AI IN COMMERCE CONSUMER RESEARCH PART 2 OF 3
The AI attribution gap: 92% of shoppers buy AI recommendations, just 31% click trackable links
We surveyed 1,128 consumers in the US to understand how AI is reshaping the path from discovery to purchase—and how many of those sales are already invisible to your analytics. See how many AI-driven sales go untracked, what makes a cited source trustworthy enough to close, and where consumer trust in AI still breaks down at the moment of purchase.


Monique Göttert
Manager, Market Research and Analytics
impact.com
Among AI-active consumers, 92% have purchased a product an AI tool recommended, but only 31% completed that purchase through a link anyone can track. Another 55% deliberately broke the chain, completing the sale in a retailer app, on a brand site, in-store, or through a fresh search.
For partnership marketing teams, every one of those exits is an attribution gap. The AI recommendation and sale happened, but nothing connects them. The instinct is to credit brand awareness or SEO, but some of that growth may be unmarked AI-driven demand, sales your program can’t claim or learn from.
What closes those sales is the source behind the recommendation. For 36% of consumers, the original source an AI cites drives the purchase decision. You may never be able to prove AI drove a given sale, but you can earn a place in the sources AI cites or synthesizes into its answers, and that is how you win the recommendation that closes it.
This report, the second in our three-part series on AI in commerce, maps where AI-influenced purchases finalize outside your program’s view, what makes a cited source close the sale, and why some consumers skip AI altogether to buy direct.
Key findings:
The sale doesn’t always close in the AI tool
Nearly all AI-active consumers (92%) have purchased an AI-recommended product, but over half (55%) finalize the sale somewhere a partnerships program can't track.
The cited source shapes decisions in two different ways
For 36% of consumers, it's the deciding factor—they trust the recommendation because of the source. For a comparable 32%, the source is helpful, but they value the AI's synthesis across multiple sources more.
Sponsorship disclosures in AI build credibility
43% say a "Sponsored" label makes a recommendation more credible—more than double the 19% who find it less reliable.
Delayed purchases fragment further, opening a door for competitors
By the time they return to buy, 21% of AI-active consumers forget the exact product name and fall back to a general search, which a competitor can win.
Trust in AI has limits at the moment of purchase
Purchases that are urgent, require physical inspection, or are high stakes still send consumers past AI and directly to a brand or retailer, in-store or online, no matter how much trust they have in the recommendation.
About the data
This report draws on an impact.com study of 1,128 AI-active U.S. consumers, conducted in May 2026.
All respondents:
Were ages 18-75
Identified themselves as active AI users
Made at least two online purchases in the past 30 days
The survey was conducted at a 95% confidence level with a +-2.92% margin of error.
Researchers intentionally oversampled high-earning consumers who self-reported an annual household income of $100K or greater as a predictive benchmark for the broader market.*
*For full methodology, see Appendix A.