The first statistically rigorous sizing of AI-generated junk inventory puts it at between 1.3% and 2.4% of open web programmatic spend, and finds it scores better than clean supply on almost every quality signal media buyers currently use.

The Trustworthy Accountability Group, the Association of National Advertisers and the technology firm Fiducia today published an analysis quantifying the share of programmatic advertising spend flowing to AI slop, finding it accounts for between 1.3% and 2.4% of open web programmatic investment. That range sits close to the 1.1% measured for made-for-advertising inventory in the same quarter.

The finding that will occupy buyers, however, is not the size of the problem. It is the direction of the quality signals. According to TAG, slop inventory recorded an invalid traffic rate of 0.05% against 0.32% for clean supply, viewability of 77.2% against 74.9%, and, once measurability was accounted for, graded as premium more than 70% of the time. Those scores translated into higher prices: a TrueCPM of $7.08 for slop against $6.15 for clean inventory.

The work forms part of the Q1 2026 ANA Programmatic Transparency Benchmark, a joint initiative of TAG TrustNet, Fiducia and the ANA. It was conducted and drafted by Scott Cunningham of Cunningham.tech Consulting.

What the analysis measured

Two separate methods were run across the same underlying dataset. Domain-level classification was supplied by DeepSee, and rendered page-level evaluation by Mobian.

The domain-level pass covered $33.97 million in matched open web spend. The page-level pass covered 233 million URLs, 4.45 billion impressions and $14.84 million in measured spend, with more than 30,000 page-level evaluations and human verification of flags.

Each method produced a different bound. The lower figure of 1.3% comes from the rendered page-level evaluation with human verification, carrying a 99% confidence interval of 0.98% to 1.56%. The upper figure of 2.4% comes from domain-level classification across the matched population. Data was processed, aggregated and anonymised by Fiducia through its data intelligence platform.

Definitions came before measurement. Through engagement with supply chain quality vendors, the analysis settled on an emerging consensus reading of AI slop as low-value, mass-produced content generated primarily by AI for monetisation, with little or no human input, originality or audience value. The analysis noted that such content carries no demonstrable human fingerprint and fails to meet a premium content experience threshold. Vendors interviewed described it as "zero originality," marked by "semantic shallowness," and as "content that cannot demonstrate what the human author contributed."

That framing draws a line the advertising industry has struggled to hold. The defining characteristic was content quality, not the use of AI in content creation. Categories built with AI assistance and excluded from the classification include AI-generated data summaries such as box scores or earnings recaps, AI-supported editorial where a human adds edits and original perspective, high-quality AI-native products, transparent AI content aggregators, and AI-enabled tools such as graphic design software. No vendor consulted defined AI slop as simply being AI-generated content.

Quality metrics point the wrong way

The inversion of conventional quality signals is the operational finding. Invalid traffic filtering, viewability thresholds and measurability rates form the backbone of pre-bid and post-bid inventory screening across the buy side. On slop domains, all three read clean.

The mechanism is not mysterious. Automated content sites are built to render fast, serve measurable ad slots, and avoid the bot traffic that would trigger fraud filters, because their revenue depends on impressions passing verification. A page assembled by a language model and wrapped in a template performs well against tests designed to catch fraud and hidden placements. It performs badly against no test at all, because none of the standard signals ask whether a human wrote anything.

Higher scores then feed into pricing. Slop inventory cleared at a TrueCPM of $7.08 against $6.15 for clean supply, meaning advertisers paid a premium for impressions rated highly by systems that were never designed to assess authorship.

Concentration and exposure

Exposure varies sharply between advertisers drawing on the same market. Across advertisers in the dataset, AI slop ranged from 0.11% to 13.84% of spend, with higher concentrations in long-tail inventory and certain exchange environments.

The phenomenon is overwhelmingly a long-tail one. Roughly 1 in 27 impressions on unknown domains, or 3.7%, was classified as slop. Large, established exchanges ran single-digit rates. Smaller exchanges and native-format exchanges reached 4% to 8%. Known publishers showed effectively zero AI slop.

One structural marker separates the two populations. The analysis found a 30.0% templated-site rate on slop inventory, 25 times the 1.2% rate recorded on clean inventory. High viewability and low invalid traffic on template-driven domains functioned as a signature rather than a quality endorsement.

Across 11,552 domains classified as AI slop, the topic mix followed a predictable content farm pattern: parenting, travel, recipes, hairstyles, personal finance and how-to content. Common formats included fabricated viral stories, revenge fables and engagement bait built on AI-generated imagery, mass produced across near-identical domain templates and clone networks.

Social platforms and the growth edge

Social platforms are the primary and most rapidly growing AI slop environment, according to the analysis. Platforms are putting slop policies and identification procedures in place, but one vendor estimated that 25 to 40 percent of social video inventory is misaligned, with slop a large and growing share of that.

The estimate lands on territory PPC Land has tracked since 2025. Platform monetisation programmes were identified as a driver of mass-produced low-quality AI content across social media in June 2025, and one-third of the YouTube Shorts feed was described as AI-generated slop by December 2025.

The overlap with MFA

Structural similarity between slop domains and made-for-advertising sites is high but not total. The analysis found that 88% of AI slop inventory was also identified as MFA. The remaining 12% falls outside existing frameworks, escapes current tools, and costs more per verified impression than clean supply.

That residual is the practical gap. Buyers running MFA suppression capture most slop as a side effect. What they miss is the portion that does not present the ad density and arbitrage traffic patterns that MFA classifiers look for, and which is priced above clean inventory precisely because it passes every other check.

The analysis put forward recommendations to advertisers and agencies, including a review of invalid traffic and viewability metrics with verification partners to establish whether those partners distinguish AI-generated content from AI slop, closer examination of exposure to smaller and native-format exchanges, and review of suppression lists and other tools aimed at long-tail risk. For publishers and content producers, it pointed to broad transparency, responsible AI use and independent validation through certification such as the AAM Ethical AI Certification, the framework Integral Ad Science became the first company to obtain in July 2025.

Inside the wider Q1 2026 benchmark

The slop findings sit within a quarter that the benchmark reads as a broad reset. TrueAdSpend, the share of total spend reaching impressions that are fraud-free, measurable, viewable and non-MFA, rebounded to 43.3% in Q1 2026 from 36.3% in Q4 2025. Loss of media productivity fell to 30.8% from a Q4 peak of 40.1%, driven largely by non-viewable impressions dropping from 22.7% to 13.3%. Transaction costs moved the other way, rising to 25.9% from 23.7%.

Pricing eased across the board. Total CPM declined to $4.42 from $5.55, and TrueCPM fell to $6.47 from $12.77 in Q4 2025. The TrueCPM Index, which expresses the premium paid for quality impressions, contracted to 33.9%, the lowest level recorded in the period covered.

MFA is where the two strands meet. The benchmark records MFA sitting consistently between 0.4% and 0.6% through 2025, with Q1 2026 marking the first meaningful increase, to 1.1%. The report attributes that movement in part to growing sub-types including AI slop.

Divergence between advertisers widened. The higher-performing cohort recorded TrueAdSpend of 54.0% against 32.1% for the lower half, a gap of 21.9 percentage points. Transaction costs differed by only 2.4 percentage points between the two groups, while media productivity losses differed by 19.4 points. The higher cohort paid an average CPM of $5.45 and a TrueCPM of $7.46. The lower cohort paid $7.40 and $19.04, or 2.6 times more per quality impression. The upper half ran across 32,998 domains and apps, the lower half across 67,049.

Participation grew. The Q1 2026 sample covered 86 participating marketers and 66 active data contributors, up from 54 and 35 in Q4 2025, spanning 20.9 billion impressions and $160 million in ad spending. CTV accounted for 47.5% of ad spend, web 42.3%.

Why the finding matters for media buyers

Quantification changes what can be procured against. Until today, the scale of AI slop in programmatic spend rested on vendor estimates and investigative case studies rather than a matched dataset with confidence intervals.

The industry has been circling the problem for two years. Integral Ad Science flagged AI-generated slop sites as a threat to programmatic effectiveness in July 2025, and PPC Land published an explainer on the phenomenon and its implications for advertisers in January 2026. DoubleVerify exposed AutoBait, a network of more than 200 MFA domains producing clickbait at roughly $2.25 per article, in March 2026. IAS opened a beta for low-quality AI content avoidance in April 2026 and moved it to general availability in May 2026, reporting a 49% higher success rate on non-slop inventory, though coverage at that point was limited to English-language text on the open web.

The 12% of slop that sits outside MFA frameworks maps onto that coverage limitation. Tooling built for MFA detection and tooling built for text-based AI content detection both leave a residue, and the analysis prices that residue above clean supply.

Broader context on where working media ends up has been documented across the same period. IAB Spain's first technical guide dedicated to supply-side platforms, published in April 2026, cited ANA benchmark data showing that only 41% of total programmatic investment reached genuine, measurable, viewable impressions free of invalid traffic and MFA inventory. The Q1 2026 figure of 43.3% represents movement in that measure rather than a resolution of it.

For buyers, the templated-site rate and the unknown-domain impression rate offer signals that do not depend on a vendor classification being in place. For sellers, the finding that known publishers showed effectively zero AI slop is the counterweight to a market where automated content competes on price with human-produced work.

Timeline

Summary

Who: The Trustworthy Accountability Group and TAG TrustNet, the Association of National Advertisers, and technology partner Fiducia, with the analysis conducted and drafted by Scott Cunningham of Cunningham.tech Consulting. Domain-level classification came from DeepSee and rendered page-level evaluation from Mobian.

What: The first statistically rigorous sizing of AI slop in programmatic media, finding it accounts for between 1.3% and 2.4% of open web programmatic spend, close to the 1.1% MFA level. Slop inventory recorded an invalid traffic rate of 0.05% against 0.32% for clean supply, viewability of 77.2% against 74.9%, premium grading more than 70% of the time, and a TrueCPM of $7.08 against $6.15. Across 11,552 slop domains, 88% also classified as MFA, and the templated-site rate reached 30.0% against 1.2% on clean inventory.

When: Published today, July 28, 2026, covering the Q1 2026 data period from January to March 2026.

Where: Open web programmatic inventory, with the highest concentrations in long-tail supply, unknown domains, and smaller and native-format exchanges. Social platforms were identified as the fastest-growing environment, with one vendor estimating 25 to 40 percent of social video inventory as misaligned.

Why: Standard verification signals rate AI slop as high quality, so inventory with no human authorship passes the pre-bid and post-bid checks buyers rely on and clears at a higher price than clean supply. The 12% of slop that falls outside MFA frameworks escapes current suppression tooling entirely, and the quantification gives buyers a measured baseline where previously only vendor estimates and case studies existed.