We checked which ad platforms an AI agent can actually operate. One of thirteen.
Agentic AI in campaign management is the headline trend of the year. Every deck has it on slide four. So we stopped reading about it and went through the platforms we actually buy on, checking one narrow, testable thing.
The short version
- One of thirteen. Only Taboola’s Realize platform exposes an interface an AI agent can operate directly. The other twelve are hand-operated.
- An API is not an agent interface. Google Ads has one of the best advertising APIs in existence. It is a build, not a connection — weeks of engineering, not an afternoon.
- Better models won’t fix it. The constraint is access, not intelligence, and it is a commercial decision by platform owners.
- The advantage available from AI today sits in the strategy layer, not the execution layer.
The claim doing the rounds is that AI has moved past automation, and that marketing teams now run systems which plan, execute and adjust multi-step campaigns with limited manual input.
Some of that is true. A lot of it describes software that already existed, renamed. What we wanted to know was narrower and more useful: can an AI agent connect to this ad platform and operate the account directly — read a campaign, change a bid, pause a line item — without a person clicking through the interface?
That question has a yes or no answer. So in July 2026 we went through thirteen platforms and answered it.
The distinction that decides everything
This is where most of the confusion lives, and it is worth being precise about.
Google Ads has one of the most mature advertising APIs in existence. It is excellent. It is also built for engineering teams, with credentials, developer tokens, review, and a codebase to maintain afterwards. That is a build, not a connection.
The gap between those two things is months of engineering, and almost every “AI is running our media” claim quietly sits on the wrong side of it.
What we found
| Platform | Agent interface |
|---|---|
| Taboola / Realize | Yes |
| Google Ads | No |
| Meta | No |
| No | |
| Microsoft Advertising | No |
| Amazon Ads | No |
| The Trade Desk | No |
| DV360 | No |
| Criteo | No |
| Outbrain | No |
| JioHotstar | No |
| Spotify | No |
| MGID | No |
One of thirteen, or 7.7%. Every other platform on this list is hand-operated, or operated by software somebody built and maintains.
What this means if you buy media
- If an agency tells you AI is running your campaigns, ask which platform and what it changes. On twelve of these thirteen, the honest answer is that a person is doing it, possibly assisted by a model that suggested it. That is not the same claim, and the difference matters when you are paying for the second one.
- Better models will not close this. The constraint is access, not intelligence. Platform owners decide what to expose and when, and they have commercial reasons to be slow — an agent that reallocates budget away from a platform is not a feature that platform is in a hurry to ship.
- Native and discovery is the exception, and it is not a coincidence. The platform that opened up is the one under most competitive pressure for spend. That pattern suggests which platforms open next, and it is a reason to have native architecture in place already rather than after everyone else arrives.
Where AI does move the needle today
None of this is a case for scepticism about AI in marketing. It is a case for putting it where it currently works.
Performance data analysed at a speed no human matches. Creative variants generated in bulk. Competitive review, drafting, and stress-testing a plan before budget commits. Every one of those is judgement work being accelerated.
Which lands somewhere specific. The advantage available from AI today sits in the strategy layer, not the execution layer. The part genuinely being transformed is the thinking — and the part everyone is talking about, the machine quietly running your account overnight, is on twelve of thirteen platforms not yet a thing that happens.
How we tested this, and what would make it wrong
- Our own assessment as of July 2026, not a vendor survey or a commissioned study.
- We hold live accounts on six of these platforms and tested those directly. The remainder were assessed from published interfaces and documentation, which is a weaker method and we’d rather say so.
- We excluded third-party middleware that automates by driving the browser interface. It exists, it sometimes works, and we would not put a client’s ad account behind it.
- “No” means no agent interface at the time of testing. It does not mean no API, and it does not mean no automation — several of these platforms have very good automated bidding. That is the platform’s own automation, not yours.
- This will change, possibly quickly. We will re-run it and update this page rather than quietly leave it up. The date at the top is the date that matters.
Questions about this
- Can AI actually run ad campaigns in 2026?
- On one of the thirteen platforms we checked, yes — an AI agent can connect and operate the account directly. On the other twelve a person is doing it, possibly assisted by a model that suggested the change. Those are different claims.
- Is an advertising API the same as an AI agent interface?
- No, and this is where most of the confusion lives. An API is built for engineering teams to write software against over weeks, with credentials, developer tokens and a codebase to maintain. An agent interface is something an assistant connects to and starts operating the same afternoon, with no software written in between.
- Which ad platform can an AI agent operate?
- Taboola, on its Realize platform, was the only one of the thirteen we checked in July 2026 that exposes an interface an AI agent can operate directly.
- Will better AI models close this gap?
- No. The constraint is access rather than intelligence. Platform owners decide what to expose and when, and they have commercial reasons to be slow — an agent that reallocates budget away from a platform is not a feature that platform is in a hurry to ship.
Written by Pranay Mohan
Co-founder of Scale and Beyond. Ten years in advertising technology and performance media, built inside global advertising platforms rather than on the agency side alone, across more than 100 advertisers. This piece exists because he spent those years on the platform side, where you learn what these systems are optimising for rather than what their dashboards report. More about the founders.
Why we published this
We’re a new firm with no client results to show yet. What we can show is how we work out what’s true: take the narrow testable question, answer it against real accounts, and publish the answer including the parts that are inconvenient. If that’s the kind of thinking you want on your account, the next step is a conversation.
How We Do Paid Media → Book a Call, No Pitch →