US programmatic display spend is projected to clear $220 billion in 2026, roughly 17.4% growth year over year. The problem is that a growing share of what that money buys is machine-made. AI content mills, template sites, and synthetic engagement are entering the bid stream faster than verification vendors can classify them, and buyers are discovering that raw scale now correlates with lower media quality, not higher reach. That reversal is what is driving a wave of new supply-path scrutiny across the buy side.
The scale problem is now a data problem
The clearest picture of the flood comes from DeepSee.io, which tracks suspicious domains at the source. Its flagged-domain list grew from roughly 17,000 in mid-2024 to more than 108,000 by May 2025, with around 97,000 identified specifically as AI-driven content mills. The firm’s co-founder put a number on the volume: about 143.5 billion impressions of AI-generated, low-quality supply hit the bid stream in January 2025 alone, with roughly 10,000 new junk sites appearing every month.
These properties are engineered for one purpose, which is to maximize ad calls against minimal editorial value. Some publish more than a thousand articles a day. Many are ephemeral, surviving only 30 to 60 days, which is short enough that they cycle through auctions before most exclusion lists catch up. The result is a supply chain where a meaningful slice of spend lands in environments that were never meant for a human reader. Estimates across the sell side put 25% to 30% of open-web spend in wasteful or fraudulent environments.
What the ANA benchmark actually measured
The Association of National Advertisers gave the industry its first repeatable yardstick for this. Its Q1 2026 Programmatic Transparency Benchmark reports a market-level TrueAdSpend Index of 43.3%. That figure is the share of programmatic investment reaching impressions that are fraud-free, measurable, viewable, and free of made-for-advertising inventory. Put plainly, fewer than half the dollars in the sample cleared the quality bar.
The more useful number is the spread. Higher-performing advertisers converted 54.0% of programmatic spend into qualified impressions, while the bottom cohort converted just 32.1%. The 21.9-point gap between them is the widest the benchmark has recorded, and it is built from 86 participating marketers rather than a projection. The gap is the real finding: media quality is not a fixed tax on everyone. It is a function of how disciplined a buyer’s supply path is, and the difference between good and bad discipline is now worth more than 20 points of working spend.
The benchmark also caught the newest variable. MFA exposure in the sample rose to 1.1% in Q1 2026 after sitting between 0.4% and 0.6% through 2025, and the ANA named AI slop as an emerging subtype that needs its own mitigation. The absolute figure is small, but the direction reversed a year of decline, which is what has practitioners paying attention.
MFA did not die, it mutated
For two years the story was that made-for-advertising inventory was in retreat. Jounce Media documented exactly that: MFA sites went from the single largest source of supply-chain waste to under 5% of all web bid requests by June 2025, pushed down by DSP-native flagging, blocklists, and supply-path optimization. On the open exchange specifically, MFA still absorbs around 15% of impressions, down from the roughly 21% of open-web impressions Jounce classified as MFA in 2024.
The catch sits underneath the impression count. Jounce also flagged a 38% year-over-year rise in the number of active MFA domains. Fewer impressions, more properties. That is the signature of AI-assisted site generation, where spinning up a new domain costs almost nothing and the economics work even if each site captures a thin slice of spend before it is caught. The category did not shrink so much as fragment, which makes it harder to block with a static list and easier to reconstitute under a new domain the following month.
The buy side’s answer is curation
Faced with an open exchange that keeps regenerating low-quality supply, buyers have voted with their budgets by moving upstream. More than 66% of the $150 billion-plus open-exchange programmatic market now flows through curated private marketplaces, a near-reversal of 2023, when about 59% of spend ran through the open exchange and 41% through PMPs. eMarketer clocked PMP spending growing close to 13% in 2025 against roughly 3% for the open exchange.
Curation reframes supply-path optimization from a manual auditing chore into a purchasing decision. Instead of bidding into raw supply and cleaning up afterward, buyers transact against pre-vetted Deal IDs where inventory quality is verified at the source. The appeal is not lower CPMs, because curated deals often cost more. It is that a curated path filters out the AI-generated churn before it reaches the bidder, which is the only durable defense when the supply base itself is being manufactured at machine speed.
Where the scrutiny goes next
Standards are the layer buyers lean on to make curation enforceable. The IAB Tech Lab’s supply-chain stack, meaning ads.txt, sellers.json, and the SupplyChain object, lets a buyer trace every participant and every dollar in a transaction down to the final publisher. Coverage is close but not complete: 8.2% of web bid requests and 9.1% of app bid requests still come from sellers with no known sellers.json entry, and that unmapped fraction is exactly where synthetic supply hides. TAG and MRC quality programs sit alongside these specs, but the enforcement gap is the unresolved edge, because a site that lives 30 days can slip through a declaration-based system before anyone audits it.
The practical read for a buyer is that scale can no longer be trusted as a proxy for reach, and quality now has a public benchmark to measure against. Pull the ANA TrueAdSpend figure for your own supply path and treat anything near the 43.3% market line as a starting point, not a passing grade. Move meaningful budget into curated deals where the seller identity resolves through sellers.json, weight verification toward AI-slop and freshly registered domains rather than last year’s blocklists, and audit the gap between impressions served and impressions that actually qualify. The 21.9-point spread the ANA found is the prize, and it goes to whoever treats supply-path discipline as a media-buying skill rather than a compliance afterthought.