Inventory padded with AI comes out premium on the quality grade
On July 28, 2026 TAG, ANA and Fiducia produced the first statistical measurement of AI generated low value inventory in open web programmatic media. The result: 1.3% to 2.4% of spend. The uncomfortable part is not the share but that this inventory looks better than clean supply on the traditional quality metrics.
The industry finally put a number on it
The Trustworthy Accountability Group (TAG), the Association of National Advertisers (ANA) and technology partner Fiducia published a joint analysis on July 28, 2026. It sets out to measure the share of programmatic media made up of what the industry calls AI slop, inventory built from low value content produced quickly with AI.
The range found is 1.3% to 2.4%. That is given as a share of open web programmatic spend, and sits at the same order as the made for advertising level the industry has long tracked, 1.1%. The work was carried out as part of ANA's Q1 2026 Programmatic Transparency Benchmark.
Two methodologies were used and both produced consistent results. The lower bound of 1.3% comes from rendered page level evaluation with human verification, with a 99% confidence interval of 0.98% to 1.56%. The upper bound of 2.4% comes from domain level classification.
The metrics look good, the inventory does not
The core finding is in the quality measurements. This inventory has an invalid traffic rate of 0.05% against 0.32% for clean supply, so on bot filtering it looks cleaner. Viewability tells a similar story: 77.2% against 74.9%.
The result shows up in price. TrueCPM for this inventory is $7.08 against $6.15 for clean inventory. Advertisers pay more for the low value supply than for the clean kind. The analysis also states that, once measurability is factored in, this inventory earns premium quality grades over 70% of the time.
A few more numbers describe the distribution. Exposure across advertisers ranges from 0.11% to 13.84%, so the problem is not evenly shared. The templated site rate is 30.0% in this inventory against 1.2% in clean supply. On long tail domains the rate rises to 3.7%, roughly one impression in 27. And 88% of this inventory is also flagged as made for advertising.
TAG CEO Mike Zaneis puts it briefly: definitions identify challenges, and data drives improvement. PPC Land relayed the findings in its August 2 roundup, foregrounding the contradiction of junk inventory earning premium grades.
What you are measuring may not be measuring quality
Quality control in programmatic buying rests largely on two metrics: invalid traffic and viewability. This analysis shows both scoring well on this kind of inventory. The existing control set was not designed to detect the problem.
The reason is understandable. An AI generated page sits on a real server, opens in a real browser, and the ad slot genuinely appears on screen. No bots, no hidden ads. What is missing is the value the content carries for a reader, and no metric in the standard set measures that.
The second consequence is on pricing. A higher CPM suggests this inventory is also being favoured algorithmically; with good quality signals, bidding engines may be rewarding it. That reading is ours, the analysis makes no causal claim.
Third, scale. 1.3% to 2.4% may look small, but in a distribution where exposure reaches 13.84% the average misleads. If a seventh of one advertiser's budget flows into this inventory, the issue is not small for that account.
What it means for businesses in Türkiye
The assessment below is our reading, not something stated in the sources. The analysis covers open web programmatic and gives no country breakdown.
A significant share of advertisers in Türkiye spend display budget through broad networks without tight domain level scrutiny, and this picture bears on that habit directly. Four practical notes. First, the placement report: actually read which domains your campaigns ran on, because the higher rate in the long tail means risk rises with broad targeting. Second, exclusion lists: templated sites with no author information that repeat the same content under different headlines belong on the exclusion list, and that is manual work rather than automated. Third, trust in metrics: clean viewability and invalid traffic reports do not mean the inventory is good, which is exactly what this analysis says. Fourth, the axis of measurement: pulling evaluation away from impression quality and toward outcomes, meaning conversions and sales, naturally reduces the impact of this kind of inventory.
In short, a quality grade is a signal, not a guarantee. Looking at outcomes rather than signals is the sturdiest defence here. That last sentence is our own comment.
The UNALSOFT take
In our ad management approach, display budget never runs on a leave it broad and let the algorithm find it logic. The placement report gets read on a schedule and the exclusion list is a living document. This analysis shows what that discipline is worth: quality metrics can look fine while part of the budget goes to pages nobody reads. Moving the measure from impressions to conversions is the fastest way to cut that risk.
Where is your budget actually running?
Let's review your placement report and exclusion list together.