PubMatic CEO Rajeev Goel sees AI changing more than how programmatic advertising is executed. He thinks it is changing where the decisions get made.
“With that comes a shift in decisioning to the PubMatic platform,” Goel said during the company’s Q2 earnings call. Read transcript. PubMatic has run more than 80 campaigns through AgenticOS, up from 30 a quarter ago, while deploying more than 20 agents across buying and selling workflows. PubMatic did not disclose AgenticOS revenue.
The bigger change may be Decision Fabric, which lets advertisers, DSPs and technology partners run their proprietary models directly inside PubMatic’s infrastructure, closer to its data, inventory and the point of auction. Goel says that gives buyers more time to make bidding decisions and lets them use more sophisticated models.
Why might decisioning migrate? Goel argues that AI is eroding one of the DSP’s historical advantages: the buying interface. “The lock-in from that UI layer is now eroding,” he said. Increasingly, he argues, competitive advantage will come from “AI-native infrastructure, proprietary intelligence, and the ability to consistently deliver advertising performance.”
Read: PubMatic Announces Second Quarter 2026 Financial Results (August 6) – PubMatic
Related research: Could publishers and advertisers both benefit from changing where auction decisions get made? A recent paper highlighted by analyst Eric Seufert models an alternative auction design that increased publisher revenue and advertiser welfare in simulations. Seufert on LinkedIn | Paper
From tipsheet: The last era moved data to the decisioning system. AI makes it increasingly possible to move decisioning to the data. The SSP doesn’t have to become a DSP to take on more of the DSP’s traditional role. Buyers can bring their intelligence closer to the seller’s data and inventory without either side necessarily surrendering its core asset.
Related: PubMatic CEO Rajeev Goel on building an AI-native advertising platform (July 23) – tipsheet
PLATFORMS
The Trade Desk: Defending buyer decisioning
During last Thursday’s Q2 2026 earnings call, The Trade Desk CEO Jeff Green focused much of his commentary on an issue larger than a quarterly earnings miss: who gets to make the decisions in programmatic advertising.
Green argued that some advertisers are trading decisioning for lower transaction costs, particularly through programmatic guaranteed and fixed-price deals. “Doing so essentially means buyers will give away their decisioning in a great buyer’s market to the sellers in exchange for lower costs of transactions,” he said.
Digiday’s Seb Joseph connects Green’s argument to The Trade Desk’s push into measurement. If advertisers can better measure incremental business results, the theory goes, they can distinguish cheap media from valuable media and make better buying decisions.
Read: The Trade Desk is betting part of its recovery on measurement nobody can argue with (August 6) – Digiday
More: The Trade Desk Q2 2026 Earnings Call Transcript (August 6) – Yahoo Finance
From tipsheet: Put Green’s argument next to PubMatic’s and a bigger question emerges as AI spreads across advertising: Why should decisioning belong exclusively to one side of the market?
Buyers should have the opportunity to apply their own data, objectives and intelligence to determine what an impression is worth to them. Publishers should have the same opportunity to apply their intelligence to inventory, audiences, scarcity, yield and other considerations.
AI could make independent decisioning on both sides possible. Rather than giving decisioning to a marketplace or intermediary, buyers and sellers can bring their own intelligence to the transaction. Market infrastructure will connect intelligence, not own it.
Also: Magnite is describing a similar architecture, using containerization and real-time data to let partners run proprietary decision engines closer to the auction. Mike Laband on LinkedIn | Magnite blog
MEASUREMENT
Nielsen + DoubleVerify: Speed the loop
Nielsen’s $2.15 billion acquisition of DoubleVerify (Variety, 8/6) brings audience measurement together with media quality, performance and optimization.
From tipsheet: Advertising has always operated through a feedback loop: buy media, measure what happened, learn and adjust. AI doesn’t create that loop. It compresses it.
Nielsen says “AI-driven planning, activation, and optimization” will increasingly rely on “verified data, outcome signals, infrastructure, and platform integrations.” That helps explain the strategic logic of the deal.
As former IAB executive Angelina Eng writes, measurement is becoming “less about delivering individual metrics and more about providing decision support.” Nielsen + DoubleVerify brings more of those signals together. AI can shorten the distance between those signals and the decisions they inform.
Call it feedback-loop compression: reducing the time between an advertising decision, the resulting signal and the next decision. The change isn’t what measurement does. It’s how quickly its signals can inform what happens next.
More: Nielsen to acquire DoubleVerify, creating a leading independent media intelligence platform (August 6) – Nielsen
AGENTS
Debate: Time serves ads to AI agents
Last week’s announcement by Mobian of its Agent Ads product inspired a thoughtful discussion on LinkedIn that’s worth a look.
- “Serving ads to AI agents will not be effective and I would caution publishers and advertisers against doing it…” (August 5) – Software engineer William Cichowski on LinkedIn
- “Respectfully, I think the test and the product are measuring different things. Your methodology tests whether off-topic content survives retrieval, and that is not what we do…” (August 6) – Jonah Goodhart, CEO, Mobian in the comments of Cichowski’s LinkedIn post
- “You 100% can get publisher-injected ads to show up in organic Google/ChatGPT/Claude summaries. The trick is inferring the search agent’s intent well enough…” (August 6) – Steven Liss, CEO, OpenAds in the comments of Cichowski’s LinkedIn post
From tipsheet: The better agents get at understanding advertising, the more advertisers may have to compete on verifiable information rather than traditional persuasion.
In the wild: “Time Is Serving AI Bots a Different Website, With Ads Built In” (August 5) – Software engineer Vincent Schmalbach on his personal blog
Commercial demand, buyer skepticism: Clients are hungry for AI visibility aids, but buyers are skeptical of specialized markdown ads (August 7) – Digiday (subscription)
LLMs & CHATBOTS
Developments
- Responding to the next frontier of critical cyber capabilities (August 7) – OpenAI
- Improving Fable 5’s biology safeguards (August 7) – Anthropic
- Legal AI startup Harvey in talks to raise funding at $15.5 billion valuation (August 7) – The Information (subscription)
MORE
- VideoAmp Cuts Staff as AI Agents Take Center Stage (August 7) – The Wall Street Journal (subscription)
- Cloudflare Wants to Tell Brands Whether AI Is Actually Recommending Them (August 6) – Adweek (subscription)
- Behind Reddit and YouTube’s roles in AI visibility (August 7) – Marketing Dive
- Kroger Adds ‘Product Listing Ads’ To AI Shopping Assistant (August 4) – MediaPost