In 2024, one in every 25 ad tech stories we collected at Advertising Perspectives was about the death of the third-party cookie. A year later, the number was closer to one in 600.
The cookie didn’t die. The story did.
We have been building marketing technology long enough to remember when a cookie sync was a clever idea and not a liability. For the last six years, a single question sat underneath nearly every data and identity conversation in the industry: what do we do when Chrome turns them off? Google answered it in April 2025 by announcing that Chrome wouldn’t turn them off at all.
This site has been collecting the trade press since 2016: AdExchanger, Digiday, TechCrunch, The Verge, Beet TV, MediaPost and others, more than 140,000 stories in all. We went back through the archive and read the cookie coverage in order. Read that way, it’s less a news story than a case study in how an entire industry plans around a deadline that keeps moving.
The chart
Here is how often the third-party cookie showed up in the stories we collected, as a share of everything published that year.
| Year | Stories that mention third-party cookies | Per 1,000 stories |
|---|---|---|
| 2018 | 5 | 0.3 |
| 2019 | 30 | 2.0 |
| 2020 | 97 | 6.5 |
| 2021 | 102 | 8.9 |
| 2022 | 66 | 19.0 |
| 2023 | 50 | 14.6 |
| 2024 | 132 | 40.6 |
| 2025 | 20 | 1.6 |
Attention peaked in the same year the plan quietly fell apart. Hold that thought.
2019–2020: “Within two years”
The first tremors showed up in the spring of 2019, when AdExchanger asked how Google’s move to restrict third-party cookies would affect publishers. By Christmas the industry was reassuring itself: Digiday’s headline on December 24 was “Google’s Chrome browser update won’t kill the cookie.”
Three weeks later, on January 14, 2020, Google said it would. TechCrunch reported that Google wanted to phase out support for third-party cookies in Chrome within two years. AdExchanger’s version was blunter: “Google Chrome Will Drop Third-Party Cookies In 2 Years.” Within days Digiday was cataloguing the collateral damage beyond ad targeting: frequency capping, attribution, measurement. All the plumbing nobody puts on a slide.
Then the world shut down. On March 25, a W3C business group said it would petition Google to postpone because of COVID-19. Two days later Google said it had no plans to postpone.
That was the posture for the next five years. The date is real until it isn’t.
2021: The replacement that got replaced
Google’s answer to “what comes next” was FLoC, a browser-based scheme that grouped people into cohorts instead of tracking them one by one. The industry spent a good part of 2021 evaluating it. Publishers including The Guardian became conscientious objectors. In June, Amazon blocked it. Nine days later, Google granted a reprieve and pushed deprecation back by two years. By early 2022 FLoC itself was gone, replaced by a different proposal called Topics.
That was delay number one. More important than the delay was the pattern it set. Every replacement Google proposed became its own project for the industry to test, debate and wait on, and every one of those projects inherited the original deadline’s uncertainty.
2022–2023: “I highly doubt this is the last deadline”
In July 2022, Google delayed again, from 2023 to the end of 2024. The best line of that news cycle came from a source in Digiday’s coverage: “I highly doubt this is the last deadline.” They were right.
By mid-2023 the CEO of The Trade Desk, Jeff Green, was calling the plan a strategic mistake. In December, after nearly four years of notice, Digiday was still explaining why the ad industry wasn’t ready. Three days later AdExchanger reported that Google planned to ring in the new year with deprecation, “for real.”
Four years of warning, and readiness was still the headline. That isn’t a criticism of the reporters. It’s an accurate description of how most organizations handle a deadline they don’t control: they wait for it to be real.
2024: The 1% and the peak
In January 2024 Chrome finally turned off third-party cookies for 1% of its users. Ad executives said the process had been dragging on for too long. A month later, Digiday found marketers still apathetic.
In April came delay number three, which AdExchanger marked with a collection of cookie-related quips. By then gallows humor was a reasonable editorial choice.
In July, Google changed the plan entirely. Instead of removing third-party cookies, it would let users choose, and AdExchanger asked what that meant for Google’s relationship with regulators, who had been overseeing the transition. By October the question was what the user choice mechanism would even look like.
This is the year coverage peaked, at roughly 40 stories in every thousand. The industry was writing more about the cookie than at any point in the decade, at exactly the moment the thing everyone was writing about stopped being likely to happen.
2025: The U-turn
The spring of 2025 tied it off. In March, Digiday reported that without the cutoff hanging over them, buyers were lukewarm on alternative IDs. On April 22, Google said it would not roll out the user-choice prompt either. Chrome would continue to use third-party cookies, and The Verge reported that Google was scrapping its planned changes.
The industry’s reaction, per Digiday the same day: “endless millions have been wasted.” Ad Tech Explained called it the end of the Google cookie saga. In October, AdExchanger wrote the epitaph for the rest of Google’s replacement technology: “Rest In Privacy, Sandbox.”
What the scramble actually bought us
“Endless millions have been wasted” is a satisfying quote, and it’s partly true. A lot of money went into testing proposals that no longer exist and into identity products whose sales pitch depended on a date. If you were on the receiving end of those invoices, you’re entitled to the frustration.
But “wasted” assumes the only point of the work was surviving Chrome. It wasn’t, and the teams that treated it that way are the ones who feel cheated now.
The cookie was already gone for a large share of your audience. Safari blocked third-party cookies outright in 2020. Firefox started blocking third-party tracking cookies by default in 2019. Apple’s App Tracking Transparency arrived in 2021 and changed mobile app tracking overnight. Chrome was the holdout, not the market. Anyone whose measurement and audience strategy still depends on the cookie is building on the shrinking part of the problem that happens to look solved.
Deadlines are not a strategy. The organizations that did well over these six years didn’t plan around 2022, then 2023, then 2024. They planned around the direction of travel: less tracking, more consent, more regulation, fewer shared identifiers. That direction didn’t change in April 2025. Only the date did.
And while everyone watched the date, the industry quietly built the cookie’s replacement anyway.
The modern cookie
The third-party cookie did three jobs at once. It recognized a person across sites, it carried what was known about them, and it let two companies match their data without much ceremony. Nothing replaced it one-for-one. Instead those three jobs split apart, and each one became its own layer of the stack.
Consumer identifiers do the recognizing. The cross-site join didn’t disappear. It moved from an anonymous browser file to an authenticated signal, usually a hashed and encrypted email address that a person handed over when they logged in. LiveRamp got there first with RampID and its Authenticated Traffic Solution, and it is all over our archive: 183 stories, peaking at 46 in 2020, the year everyone went looking for a plan B. By the spring of 2021, publishers like CafeMedia were selling ads against LiveRamp’s email-based IDs. The Trade Desk’s Unified ID 2.0 became the other big name, and in October 2020 the two united on identity, with UID2 becoming part of LiveRamp’s ATS. Our archive tracks UID2’s rise too: 25 stories in 2021 alone, as one company after another signed on, from InfoSum to MediaMath, until in 2023 The Trade Desk scrapped its plan for independent public operators. The difference from the cookie matters. An identifier like this only exists because a customer chose to log in, so its reach is earned, and it’s only as good as your consent records and your match rates.
The CDP does the remembering. Once you have a recognizable customer, you need somewhere to resolve their identity across every system that has touched them and keep a profile you’re allowed to use. That’s the customer data platform. In June 2019, before Google had even set a date, the big marketing clouds were already building CDPs of their own. By 2021 the category had grown up, and identity vendors were moving into the same territory: LiveRamp launched identity resolution for first-party data that May. Today the same idea is built into most of the major marketing and data platforms: identity resolution, consent and audience insights in the system where the customer record already lives. In our experience this is the layer most companies underestimate. They buy a CDP to replace the cookie, then discover the hard part is that their own customer data has never been in one place.
The clean room does the matching. When a brand and a publisher, or a brand and a retailer, want to compare audiences without handing each other raw customer files, they meet in a clean room. Coverage in our archive climbed to a peak of 61 stories in 2023, and in January 2024 LiveRamp bought Habu, folding one of the best-known independent clean rooms into the company that already owned so much of the identity layer. By late 2024 the FTC was pointing out that clean rooms may have a few dusty corners, and in January 2026 AdExchanger called the end of the clean room era. We’d put it differently. Clean rooms didn’t fail. They stopped being a product category and became a feature of the data platforms everyone already runs.
Put the three layers together and you have the modern cookie: a consented identifier, resolved into a profile in a CDP, matched with partners in a clean room. It is better than the old cookie in every way that matters to a customer and a regulator. It is worse in one way that matters to a business. The old cookie was effortless, and this stack is not. Plenty of companies now hold exactly the first-party audiences advertisers want, and still struggle to turn them into revenue, because every question about those audiences ends in a ticket for the data team.
What to watch now
AdExchanger opened 2026 with a warning not to let the next round of privacy shifts blindside you. The pressure that drove the cookie fight hasn’t gone anywhere. It has moved to state privacy laws, consent requirements and, increasingly, to AI agents that act on a person’s behalf.
Agents change the identity question in two directions. They need a trustworthy way to know who a customer is and what they’ve agreed to. And inside the enterprise, they change who gets to ask questions of customer data in the first place. A sales lead or a CRO shouldn’t need a dashboard, an analyst and a week to find out what an audience is worth.
The cookie made audience data easy and careless. The modern stack made it careful and hard. The next step is making it careful and easy, which is the problem we’re working on at data4orce.ai.
Every headline in this piece links to the original story. Browse the full archive of third-party cookie coverage, or the related coverage of Google’s Privacy Sandbox, LiveRamp, Unified ID 2.0 and data clean rooms.
Kevin Reynolds is the founder of Advertising Perspectives, MadTech Consulting and data4orce.ai. He is a marketing technology architect and C-level advisor, with a focus on Salesforce and data-driven advertising.