The checks programmatic buyers run to screen out junk inventory now rate AI-generated junk more highly than the publishers it imitates. An analysis published on 28 July by the Trustworthy Accountability Group, the Association of National Advertisers and the data firm Fiducia puts the first measured figure on how much advertising money reaches mass-produced AI content. It found slop takes between 1.3% and 2.4% of open-web programmatic spend, against 1.1% for made-for-advertising inventory in the same quarter.

The size of the problem is not the part that should hold a marketing manager’s attention. The direction of the signals is. Slop inventory recorded lower invalid traffic and higher viewability than clean supply. Once measurability was factored in, it graded as premium more than 70% of the time, and cleared at a higher price than the real thing. The standard brand-safety stack read clean throughout, and the budget followed it.

Why does AI slop score better than clean inventory?

TAG’s analysis recorded an invalid traffic rate of 0.05% on slop inventory against 0.32% for clean supply, and viewability of 77.2% against 74.9%. Those two metrics, alongside measurability, form the backbone of pre-bid and post-bid screening across the buy side. On slop domains all three read well.

The mechanism is not mysterious. Automated content sites depend on impressions clearing verification, so they are built to render fast, serve measurable ad slots and avoid the bot traffic that trips fraud filters. A page assembled by a language model and dropped into a template performs strongly against tests designed to catch fraud and hidden placements. None of those tests asks whether a human wrote anything.

Higher scores then set higher prices. Slop cleared at a TrueCPM of about £5.25 ($7.08) against £4.56 ($6.15) for clean inventory, a premium of roughly 15% per verified impression. Mike Zaneis, TAG’s chief executive, framed the exercise as a matter of sequence: “Definitions identify challenges, and data drives improvement.”

The definition came before the measurement, and it is narrower than the shorthand suggests. Working with supply chain quality vendors, the analysis settled on 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 defining characteristic is content quality, not the use of AI. AI-generated data summaries such as earnings recaps, AI-supported editorial where a human adds edits and perspective, and AI-enabled design tools were all excluded. No vendor consulted defined slop as AI-generated content on its own.

How much of a marketing budget is actually exposed?

Fiducia’s dataset shows exposure varies far more between advertisers than the headline range implies. Across buyers in the same quarter, AI slop accounted for between 0.11% and 13.84% of spend, with the heaviest concentrations in long-tail inventory and particular exchange environments. The gap between the cleanest and the dirtiest buyer is a function of where budget went, not of which suppression list was switched on.

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

One structural marker separates the two populations. Slop inventory carried a templated-site rate of 30.0%, 25 times the 1.2% recorded on clean supply. High viewability and low invalid traffic on template-driven domains function as a signature rather than a quality endorsement. Buyers therefore have a diagnostic that does not depend on any vendor classification being in place. Across the 11,552 domains classified as slop, the topic mix followed a familiar content-farm pattern: parenting, travel, recipes, hairstyles, personal finance and how-to. Common formats were fabricated viral stories and engagement bait, cloned across near-identical templates.

The measured figure covers open-web programmatic only, which is the caveat B2B teams running paid social should hold onto. The analysis identifies social platforms as the primary and fastest-growing slop environment without sizing it. One vendor estimated that 25% to 40% of social video inventory is misaligned, with slop a large and growing share of that. This is a vendor estimate rather than a measured finding, and it sits outside the 1.3% to 2.4% range.

What the analysis asks buyers to check

TAG’s first recommendation is a conversation rather than a purchase. Advertisers and agencies are told to review invalid traffic and viewability metrics with their verification partners, and to establish whether those partners distinguish AI-assisted content from AI slop with any precision. The analysis also asks buyers to examine their exposure to smaller and native-format exchanges, and to review suppression lists and other long-tail controls.

Most slop is already caught in passing. The analysis found 88% of slop inventory also classified as made-for-advertising, so buyers running MFA suppression pick up the bulk of it as a side effect. The remaining 12% is the working gap. That portion falls outside existing frameworks, escapes current tools and costs more per verified impression than clean supply, because it lacks the ad density and arbitrage traffic patterns MFA classifiers look for.

Verification vendors have been building against the problem for over a year, and the analysis names none of them as failing. DoubleVerify’s Fraud Lab exposed AutoBait, a network of more than 200 AI-generated domains, in March 2026. Integral Ad Science moved its low-quality generative AI avoidance segment to general availability on 29 May 2026, reporting a 49% higher success rate on non-slop inventory. Coverage at that point was limited to English-language text on the open web, a limitation that maps onto the same residue the analysis prices above clean supply. For publishers, the recommendation is transparency, responsible AI use and independent certification such as the Alliance for Audited Media’s Ethical AI standard.

Scott Cunningham of Cunningham.tech Consulting conducted and drafted the analysis as part of the Q1 2026 ANA Programmatic Transparency Benchmark. It ran two methods across the same dataset: domain-level classification from DeepSee across $33.97m in matched open-web spend, and rendered page-level evaluation from Mobian across 233 million URLs and 4.45 billion impressions. The lower bound of 1.3% carries a 99% confidence interval of 0.98% to 1.56%.

Made-for-advertising inventory sat between 0.4% and 0.6% of spend through 2025 on the same benchmark. In the first quarter of 2026 it reached 1.1%, a rise the analysis attributes in part to growing sub-types including AI slop.