According to a new research report from Tidio, artificial intelligence now influences approximately half of consumer purchase decisions while receiving attribution for less than 1% of retail web traffic, creating what researchers term "dark AI" - commercially significant influence that remains analytically invisible to standard measurement systems. The report, AI in E-Commerce in 2026: The New Shopping Funnel, draws on data from more than sixty sources including McKinsey, Contentsquare, Similarweb, and Bain, revealing a structural gap in how marketers measure and understand consumer behavior.
The mechanism creating this measurement gap is straightforward: consumers increasingly ask AI assistants for product recommendations, receive curated shortlists, then navigate directly to specific brands via new browser tabs or branded searches. These sessions register as direct or organic traffic in standard attribution models, leaving the AI that initiated the shopping journey with no credit. Contentsquare's analysis of actual retail web traffic places AI-referred sessions at just 0.2% of total visits, while McKinsey research indicates half of consumers now rely on AI as their primary or preferred source for product research.
Conversion data from the small fraction of sessions that do get properly tagged as AI-referred suggests the undercounting is substantial. Similarweb's analysis of U.S. retail data finds ChatGPT-referred sessions convert at 11.4%, the highest rate of any measured channel, ahead of direct traffic at 10.2%, paid search at 9.3%, and organic search at 5.3%. This conversion premium implies that tagged AI referrals represent only a high-intent fraction of a much larger pool of AI-influenced consumer journeys.
The attribution gap continues to widen as AI platforms evolve. TollBit's analysis of AI bot behavior across publisher sites shows click-through rates from AI applications dropped nearly threefold during 2025, from 0.8% in the second quarter to 0.27% by year-end, as AI platforms consume more content while generating proportionally fewer outbound clicks. "Brands making budget decisions based on last-click attribution are optimizing for a measurement system that cannot see what is actually driving demand," said Tytus Gołas, Founder and CEO of Tidio.
The financial implications of this measurement gap are substantial. McKinsey projects $750 billion in U.S. revenue will flow through AI-powered search by 2028, with brands that fail to prepare risking 20 to 50 percent of their traditional search traffic. Morgan Stanley estimates AI agents will influence between $190 billion and $385 billion in U.S. e-commerce spending by 2030. The report notes that the inputs determining AI visibility - including feed completeness, structured data, and review coverage - typically live across multiple organizational teams with no clear ownership because current measurement systems cannot demonstrate return on investment.
Protocol infrastructure is emerging to formalize AI's role in transactions, with Google's Universal Commerce Protocol, OpenAI's Agentic Commerce Protocol, and Visa's Trusted Agent Protocol creating standardized rails for AI agents to complete purchases on behalf of consumers. Consumer readiness for AI-assisted transactions is building faster than anticipated: Omnisend's longitudinal research found reluctance to allow AI to complete transactions dropped from 66% to 32% between February and July 2025. The full report is available for download at https://www.getlyro.ai/reports/ai-in-ecommerce.
