IAB reassigns AI-driven sales: why ChatGPT recommendations disappear from tracking

The Interactive Advertising Bureau introduced a new standard framework on the 12 of November 2026 designed to make AI recommendations visible along the purchase journey. The AI-Era Attribution Blueprint distinguishes whether people buy on their own, whether AI supports the decision, or whether an agent executes the purchase. The central question for marketing teams: who gets credit for a sale when the first decisive contact happens in an AI answer rather than on a website?

Traditional last-click reporting often only sees the last Google search result, the direct visit or the shop session. A customer may, however, have researched the product days earlier in ChatGPT, Gemini or a shopping agent.

Three purchase paths instead of a single conversion number

The industry body defines three categories:

• Human-Owned: A person develops the purchase intent, decides and completes the purchase themselves.
• AI-Assisted: An AI system helps with research, selection or recommendation; the decision remains essentially with the human.
• AI-Executed: An agent carries out the final purchase or transaction on the user's behalf.

For every journey, it should be documented who triggered the intent, made the decision and executed the transaction. Afterwards, a platform, publisher or other participant can claim credit for a contribution to the sale, but only on the basis of traceable evidence. A mere assertion is not yet recognised attribution.

The new blueprint thus builds on the earlier IAB framework for AI Visibility: while that focused on mentions and sources in AI answers, the focus now is the link to the business outcome. We already explained the differences between brand mention, citation and referral in our article on AI visibility measurement.

Why many AI recommendations show up as search or direct

Similarweb data cited by the IAB in its study illustrates how large the gap can be. Users who had been recommended a brand by an AI assistant visited that brand's website 2,5 times as often in the following seven days as other users in the compared group. Yet only 8,8 percent of these visits were recorded as direct AI referrals. 55,9 percent came via search, and 19,9 percent via direct access.

In such journeys, a classic analytics dashboard could therefore list search or direct as the most important contact, even though the brand had already entered the relevant set in an AI conversation. This misattribution directly influences where budget flows.

Measurement practice itself also shows pressure to adapt: according to the IAB Outlook Study 2026, September update, 45 percent of surveyed advertisers name comparing AI-driven and conventional journeys as their biggest measurement challenge; 44 percent rely on branded search or direct traffic as proxies for AI influence.

Publishers face a new attribution question

The role of journalistic content is particularly interesting. A user asks an assistant for a product recommendation whose answer is based on a review, and then buys in the shop. The review may have contributed to the decision without ever receiving a registered click.

That is why the IAB distinguishes between claims to the decisive contact and to a contributing touchpoint. Publishers must be able to show that their content was used and, where available, that it is linked to a later conversion. The IAB's recommendation to platforms is to separate paid and organic recommendations, disclose sources, and label AI-driven traffic consistently.

This touches on the debate about Google AI Overviews and the lack of publisher click data. At the same time, the new ChatGPT Business Analytics match brand mentions and clicks as the first part of a broader measurement stack to come.

What marketers can add to their reporting now

Rather than immediately reweighting Search, Social and Direct, teams can systematically fill the data gap: AI recommendation, visible source citation, AI referral, branded search and downstream conversion should be tracked separately. Surveys conducted after the initial product discovery and controlled lift tests can also help.

AI agents are also blurring the distinction between human visits and automated actions. The new IAB blueprint provides a shared vocabulary for this – but platforms, shops and publishers now need to make the actual transaction records jointly accessible.

Sources - IAB: Official press release on the AI-Era Attribution Blueprint, 12. November 2026 - IAB: AI-Era Attribution Blueprint – framework and categories - Digiday: The new IAB attribution model and the data gaps - MediaPost: Background on the IAB attribution framework