IAB blueprint highlights gaps in measuring AI’s sales impact

Consumers who receive a brand recommendation from an AI assistant are 2.5 times more likely to visit that brand’s website within seven days. Yet only 8.8% of those visits are attributed to AI referrals, according to Similarweb data cited by the Interactive Advertising Bureau.

Search receives credit for 55.9% of those visits, while direct traffic accounts for 19.9%. The figures illustrate one of the problems at the center of the IAB’s new AI-Era Attribution Blueprint.

According to Digiday, the project set out to create a framework for measuring conversions influenced by artificial intelligence. The working group included People Inc., Walmart, WPP Media, Criteo, Microsoft, Prudential Financial and other companies.

The final blueprint distinguishes among who or what created intent, influenced a decision and executed a transaction. It also identifies the evidence needed to support different attribution claims and highlights where that evidence remains incomplete.

Caroline Giegerich, the IAB’s vice president of AI, said her own expectations for the project changed during the four-month process.

“So it kind of started in one place, and over the course of four months, ended in another,” she said at a Monday roundtable. Giegerich described the finished document as more of a call to action for the industry.

One example in the blueprint shows how the attribution problem can affect publishers.

A shopper might ask an AI answer engine for the best product in a category. The response could draw from a publisher’s review, but the user may never click the citation before buying the product elsewhere.

The IAB says a publisher could have a contributory claim in that scenario. But content used during model training is generally difficult to trace from outside the model, while a visible citation does not reveal how much a source influenced an answer.

The larger gap is connecting those signals with a later conversion.

“Publishers are feeding the answers but they don’t have the receipts,” said Jaime Schultheis, head of global data partnerships at Bombora.

The IAB calls on AI platforms to identify whether recommendations are paid or organic, disclose their sources and label AI-originated traffic consistently. Publishers, meanwhile, are encouraged to identify the claims they want to make about their content and determine what evidence would support them.

Some publishers and measurement companies are already tracking pieces of that activity.

Time is working with Mobian to measure advertising served to AI agents, while Bustle Digital Group uses Profound to measure generative engine optimization campaign performance. Those efforts provide information about specific activity and visibility, but they do not resolve the broader challenge of attributing AI influence to downstream purchases.

Josh Blyska, head of insights at Profound, said tools can measure factors such as AI visibility and citations. They cannot yet reliably establish whether those interactions ultimately result in a purchase.

Profound’s browser-session analysis found that one in three consumers had a relevant AI conversation about a purchase within a month that went unattributed. The dataset did not include Google AI Mode or AI Overviews.

Additional platform data could provide more visibility.

The U.K. Competition and Markets Authority is requiring Google to provide U.K. publishers with click data for its AI features by Dec. 3. OpenAI is also testing a dashboard that shows how often businesses appear in ChatGPT answers and how many clicks those appearances generate.

Neither development, on its own, solves the broader attribution problem described by the IAB. They instead provide additional evidence about specific stages of the customer journey.

For publishers developing AI-related advertising or measurement products, the blueprint offers a shared vocabulary for comparing what different tools actually measure.

People Inc. chief innovation officer Jon Roberts said vendors sometimes use the same terms to describe different things. He also said publishers remain limited by the data available from AI platforms.

The central problem is therefore not that AI’s influence is entirely invisible. Publishers and vendors can observe citations, visibility, clicks and some campaign activity.

What remains difficult is reliably connecting those signals with a consumer’s eventual purchase. Roberts said publishers are still “flying blind” without more complete data, while independent measurement would remain important for verifying information supplied by platforms.