AdMarketplace Priced AI Chat Ads at CPC While ChatGPT Was Charging $60 CPMs
AdMarketplace, an independent search ad network founded in 2000, ran more than 2 million sponsored placements through roughly 700,000 AI chat queries during a six-month alpha and is now expanding the beta into Dupe and Sezzle alongside Opera's Aria assistant. Pricing is CPC only, no CPM minimums. ChatGPT, by contrast, launched its ad pilot at a $60 CPM and only recently opened CPC bidding in the $3 to $5 range.
A 24-year-old ad network beat the platforms to performance pricing
The story most marketers expect with AI ads goes like this. OpenAI, Google, and Meta build chat-native ad surfaces, set the rules, set the prices, and the rest of the ecosystem either plugs in or gets left out. Two duopolies, maybe a triopoly, and a long tail of nobody.
That story is already getting messier. AdMarketplace, a New York network that started in 2000 as eBay's exclusive search ad partner, has been quietly running performance-priced placements inside Opera's Aria AI assistant for six months. The product is called AMP Discover. It just moved from alpha to beta with five LLM partners, around ten advertisers, and an inventory pool of roughly 200 million product ads pulled from AdMarketplace's existing Arena index. Opera was the first publisher; Dupe and Sezzle are next.
The number that matters is the placement density. Sam Cox, AdMarketplace's chief product officer, told AdExchanger the alpha averaged around three sponsored placements per consumer interaction, and up to fifty ad slots in some longer responses. Two million placements over 700,000 queries works out to almost three ads per query, which tracks. That's a high impression rate for a surface most marketers haven't even tested yet.
What AMP Discover actually does inside an AI chat
The mechanism is straightforward. The LLM partner sends AdMarketplace a query and the conversational context. AdMarketplace returns relevant product ads from its Arena index, and the partner injects them into the response. The advertiser pays only when a user clicks through. No exposure fees. No minimum CPM commitment. No engagement-time pricing.
What's interesting is where the placements end up landing. Cox said the majority fall into "some form of a discovery category." That's a polite way of saying users aren't typing in transactional queries like "buy ASICS Gel-Kayano size 10." They're asking the AI to compare options, suggest alternatives, or build a shortlist. The funnel compresses, and the ads sit closer to the moment a buyer makes a decision than they would on Google or Meta. Cox's exact line: "We're seeing a compression of the funnel."
I think this is the part of AI search ads that most marketers still haven't internalized. Top-of-funnel awareness on AI chat is not the prize. The prize is being in the recommendation list when someone asks "what should I buy." That's where AdMarketplace is plugging in, and it's a different inventory class than what ChatGPT's launch partners were testing.
The CPM problem AdMarketplace is exploiting
OpenAI's first ChatGPT ad units shipped at a reported $60 CPM, which eroded to roughly $25 within ten weeks as advertisers pushed back. OpenAI eventually turned on CPC bidding, dropped its self-serve minimum from $250K to $50K, and is now chasing a $2.5B ad-revenue target for 2026. None of those concessions would've happened if advertisers were happy paying $60 CPMs for exposure inside a chat thread.
Cox put it bluntly: "A CPC model is good for advertisers, because they're not paying $60 or $100 CPMs just for exposure." That quote is doing a lot of work. It implicitly references both ChatGPT's launch CPMs and Perplexity's higher-end programmatic pricing, and it stakes out a position where AdMarketplace's pricing is the friendly version of what the platforms are still trying to defend.
From what I've seen, that's the actual buying pattern emerging right now. Performance teams want guaranteed downside protection on a channel they don't yet have attribution clarity on. CPC gives them that. CPM doesn't.
The platforms charge for exposure inside a thinking surface. AdMarketplace charges for the click out of it. Marketers will pick the second one almost every time, until ChatGPT and Perplexity prove they can match.
Why this matters more than another AI ads pilot
The default assumption has been that AI chat ad spend will consolidate into OpenAI, Google, and a handful of platform-owned surfaces, the same way search and social inventory consolidated. eMarketer's most recent forecast pegs US AI search ad spend at $25.9 billion by 2029, up from 0.7% of search spend in 2025 to 13.6%. That's the number every platform pitch deck is built around.
What AdMarketplace's pilot suggests is that the inventory might not consolidate the way search did. Browsers, second-tier chat apps, shopping assistants, vertical AI products: each one needs an ad partner, and most of them aren't going to build that themselves. AdMarketplace is positioning to be the network in the middle of that fragmented surface, the way OpenX or Magnite ended up running programmatic for publishers that didn't want to build a stack.
That framing isn't certain, and it's worth hedging on. There's a reasonable scenario where Google extends AdSense into AI chat, OpenAI buys an inventory network instead of building one, and the consolidation thesis wins anyway. But the fact that AdMarketplace got a six-month head start with Opera and is already at five LLM partners suggests at minimum that the platforms aren't a foregone conclusion in this category.
There's also a context point most coverage of this beta will miss: AdMarketplace's existing book of business is search. Bing's network, third-party search syndication, browser default placements. AI chat looks structurally similar to that book. Brand-safe, query-driven, performance-priced. AdMarketplace is in a unique position to layer this onto inventory it already monetizes, which is part of why the alpha could scale so fast.
What working marketers should do this week
If you're already testing AI chat ad placements through ChatGPT, Perplexity, or any of the platform-owned surfaces, your benchmark exposure right now is paying somewhere between $25 and $40 effective CPC equivalent on CPM-priced inventory. Get on AdMarketplace's beta waitlist this week, even if you don't activate. Knowing what a clean CPC reference looks like will reset what you'll tolerate from the platforms.
If you're not testing AI chat ads at all yet, this is a reasonable on-ramp. CPC pricing means your downside is bounded by your bid. There's no exposure fee, no minimum spend, no CPM commitment. The product categories that fit best are the ones that already do well on shopping comparison: apparel, consumer electronics, home goods, travel, anything where buyers shortlist before they buy. (See our earlier piece on ChatGPT's logged-out ad surface for the parallel question of who's actually seeing these placements.)
And honestly, this is the part I'd flag to a CMO this week. The story isn't "AdMarketplace launched a beta." It's that the supply side of AI chat advertising is cracking open faster than the platforms wanted, and the floor on what advertisers will accept just dropped. Anyone setting a 2026 AI ad budget on the assumption that CPMs will stabilize at $40 to $60 is probably building a bad model.
The three things worth tracking next
First, watch which advertisers come out of the beta as case studies. AdMarketplace hasn't named the ten beta accounts. If those names are mid-market DTC and retail, the network plays to its strengths. If they're enterprise CPG and finance, the pitch shifts and the inventory probably looks different.
Second, watch the CPC distribution. ChatGPT's $3 to $5 CPC range is a useful comparison point. If AdMarketplace's effective CPCs come in lower (say $1 to $3 across categories), that's a real arbitrage window for performance teams. If they come in higher than ChatGPT, the pitch loses some of its edge.
Third, watch which LLM partners join after the initial five. If the additions are independent shopping assistants and browser-native AI like Opera, the fragmentation thesis holds. If the next partners are platform-aligned (a Microsoft surface, a Meta product), the consolidation case strengthens.
The honest read: AI chat ads are going to be a messier market than search was, and the early money is on whichever network can move first across the most surfaces. That's an unusual position for a 24-year-old company most marketers stopped thinking about a decade ago to be in.
Notice Me Senpai Editorial
