AI-generated slop sites take between 1.3% and 2.4% of open-web programmatic ad spend, according to an analysis published on 28 July 2026 by the Trustworthy Accountability Group (TAG), the Association of National Advertisers (ANA) and the technology firm Fiducia. The more uncomfortable finding sits underneath that number. The same inventory was graded premium more than 70% of the time, it beat clean supply on both invalid traffic and viewability, and it cost more per thousand impressions.
Two things belong up front. TAG and ANA are industry bodies, but the measurement itself was performed by commercial vendors: DeepSee, Mobian and Fiducia, all of which sell into the problem they measured. And the primary report document could not be located. PPC Land does not link it, so every figure here is as reported by PPC Land rather than read from the study.
What TAG, ANA and Fiducia measured on 28 July 2026
TAG, ANA and Fiducia described their 28 July 2026 analysis as the first statistically rigorous quantification of AI-generated slop inventory in programmatic advertising. Two methods ran in parallel over the same dataset. DeepSee classified inventory at domain level across $33.97 million of matched open-web spend. Mobian evaluated rendered pages, covering 233 million URLs, 4.45 billion impressions and $14.84 million in measured spend.
Those two methods produced the two ends of the headline range, and the distinction is worth getting right. The 1.3% lower bound comes from the page-level evaluation, and it is that figure which carries the stated 99% confidence interval of 0.98% to 1.56%. The 2.4% upper bound comes from the domain-level classification. The interval describes the lower-bound measurement. It is not the confidence interval of the whole 1.3% to 2.4% span.
Slop cleared the checks that exist to stop it
The comparison below is the part of the TAG and ANA analysis that should change how a media team thinks about verification. Every figure is from the 28 July 2026 study as reported by PPC Land.
| Metric | AI slop inventory | Clean supply |
|---|---|---|
| Invalid traffic rate | 0.05% | 0.32% |
| Viewability | 77.2% | 74.9% |
| TrueCPM | $7.08 | $6.15 |
| Templated-site rate | 30.0% | 1.2% |
Slop recorded roughly a sixth of the invalid traffic that clean supply did, and it was the more viewable of the two. On top of that it was graded premium more than 70% of the time and cost about 15% more per thousand impressions. The signals most advertisers rely on to police quality are the signals that clear this inventory, and buyers paid extra for it. A page generated by a model, rendered quickly, with a real person loading it in a viewable slot, looks excellent to a verification vendor. It simply has nothing on it.
The signals that did separate the two were structural
One measure in the TAG and ANA study split slop from clean inventory cleanly: templated-site rate, at 30.0% on slop against 1.2% on clean supply. Of the 11,552 AI slop domains examined, 88% were also classified as made-for-advertising, which puts most of this supply in a category many buyers already believe they exclude.
That is why the recommendations attached to the analysis point away from scoring thresholds. TAG and ANA advise reviewing invalid-traffic metrics with verification partners, examining exposure to smaller exchanges, and auditing suppression lists for long-tail risk. None of that is a matter of raising a viewability floor. On this data, raising the floor would buy more slop, not less.
What this means for Thai marketers
The study covers open-web programmatic only. It says nothing about walled gardens, search or social, it gives no geographic breakdown, and Thailand is not mentioned in it. What carries over is the shape of the exposure rather than any local figure. A Thai advertiser buying open-web inventory through a DSP is reaching the same long-tail supply pool through the same exchanges, so the audit steps transfer unchanged.
In practice that means pulling a domain-level spend report instead of a campaign-level one, checking how much of the budget clears through smaller exchanges, and treating a suppression list as something maintained monthly rather than inherited. It also means separating media quality from measurement quality: if placement data never reaches your analytics setup at domain granularity, none of this is visible to you in the first place. Retail advertisers running prospecting at scale should fold the same check into ecommerce campaign planning, where open-web display often carries the top of the funnel. Search budgets sit outside this study entirely, so Google Ads search activity should be judged on its own evidence.
Frequently asked questions
How much programmatic spend goes to AI slop?
Between 1.3% and 2.4% of open-web programmatic investment, according to the TAG, ANA and Fiducia analysis published on 28 July 2026. The 1.3% lower bound comes from page-level evaluation and carries a stated 99% confidence interval of 0.98% to 1.56%. The 2.4% upper bound comes from domain-level classification.
Does brand-safety and viewability scoring catch AI slop?
No, and on this data it points the wrong way. Slop inventory was graded premium more than 70% of the time, recorded an invalid traffic rate of 0.05% against 0.32% on clean supply, and reached 77.2% viewability against 74.9%.
Is AI slop cheaper than clean inventory?
No, it cost more. TrueCPM was $7.08 on slop against $6.15 on clean supply in the 28 July 2026 study, roughly 15% higher.
Does the study say anything about Thailand?
No. The analysis gives no geographic breakdown and does not mention Thailand, and it covers open-web programmatic only rather than walled gardens, search or social.
Who performed the measurement?
DeepSee handled domain-level classification and Mobian handled the page-level evaluation, with Fiducia publishing alongside TAG and ANA. All three are commercial vendors operating in this market, which is worth weighing when reading vendor-produced figures, and the primary report document could not be located, so the numbers are as reported by PPC Land.
Where to look first
If you buy open-web programmatic, the fastest useful move is a domain-level spend export for the last 90 days, sorted by spend, read against your current suppression list. Templated pages and small exchanges are where this study found the difference, and neither surfaces in a dashboard built on viewability and invalid traffic alone. If you want a second pair of eyes on where a display budget is actually landing, our team can go through the placement data with you.
