In 2017, ad fraud came up in 60 of the stories we collected at Advertising Perspectives. By 2024 it was seven.
If you only looked at that number, you’d conclude the industry solved ad fraud. It didn’t. But reading the past decade of headlines in order, it’s clear that something real did happen. The industry fixed the specific kind of fraud it was fighting in 2017. The problem then moved to places those fixes didn’t reach, and some of it stopped being called fraud at all.
The count
Stories in our archive that mention ad fraud, by year:
| Year | Stories |
|---|---|
| 2016 | 4 |
| 2017 | 60 |
| 2018 | 39 |
| 2019 | 33 |
| 2020 | 16 |
| 2021 | 17 |
| 2022 | 20 |
| 2023 | 12 |
| 2024 | 7 |
| 2025 | 16 |
A peak in 2017, a long decline and a small uptick in 2025. As with every chart we’ve built from this archive, the mix of outlets changes by year, so read it as a shape. The shape matches what we remember from the inside: fraud was the dominant worry of programmatic’s adolescence, then gradually turned into a line item that gets managed.
2016–2017: “There is no collective will”
At the end of 2016, the industry got its wake-up call. A botnet known as Methbot was found faking video views at enormous scale, and while exchanges downplayed its impact, AdExchanger predicted the fallout would extend well into 2017. It did.
The coverage that year was a catalog of ways to steal. Digiday reported that domain spoofing remained a huge threat to programmatic: fraudsters would label cheap or fake inventory as if it came from a premium publisher, and buyers had no reliable way to check. AdExchanger noted that thieves didn’t care whether they were on iOS or Android. One contributor described adware that degraded the product experience itself.
The diagnosis was blunt. The Financial Times’ sales chief told Digiday that on fraud, “there is no collective will.” An AdExchanger contributor had argued the year before that standards wouldn’t bust ad fraud. Others floated more radical fixes, like an impression token to make every impression verifiable.
2017: The fix that worked
Then the industry did something that, in hindsight, is one of its best decisions of the decade. It shipped a small, boring, public standard.
Ads.txt is a plain text file a publisher puts on its own website, listing the companies authorized to sell its inventory. That’s it. A buyer can check the file before buying, and if a seller isn’t on the list, the inventory isn’t really from that publisher.
In May 2017, AdExchanger’s headline was domain spoofing be gone: ads.txt will filter out imposter sites, and a week later, that ads.txt inspired cautious optimism. Digiday explained what it was in June, the same day it reported that domain spoofing was still unsolved. The open question was whether buyers were ready to pull the trigger and actually refuse unauthorized inventory. By September, an open letter to publishers put it plainly: adopt ads.txt or be blocked.
They adopted it. The standard was later extended with sellers.json, which lists who each seller is, and Digiday explained what that was in 2019. By 2022 the industry was already working on where sellers.json fell short and how supply chain objects could help. Not perfect. But domain spoofing on the open web, the fraud that dominated 2017, became far harder to pull off at scale.
The lesson we take from it: the fix that worked wasn’t the clever one. It was the simple, public one that any buyer could check for free.
2020–2021: Fraud moves to the television
Fraudsters follow the money, and by 2020 the money was in connected TV. CTV inventory was expensive and buyers were hungry for it. And because ads are stitched into the stream on the server, much of the inventory couldn’t be verified by the old methods.
In January 2020, DoubleVerify launched a supply certification aimed at CTV fraud. That August, AdExchanger ran a piece on how the industry could step up to fight CTV fraud. By July 2021, CTV fraud was changing, and in August one firm said it had uncovered CTV fraud costing advertisers $20 million a month.
The response followed the ads.txt playbook again, with an app-ads.txt equivalent for apps and new ways to prove a device is real. In November 2025, Digiday was still explaining the IAB Tech Lab’s device attestation approach to CTV fraud. The fight continues, but it has the same shape as 2017: a public, checkable signal that makes the cheap tricks expensive.
2023–2024: The fraud that isn’t called fraud
The quietest shift in the archive is the most important one. After 2022, a lot of what buyers worry about stopped being outright fraud and became something harder to name.
In 2023 the hot topic was made-for-advertising sites. Digiday explained what MFAs were in June: sites that exist mainly to arbitrage traffic and fill pages with ads. They’re real sites, with real (if bought) visitors and real ads. Nothing is faked, which is exactly the problem. By August, Digiday reported that MFAs lacked a clear definition, and in 2024 the verification companies were arguing about shades of gray, with IAS adding another.
The same year raised harder questions about the biggest platforms. In August 2023, an AdExchanger contributor wrote about a second, worse wave of alleged ad fraud at Google, and research firm Adalytics followed with evidence of behaviorally targeted ads served against kids’ content on YouTube. The ANA’s long-running transparency work moved forward too. It parted ways with its auditor in mid-2023, and by June 2024 it reported that advertisers were spending way less on MFA, though programmatic still wasn’t fully transparent.
And AI showed up on both sides. In August 2023, AdExchanger’s roundup noted that AI was accelerating ad fraud. Cheap generated content makes it easy to spin up endless sites that look legitimate enough to buy. The same technology is now being used to detect them.
So, did anything get fixed?
Yes, more than the industry tends to admit.
The 2017 kind of fraud is much harder now. Spoofing a premium publisher’s domain on the open web used to be trivial. Ads.txt and sellers.json made it checkable, and checkable fraud doesn’t scale.
What got harder to fix is waste, not theft. MFA sites, unclear supply paths, ads in the wrong places. These aren’t crimes, and no text file stops them. They’re the result of buyers optimizing for cheap reach and the market happily supplying it.
The fraud follows the money. It moved from desktop to mobile to CTV, and it will move to whatever is expensive and hard to verify next. Agent-driven buying, which we’ve written about separately, is a natural target: automated buyers transacting with automated sellers at speed.
The fixes that work are public and boring. Every durable improvement in this archive followed the same pattern as ads.txt: a simple, open signal that any buyer can check for free. The clever, proprietary fixes come and go.
If we were advising a buyer today, the playbook is short. Buy from fewer, known paths. Insist on ads.txt and sellers.json compliance, and on device verification for CTV. Treat unusually cheap reach as a question, not a bargain. And measure outcomes, not impressions, because fraud and waste both look great on an impressions report and terrible on a sales one.
Every headline in this piece links to the original story. Browse the archive’s coverage of ad fraud, ads.txt and made-for-advertising sites.
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.