Advertising’s new ‘universal language’ for AI agents sparks old debates about power and openness

Advertising loves to talk about transformation. It’s less fond of actually doing it.
The latest case in point is the Ad Context Protocol, or AdCP. Billed as a shared, standardised language for how AI agents communicate across advertising, it’s supposed to bring order to the next wave of machine-driven media. Instead, its debut last week triggered more debate than consensus.
To some, AdCP looks like a timely fix for an industry weighed down by fragmentation and technical debt — a way to make the AI tools interoperable before the chaos sets in. To others, it looks like the start of yet another power play, a supposedly “open” initiative that will quietly advantage those who helped write the rules.
Each view makes sense – the optimism and the doubt – because it’s still very early days.
For now, AdCP is launching with two pieces: the Audience Activation Protocol and the Media Buy Protocol. The first helps advertisers define and find audiences. Picture an AI agent parsing a prompt like “travel enthusiasts planning winter vacations with high purchase intent in the U.K.”. The second enables those agents to buy media against those parameters, theoretically allowing publisher-side agents to respond with matching inventory.
Both are positioned as complementary to, not replacements, for OpenRTB — tools focused on orchestration and workflow rather than impression-level auctions. In theory, publishers and ad tech vendors could run OpenRTB and AdCP side by side, with the latter handling the coordination around it — packaging inventory, exchanging campaign constraints and synching measurement, for instance.
But it doesn’t just have to stop there. The standard could eventually spread to out-of-home, linear TV and other channels. The consortium behind it, which includes Scope3, PubMatic, Magnite and Ebiquity, is nothing if not ambitious. It plans to release full documentation and working demos within a year.
“My tech team has been looking into AdCP and quite like the ideas behind it, but in their opinion it’s like miles away from getting the adoption that is needed for it to become a real standard,” said Thomas Lue Lytzen, director of ad sales and tech at Danish publisher Ekstra Bladet. “More of the big players need to join – and what about Google?”
Whether this actually happens is another matter entirely. Every criticism of AdCP is met with an equally valid counterpoint.
Skeptics call it power grab dressed as openness, supporters counter that it’s ownerless by design —shared protocol, not a commercial product. Critics note that PubMatic, Magnite, Scope3 are driving the effort but are hardly neutral players. Fair, say backers but every standard has to start somewhere. And while some doubt the industry will ever rally behind a single language, others point to publishers and retailers already drifting from OpenRTB toward more direct, AI-driven deals. If it ends up disrupting entrenched business models? Even some critics argue that’s long overdue.
Beneath it all sits a deeper argument: whether AdCP risks automating the same flaws that have dogged programmatic since the beginning. If agents start planning and buying media through a shared protocol, the same incentives could follow them: speed over scrutiny, efficiency over accountability, automation over understanding.
“My estimation is this has a 20 to 30% chance to succeed in terms of getting to critical mass and skill,” said Ruben Schreurs, CEO of Ebiquity.
That’s not pessimism so much as realism. Advertising’s history is paved with standards and protocols, each promising to fix what the last one couldn’t. The cycle rarely changes: enthusiasm, adoption, fragmentation and then, finally, fatigue. Maybe, AdCP breaks that loop. But for now, it’s another reminder that change is easier to imagine than deliver.
“We’ll definitely get behind AdCP and build towards it,” said Joe Root, CEO of Permutive. “But for us to really put it at the front of our development list, we’d need to see demand partners. Right now, it’s easier for us to go directly to that side of the market.”
Still, the momentum is hard to ignore. Brands, agencies and publishers are already experimenting with their own agents, each trained to create, trade or optimize ads in different ways. Eventually, those systems will need a shared language to talk to one another. If that sounds idealistic, consider media management as an early example. As more trading becomes automated, advertisers still need a way to pass brand guidelines, compliance policies and creative restrictions from a media management system to a buying platform.
“Everyone is trying to figure out for themselves how to make sure that what they’re doing is somehow compatible with other constituents in the ecosystem that they want to work with,” said Schreurs. “And that’s the only thing this [AdCP) is – a place where systems can meet.”
True as this is, AdCP may also become something bigger: the on-ramp to whatever comes after real-time bidding. Its success is predicated on swapping millions of split-second bids by ad tech for AI agents that can plan, trade and optimize media through a common language. On that basis, it’s potentially the bridge between today’s automated bidding and tomorrow’s fully automated buying.
And like every shift before it, the benefits won’t be evenly distributed. Advertisers and publishers gain cleaner execution and a clearer view of where their money goes. Ad tech firms and agencies, meanwhile, get a new race to prove who still adds value once the machines start talking to each other.
“We will certainly test it [AdCP] to see how it shakes out,” said Ian Maxwell, CEO at ad tech business Converge Digital. “That said, I think the potential downside of this is that it could actually create another divide outside of the walled gardens. And if that happens, agencies may feel like it’s way too complicated to spend money there since it’s splitting in half the agentic web and the open web.”
In the end, AdCP feels like a Rorschach test for an industry that talks about the future constantly but can’t quite agree on how to build it.
“I’d estimate its success odds at 70% in 2-3 years, given AI’s rise and its open-source momentum,” said Karsten Weide, principal and chief analyst at W Media Research. “How do I define success? More than half of players adopt AdCP. The key to success is rapid developer adoption through easy, plug-and-play integrations that hook AI builders early, turning it into the go-to protocol for agentic advertising.”
More in Marketing

Walmart deepens relationship with OnePay, a one-stop finance app it helped create
Walmart-backed fintech firm OnePay this year has added features that further expand its relationship with Walmart and its customers.

How creators are becoming their own networks with Night agency’s David Huntzinger
David Huntzinger of Night says the creator economy is reaching an inflection point, as YouTube experiments with dynamic ads and brands like Blue Apron take influencer marketing in-house.

MMA repositions as Marketing + Media Alliance to emphasize in-market collaborative testing
The new moniker represents the organization’s intent to showcase the nitty-gritty marketing work it has been doing on behalf of its 825-member base of marketers.