problem to Meta. Your privacy notice, consent interface and vendor disclosures need to describe what the pixel does on your site, including new behavior you did not code yourself.
5. Can you prove what your site did?
When a demand letter, a regulator inquiry or an insurer’s underwriting question arrives, you will be asked what your pixel did on specific dates. If a feature switched on by default on November 12 and nobody captured your settings or your site’s behavior before and after, you will be reconstructing history from memory. A dated baseline is cheap now and impossible later.
Who Carries the Risk
| Role | Why the exposure is real | What to do first |
|---|---|---|
| Marketers and media buyers | They approve measurement features for performance reasons and may not see the legal downstream | Pause the assumption that “no integration changes” means “no privacy change,” and route the decision to privacy or legal |
| Agencies | They hold admin access to client datasets, so a default that switches on under their management is arguably their change | Get written client direction on each dataset, and keep a change log |
| Privacy officers and DPOs | They answer for the notice, the consent interface and the record of processing, all of which can drift without a code change | Inventory every Meta dataset, decide an opt-out policy by context, and obtain Meta’s answers in writing |
| Counsel | Pixel cases turn on facts that a default setting can alter | Preserve a dated baseline and review contract allocations with agencies and vendors |
| Healthcare, finance, education and children’s services | Sensitive context turns routine tracking into the cases with the largest settlements | Treat the feature as opt-out until Meta’s consent handling is documented |
A Special Note for Agencies
Agencies sit in the most exposed position because they have the access and the pace. A single agency may administer pixels for dozens of clients across jurisdictions, and a platform default can change them all on the same day. Four practices reduce that risk.
- Keep a pixel register. List every Meta dataset you manage, the client, the sites it fires on, the sensitivity of those sites and the current state of each Events Manager setting.
- Get client direction in writing. For each dataset, record whether the client wants the feature on or off, who decided, and on what date. Silence should not be interpreted as consent to a new default.
- Review the contract. Your services agreement and any data processing terms should say who is responsible for configuring platform features, who must approve changes that affect data flows, and how liability is allocated if a default setting creates exposure.
- Do not enable on sensitive properties without legal sign-off. Patient portals, financial application flows, children’s services and any page where a visitor’s situation is itself sensitive deserve a human decision, not a default.
The 25-Day Plan
Meta’s notice leaves 25 days between today and November 12. The plan below fits that window.
| Window | Objective | What to finish |
|---|---|---|
| Oct 18 to Oct 25 | Inventory and baseline | List every Meta dataset and Pixel across clients and sites, record current Events Manager settings with dated screenshots, run a scan of each site to capture what the Pixel and other trackers do today |
| Oct 26 to Nov 1 | Ask and decide | Send Meta the written questions below through your account contact, classify sites by sensitivity, draft an opt-out policy by context, and brief counsel |
| Nov 2 to Nov 8 | Test and align | Test each priority site with consent declined and with an opt-out signal sent, confirm the Pixel is gated behind consent, and update privacy notices and vendor disclosures where the feature will be on |
| Nov 9 to Nov 12 | Set and document | Apply the opt-out or opt-in decision dataset by dataset in Events Manager, record who decided and when, and capture settings again |
| Nov 13 onward | Verify and monitor | Rescan, compare behavior against the baseline, and set a recurring check, because Meta changes these controls frequently |
Questions to send Meta in writing
- How does the feature treat visitors who declined consent, sent a Global Privacy Control or other opt-out signal, or whose browser blocks tracking?
- What exactly does the on-device code read from the page, the form fields and the browser, and for how long is the temporary device data kept?
- What identifier links a recovered conversion to a campaign, and what does Meta receive and store?
- Does the feature behave differently for sensitive categories such as health, finance, employment, housing or children’s services?
- Can the feature be limited by site, event type or region, or only switched on or off?
- What documentation can Meta provide that we can cite in our privacy notice and in our records of processing?
To Opt Out or Not: A Decision Framework
There is no universal answer, and the business decision belongs to you and your counsel. This framework reflects a privacy-first default while Meta’s documentation is thin.
| Context | Suggested stance | Why |
|---|---|---|
| Health, finance, children, employment, housing or other sensitive services | Opt out until Meta documents consent handling in writing | Sensitive contexts drive the largest pixel settlements and regulatory attention |
| Properties with visitors in the EU, UK or other consent-first jurisdictions | Opt out unless the Pixel is gated behind valid consent and the notice describes the new behavior | Reading information on a device is regulated regardless of what leaves it |
| Consumer sites in US states with opt-out signals and wiretap exposure | Opt in only after testing that consent choices and opt-out signals are honored, and after updating disclosures | The litigation theories focus on tracking after opt-out and on undisclosed sharing |
| Low-risk properties with working consent gating and current notices | Decide on business value, document the decision and monitor | The risk is manageable when the surrounding controls are verified |
How Captain Compliance Helps
The common thread in everything above is evidence and control: knowing what your pixels do, making sure consent actually governs them, and being able to prove it. That is the work Captain Compliance is built to do.
- Gate the Pixel behind consent. Our consent management platform holds non-essential tags, including the Meta Pixel, until a visitor opts in and honors Global Privacy Control signals. When a script is blocked before consent rather than merely muted, there is no code on the page to run new behavior. Configuration matters here, so verify blocking in testing, not in the settings screen.
- See what actually fires. Patrol scans for tracking tags, cookies and consent failures, so you can capture a baseline before November 12 and compare behavior afterward.
- Keep a dated record. Patrol’s evidentiary output is SHA-256 verified, with dated scan records and screenshot archives, which gives counsel and insurers something better than recollection.
- Keep your notices true. Dynamic privacy policies and vendor disclosures update as your technology changes, so what you say stays aligned with what your site does.
- Handle the downstream requests. The DSAR portal routes access, deletion and opt-out requests and leaves an audit trail.
Our consent platform is used on more than 8,000 websites and processes over 50 million consent choices per month. If you manage Meta datasets for yourself or for clients and want a baseline before November 12, start with a free privacy audit or book a demo and we will walk through what your sites send today.
Frequently Asked Questions
What is Meta’s new on-device conversion feature?
Meta has told advertisers that, starting November 12, 2026, a new Meta Pixel feature will recover conversions that do not appear in campaign reporting today. Those conversions are measured on the customer’s device using information the customer already shared with the business, such as an email address, plus temporary device data. Meta says it learns only that a purchase happened and which campaign should get credit.
Do I have to do anything for the feature to turn on?
No. Meta says no updates to existing integrations are required and that the feature applies to datasets already sending events. That is why a privacy review should happen before the rollout rather than after.
Can I opt out of the Meta Pixel on-device feature?
According to Meta’s notice, you can opt out at any time in Events Manager. Confirm the control in your own account, record who made the decision and the date, and capture the setting as evidence.
Does measuring on the device mean it is not a privacy risk?
Not necessarily. Wiretap theories focus on whether code read the contents of a communication, and European ePrivacy rules regulate reading information from a user’s device generally regardless of whether it leaves. Data staying on the device is relevant but not a legal safe harbor, and Meta still learns that a conversion occurred and which campaign gets credit.
Why is the Meta Pixel central to privacy litigation?
Coalition’s analysis of insurance claims found the Meta Pixel cited in 43 percent of web privacy claims. A jury has found Meta liable under the California Invasion of Privacy Act in the Flo Health case, healthcare organizations have paid steadily to settle pixel claims, and statutory damages of $5,000 per violation under the California statute multiply quickly.
Does California’s SB 690 end Meta Pixel lawsuits?
No. SB 690, signed September 30, 2026, removes the private right of action under the pen register provision for website and app conduct effective January 1, 2027. It leaves the wiretapping and recording provisions, other states’ laws, federal law, common-law claims and regulatory enforcement in place.
What should an agency do before November 12?
Build a register of every Meta dataset it manages, get written client direction on whether each should have the feature on or off, review contract language on who configures platform features, and avoid enabling the feature on sensitive properties without legal sign-off.
Can a consent management platform prevent the feature from collecting data?
If the Meta Pixel is blocked before consent, rather than merely signaled, the code is not on the page to run the feature for visitors who have not agreed. The exact behavior depends on how the Pixel is loaded and configured, so test with consent declined and with an opt-out signal sent, and verify the result.
What questions should I ask Meta?
Ask how the feature treats visitors who declined consent or sent opt-out signals, what the on-device code reads, what identifier links a conversion to a campaign, whether sensitive categories are handled differently, whether the feature can be limited by site or event, and what documentation Meta can provide. Put the questions in writing and keep the answers.
Put this into practice
Turn privacy guidance into working controls.
Captain Compliance helps teams discover tracking technologies, enforce consent, manage privacy requests and document the evidence behind every decision.