The story. Does AI buy programmatic media on its own yet? Not at scale. IAB Europe’s 2026 study put agentic buying and selling at 38% of respondents asked about workflows. Programmatic optimization scored 59%. The survey, covered in Digiday’s Tuesday Ad Tech Briefing, lands beside a debate over a Google Ad Manager option that lets publishers set price floors for individual bidders. Both point at the same question: as buying decisions get automated, who can see them and who can push back? (Digiday, Ad Tech Briefing, 29 September 2026)
The numbers in this piece
01What happened
- AI is everywhere, autonomy is not. IAB Europe found 86% of its 50 respondents use AI for marketing. Among the 29 asked about workflows, 59% named programmatic optimization and 55% media planning and strategy.
- Expectations run ahead of deployment. Asked about the next year, 30% expected agents to become a main way of buying and selling ads in some markets. Another 28% expected regular use without that scale, and no forecast won a majority.
- Results are hard to prove. Respondents rated AI in current ad operations 2.76 out of five on average. Of the 29 asked how they evaluate it, 69% picked operational efficiency and 48% CPM improvement.
- Yield is barely measured. Only 3% of those 29 picked fill rate or yield, and 21% picked audience quality or targeting accuracy. Digiday notes the groups behind these splits are small, so treat them as signals.
- Governance lags. Of the full sample, 78% named an AI governance owner but only 48% had marketing-specific AI guidelines.
- Publishers are debating a lost lever. A MonetizeMore LinkedIn post calls bidder-specific floors in Google Ad Manager a return of control lost in 2019. The final judgment’s language is still being worked out.
02What it means inside a GAM network
The survey measures how buyers evaluate AI: efficiency and CPM first, yield almost never. That is the asymmetry a sell-side operator inherits. If the buy side judges its agents on cost savings, those agents will push clearing prices down unless something in the auction pushes back. Bidder-specific floors are one such lever, but only if they are set against net revenue. Google’s own example, as Digiday relays it, tests a $1 bid under an 80/20 revenue split as $0.80 against the floor.
Digiday also flags what is not confirmed. The claim that bidder-specific floors stay hidden from buyers goes beyond what Google’s documentation explicitly says. Until a publisher can compare floor performance per buyer, a floor is a guess, and an automated buyer will probe it faster than a human trader can adjust it.
The IAB Europe data says agents are early, uneven and mostly unproven on yield.
03What publishers should do about it
04The bottom line
The IAB Europe data says agents are early, uneven and mostly unproven on yield. The price-floor debate says publishers may be getting a lever back at the moment automated buying starts to test them. Digiday puts the test plainly: the value of a control depends on whether publishers can see its effects and use it without sacrificing worthwhile bids.