AI Slop Is Turning Media Quality Into A Procurement Test

On 20 October, AdExchanger reported that advertising supply-chain intelligence company DecryptAds had identified 2,116 recipe domains with signs of fabricated authorship, AI-written content and programmatic monetisation. The investigation found major brands appearing across the sites, turning a familiar content-quality complaint into a media-buying question: what exactly did the advertiser pay to reach?

The finding does not establish that every named intermediary sold ads on every domain. It does show why authorised seller records, viewability and low CPMs are no longer enough to establish inventory quality when automated publishing can create plausible sites at industrial scale.

What The Investigation Found

DecryptAds began with roughly 400 sites that researchers manually reviewed for AI-style author photos and text indicators. It then used more than 190 technical fingerprints, including advertising account IDs, DNS records and site infrastructure, to expand the pool to 2,116 domains during analysis conducted from July to October.

Some sites presented invented chefs, detailed biographies and synthetic or altered portraits. DecryptAds said one appeared to use manipulated images of actor Emily Blunt for a fictional baker. The researchers also found crowded ad placements and links through ads.txt and sellers.json records to programmatic sales partners.

Those files matter, but they have limits. An ads.txt entry shows that a seller is authorised; it does not prove that the seller handled a specific impression, that the relationship remains active or that the page delivered value.

9 Ad Fraud and Verification Platforms for APAC Advertisers

Compare nine ad fraud and verification platforms for APAC advertisers, with guidance on media quality, brand suitability, click protection, mobile attribution fraud, regional coverage, evidence access and the proof to demand in a pilot.

Mission MediaMission Media

The Buying Record Is Not A Quality Certificate

That distinction became clearer when vendors responded. MonetizeMore said it was not actively running demand on any of the 11 sites associated with it in the report. Ezoic, which disputed the idea that AI authorship automatically makes content worthless, said quality and usefulness should determine value.

“Whether AI helped produce the recipe does not, by itself, determine its value,” Ezoic said. That is a fair challenge. The issue for buyers is not machine assistance alone, but deception, audience quality and whether a placement contributed to a commercial outcome.

The practical result is that blocklists based only on publisher names or AI detection could create false positives. CMOs need evidence that connects the domain, supply path and actual transaction with post-click behaviour and conversion quality.

Verification Is Moving Beyond Basic Viewability

Other firms are finding the same pressure from a different direction. DoubleVerify reported in March that it had uncovered a coordinated network of more than 200 made-for-advertising domains. Its researchers said pages could contain as many as eight ad banners per slide, with ads refreshing every few seconds.

Google also updated its publisher guidance in October to call fabricated creator profiles a form of deception and “a signal of a low-quality page.” Search policy is not advertising policy, but the change gives marketers another reason to question inventory that borrows credibility from invented expertise.

The adjacent market is already responding with pre-bid fraud controls, content classification and post-click validation. The gap is operational: verification products observe different points in the funnel, so one dashboard cannot prove that every impression was suitable, human and commercially useful.

What CMOs Should Change

For CMOs and regional media leaders, the decision is whether media quality remains a reporting metric or becomes a procurement requirement. Agency scopes and platform contracts should specify domain-level placement reporting, supply-path evidence, reason codes for exclusions and a process for challenging disputed classifications.

APAC adds complexity because buying spans local-language publishers, mobile apps, marketplaces and global exchanges across very different regulatory and device environments. Regional governance should be consistent, but thresholds and coverage need market-by-market testing.

The DecryptAds report suggests that cheap programmatic reach can conceal weak content and uncertain provenance without necessarily producing conventional fraud. The executive response is not to ban AI-assisted publishing. It is to require proof that a placement reached a genuine audience, in a defensible environment, through a supply path the buyer can inspect.