AI Doesn't Fix Bad Media Decisions.

Updated: Sep 2

Aren't the robots getting very good at making decisions?
They can analyse billions of signals, identify patterns, adjust bids and optimise towards outcomes in milliseconds. They can even browse the web themselves.
In April 2026, browser-based AI agents accounted for roughly 71% of observed agentic activity on HUMAN's platform. Yet only 3.16% of that activity reached checkout or payment routes.
That's not necessarily a bad thing. AI agents are becoming a legitimate part of how people navigate the web.
But it does make one thing clear: The web is becoming harder to interpret.
And as more media decisions are handed to machines, the quality of the media going into those decisions matters more than ever.
Because automation can make a good decision faster. But it can also make a bad decision very efficiently.
More optimisation doesn't automatically mean better outcomes
Digital advertising has spent years making buying more automated. Bidding, audience selection and creative optimisation are increasingly driven by algorithms.
In June 2026, global invalid traffic rates reached 40%, according to eMarketer, up from 26% in January. The increase was attributed to increasingly sophisticated invalid-traffic operations leveraging AI.
These aren't simply verification problems. They're media quality problems.
Automation doesn't make a signal true. It makes the system better at acting on the signal it is given.
So the question becomes: Are we optimising media, or are we optimising the outcome?
Every campaign starts with an outcome
The answer starts with the brief. Not the format. Not the audience segment. Not the platform. The outcome.
A brand might want to increase consideration. Another might want to drive sales. Another might want to reduce acquisition costs or improve ROAS.
The objective changes, but the principle shouldn't: Every media decision should be made with the end outcome in mind.
That sounds obvious, but much of the media ecosystem is still built around optimising metrics because they're easy to measure.
Those metrics can be useful. But they aren't necessarily what the business asked the media to deliver. If the objective is the destination, every decision along the way should help move towards it.
That means asking better questions about the media we're buying.
Is there genuine demand? Is the supply relevant to the brief? Does it match the opportunity when it goes live?
Demand. Relevance. Match. Three questions that turn media selection from a volume exercise into an outcome-led decision.
Better inputs. Better optimisation.
The answer isn't replacing automation or adding another platform for clients and agencies to navigate.
It's making the existing media ecosystem work better.
AI is exceptionally good at processing complexity, identifying patterns and acting at scale.
Give it better inputs, and it can make better decisions. That's where modern media buying needs to go: less focus on simply optimising the buy, and more focus on connecting every decision back to the outcome.
Because ultimately, the question isn't: How efficiently can we buy media?
It's: How effectively can we turn media into an outcome?
The smarter the machines become, the more important the decisions we make before they start become.
Better inputs. Better decisions. Better outcomes.



Comments