How AI Agents Can Sweep Fragmentation, Frustration, and Fraud Out of CTV
Scott Fitzgerald’s famous quote about “first-rate intelligence” unfortunately meets cynical reality when it comes to the world of programmatic-led advertising. Holding two opposing ideas in your head may well be a sign of genius, but in programmatic advertising it’s become a recipe for dysfunction.
Take Connected TV (CTV): first, there’s the joint promise of the scalability and ease of digital plus the engagement of the big screen. All good so far, but then second, comes the opposing idea: that CTV as a powerful provider of outcomes can continue to function properly when it’s dealing with an infestation of programmatic bad habits, old and new. Too many integrations. Too many access points to premium inventory. Too much arbitrage. Too many fingers in the till. Too much black box logic gumming up transactions between buyer and seller.
What could save CTV is the arrival of agentic AI, a technology that allows buyer and seller to transact directly with purpose-built agents trained to fulfil all necessary trading, targeting, and verification functions. But before we get to the solution, let’s have a look at some of the problems that hide in CTV’s murky supply chains.
Before agents can get to work, CTV needs a data deep clean
Ask anyone transacting CTV advertising what keeps them up at night and fragmentation will be high on the list. But fragmentation itself isn’t necessarily a problem. In fact, it’s good for the market to have a range of competing platforms and devices, and the last thing we need is more walled gardens encroaching into CTV. The problem is the lack of shared definitions that result from fragmentation.
The consequences range from frustrating to fraudulent. At the less nefarious end of the scale, various CTV sellers not aligning on how to tag and define their inventory makes it a nightmare for buyers to target the right content at scale. You have a broadcaster on one side tagging a Champions League game as simply “Champions League Football” and a streamer on the other calling it “champions-league-arsenal-intermilan-1-20-2026”.
This is an issue agents can fix, but first we need standardised data, otherwise it’s the classic case of SISO (“sh*te in, sh*te out”).
Then there’s the fraud. I’m not here to sling mud so I won’t name names, but there are resellers exploiting the lack of a standardised definition of what exactly CTV means (despite the IAB’s best efforts) to sell anything that might potentially appear on the big screen as CTV inventory. We see online video that is clearly played on a browser packaged up, slapped with a CTV sticker, and sold to unsuspecting buyers.
You could be watching a video on your commute and a reseller could serve you an advert from an unsuspecting buyer who assumed you were kicked back on the sofa with the full living room experience. The result is a creative and targeting misfire, and a total waste of advertising budgets, that can easily slip through the cracks caused by fragmentation.
Agents can deliver programmatic without the pitfalls
Programmatic trading is necessary for CTV to achieve its growth potential. Big Tech made its big bucks not through big brands, but through being able to serve the long tail of smaller buyers who want to be able to easily target a narrow range of consumers (and pay a higher CPM for the privilege). But for this to happen, advertisers need to have confidence that the system works.
Currently, advertisers are having to make investment decisions without a clear understanding of the content, environment, or audience they are reaching. Fragmentation and measurement challenges are often described as technical problems that can only be solved by throwing yet more technical solutions at them, but the reality is they are symptoms of inconsistent definitions and unreliable data clogging up the supply chain.
Should sellers do the necessary legwork to standardise their inventory definitions, agentic AI can smooth over the issues caused by fragmentation. Instead of relying on self-described content and audience labels, buyers can rely on agents trained to identify observable reality itself.
By learning to recognise viewing behaviours that mirror a genuine television experience, agentic systems can distinguish between professionally produced, big-screen content and video that only fits the category with the most loose and generous definitions.
For sellers, the onus becomes what it ought to be: make and distribute content audiences want to watch, and monetise their attention through advertising. The whole wobbling Jenga tower of a tech stack that they have to currently keep upright can be swapped for simple agent-to-agent transactions.
This approach does not railroad the industry into a single definition of CTV. Sellers remain free to describe their inventory as they choose, while buyers gain the ability to interrogate those claims. Agentic AI trained to understand content, environment, and demographic signals in detail allows advertisers (or, more accurately, the agents representing them) to decide what qualifies as CTV and what does not.
These agents still need regular attention to ensure they remain on the right track, but the difference is buyers and sellers won’t have to take it on faith that intermediaries are acting in their best interests and charging a fair fee. Everything is out in the open, giving fraudsters nowhere to hide.
[Editor's note: This is a contributed article from Converge. Streaming Media accepts vendor bylines based solely on their value to our readers.]
Related Articles
By this time next year, agentic AI will sit inside everyday operations no longer as an experiment, but as part of the connected backbone that powers how content is created, managed, and delivered. The question has shifted from if to how, and specifically, how to make it usable across the whole business.
16 Jan 2026