Attribution Models Systematically Undercount AI Search Demand by Up to 10x
The Development
George Bonaci, VP Growth at Ramp — a company processing over $100 billion in annualized payments for 70,000+ businesses — argues publicly that click-based attribution is now a liability masquerading as a measurement system. As AI Overviews, ChatGPT, and Perplexity synthesize answers without routing users through trackable clicks, the classic search-to-click-to-convert funnel is collapsing. Graphite's research puts AI underattribution at roughly 10x. Separately, a BCG survey of 3,000 senior measurement professionals found that integrated approaches combining MMM, incrementality testing, and multi-touch attribution together produce up to 70% stronger revenue growth — yet only 39% of IAB respondents use all three methods simultaneously.
Our Take
Bonaci's sharpest observation is not about measurement mechanics — it's about what over-reliance on attribution actually does to a portfolio. At Ramp, rigid attribution would have eliminated brand marketing, events, and direct mail in favour of generic paid performance. That's not a measurement failure; it's a strategic one. The deeper issue is that AI commoditizes everything measurable faster than measurement teams can adapt, which means the channels attribution cannot capture — brand, earned presence, AI visibility — are precisely where differentiated demand is now being built. Teams that keep demanding attribution for these channels will systematically defund the work that drives it.
What Changed
Teams now have a structured alternative to attribution for unmeasured channels: triangulating AI Share of Voice, self-reported CRM discovery tags, and conversion outcomes as a three-signal confidence system — rather than waiting for click data that will never arrive.
Marketing Impact
Marketing mix and demand generation functions are most exposed. Budget allocation decisions anchored to last-touch or MTA models will structurally underinvest in AI-visible content, brand programming, and organic authority — the inputs that drive direct visits from AI-mediated journeys.
Competitive Implication
Teams running triangulated measurement — Share of Voice plus self-reported attribution plus incrementality — gain the ability to defend and grow non-paid channel investment. Teams demanding click-based ROI for every channel will cut exactly the programs building AI discovery advantage, ceding ground to competitors willing to operate with looser proof standards.
Strategic Outlook
Expect measurement vendors to accelerate Share of Voice and AI visibility tracking products into Q4 2026 budget cycles. The IAB's finding that only 39% of buy-side teams use all three measurement methods together signals a near-term consolidation toward integrated stacks — and a window for vendors who can package incrementality, MMM, and GEO visibility in a single interface.
The Exploit
Action Item
Growth and marketing ops leaders should run a geo experiment in Q4 2026 — hold out AI visibility investment in two matched markets and measure direct traffic and pipeline delta — to build the internal proof of concept that justifies budget defence without relying on click attribution.