By Nisim Tal, CTO, DoubleVerify
Years ago, the industry worried about the “fat-fingers” problem — mobile users accidentally clicking on ads. It was an annoyance, but one we learned to live with. What we’re facing now is something different, bigger, and far more structural: the “bot-fingers” problem.
Recent data from our Fraud Lab shows how quickly AI is reshaping the clickstream. Our team tested how AI browsers and agents behave when navigating webpages. We found that in unprotected media, AI bots accounted for up to 15 percent of all clicks. In certain client-testing and research studies, an ad click was 10 times more likely to be from an AI bot than from a human.
This fundamentally changes what ad “engagement” even means.
As AI agents become more capable, autonomous, and popular, the ad industry needs to move quickly to redefine metrics, guardrails, and expectations. Three major questions now sit at the center of that conversation:
What counts as “acceptable” bot engagement?
AI bots are no longer binary. It’s no longer as simple as “block all” or “allow all.” Some AI agents will act as legitimate proxies for human intent. Imagine a not-too-distant future where you ask an AI agent to find the best laptop deal for Black Friday. If that agent encounters a relevant ad along the way, shouldn’t it click it? It should evaluate that option just as a human shopper would.
But this is where things get complicated. This introduces entirely new forms of bot-driven value that go well beyond traditional use cases like SEO indexing or site testing. If AI agents are going to participate in product discovery and commerce, the industry will need to account for that behavior explicitly. This means redefining what counts as “legitimate” ad engagement and how performance is attributed when a click may no longer come from a human.
What happens when AI traffic starts influencing optimization?
Even when AI bots don’t convert, their activity still feeds the systems advertisers rely on to optimize campaigns — from bidding strategies and creative rotation to audience modeling and budget allocation. If AI-driven bots generate repeated interactions at scale, even without producing real outcomes, that activity can quietly skew performance signals, pushing campaigns to “work” based on engagement data that was never human to begin with.
The risk, then, isn’t just media waste. It’s training optimization systems on behavior that doesn’t reflect real consumer intent at all. As AI-driven traffic becomes a larger share of the clickstream, advertisers will need to rethink how their models adapt — and how to prevent dynamic, automated learning loops from reinforcing the wrong signals.
What does this mean for retail and commerce media?
Retail/commerce media is the fastest-growing segment in advertising, with advertisers expected to invest $175 billion globally this year. The category has thrived because it sits closest to the transaction, where intent is explicit, measurable, and high-value. But that model assumes humans are doing the browsing and buying. What happens when AI agents start doing both?
Shopping environments are uniquely suited to agents designed to execute complex workflows, from product comparison to checkout. Even if AI agents handle only a small share of shopping journeys, the foundations of retail media measurement begin to shift. CPMs, attribution models, and performance benchmarks were built around human behavior, after all — not autonomous agents clicking, comparing, and transacting at robot speed. While it’s still early days, that dynamic could give advertisers who’ve gone all in on commerce media real pause.
The bottom line: We’re entering the “bot fingers” era, where not all clicks are equal, not all bots are bad, and not all shopping journeys involve humans. The industry will need new frameworks and better technology to help clarify what “real” engagement actually means, and it needs them today, not tomorrow.
