Editorial: The AI attribution gap and the case for human judgment
AI powered attribution tools promise clarity, but the data gaps behind them mean marketers still need human judgment to trust the numbers.
Mariano De Vitto · August 2026
Marketing Dive's 2026 predictions roundup surfaces a problem the industry has been avoiding: agencies and brands are pouring budget into AI powered attribution tools while quietly admitting those tools cannot reliably tell them which channels actually drove a sale. That gap between confidence in AI outputs and the messiness of real customer journeys is now shaping how marketing leaders plan next year's spend, and it deserves more scrutiny than it is getting.
The predictions cited in the Marketing Dive piece point to continued growth in AI driven measurement platforms, the kind of tools that promise to stitch together touchpoints across search, social, retail media and connected TV into a single tidy model of influence. These platforms are marketed as the antidote to years of guesswork, promising executives a single source of truth for where marketing dollars are best spent heading into next year's budget cycles. Vendors selling these systems are happy to let marketers believe the math is settled.
The walled garden problem
It is not settled. Multi touch attribution has always struggled with walled gardens like Google and Meta withholding data, and layering machine learning on top of incomplete inputs does not fix the underlying visibility problem, it just makes the output look more authoritative than it deserves.
This is not a new tension. For years, performance marketers have known that platforms controlling the largest share of ad spend also control the data needed to prove that spend worked, and the incentive to share that data openly has never been strong. Retail media networks and connected TV platforms add even more silos to the mix, each with its own definition of a conversion and its own reluctance to expose raw signal to outside measurement vendors. An AI model can only be as honest as the data it is fed, and right now much of that data arrives filtered, aggregated or withheld entirely.
When dashboards become gospel
That authority is the real risk. When a dashboard produces a clean number, marketers are tempted to treat it as fact rather than an estimate built on assumptions the model rarely explains. Budget decisions, agency reviews and even headcount can hinge on figures that no one on the team fully understands or can defend under questioning.
The danger compounds inside organizations where marketing has to justify spend to finance leaders who expect precision. A confident looking output from an AI attribution platform can become the basis for cutting a channel that was actually working, or scaling one that was riding on the coattails of demand generated elsewhere. Without someone in the room asking how the number was produced, the model's blind spots simply become the company's blind spots.
Pairing AI with judgment
Why it matters: marketers should keep investing in AI measurement, it genuinely improves on manual guesswork, but pair every model output with human review. Ask vendors what data the model excludes, test attribution claims against controlled experiments where possible, and train teams to treat AI generated insights as a starting point for judgment, not a replacement for it. As 2026 planning cycles ramp up, the agencies and brands that build in that skepticism from the start will make better calls than those that simply trust the dashboard.
The Signal Brief · Mariano De Vitto — Head of Marketing, Barcelona