IAB: AI Attribution Hindered by Lack of Platform Data

The IAB publishes a framework to credit the influence of LLMs on conversions, while noting the lack of actionable evidence within platforms. However, openings are emerging: Google is required to share clicks in the UK, and OpenAI is testing a dashboard, although the link between AI visibility and sales has yet to be established.

Data promised by Google and OpenAI, but blind spots remain

The UK Competition and Markets Authority is requiring Google to provide British publishers with click data from its AI search features, with a deadline set for December 3. Meanwhile, OpenAI is experimenting with an analytical dashboard that tracks the number of times a business appears in ChatGPT responses and the clicks generated to its site. This information is eagerly awaited by publishers like BDG, who consider it valuable for their geographic offerings. According to some officials, the absence of this data leaves publishers in the dark. Sharing analytics is presented as beneficial for both platforms and media, and the need for third-party verification to audit the provided figures is emphasized. Despite the publication of the IAB framework, many aspects remain unmeasured.

Publisher experiments: increased visibility, incomplete attribution

Publishers are testing solutions in the field. Time is collaborating with Mobian to measure the performance of its ads shown to AI agents, while Bustle Digital Group relies on Profound to evaluate its geographic campaigns. These efforts provide visibility on certain segments of the journey but still do not reliably connect AI influence to sales. They come at a time when groups are expanding their geographic offerings, a development hampered by the ongoing difficulty of measurement.

What the framework formalizes: decisions, expected evidence, and transparency gaps

The IAB framework proposes distinguishing who creates the intent, who makes the decision, and who executes the transaction, associating each claim with specific evidence. It highlights transparency deficits that complicate the verification of these claims and makes several recommendations: indicate whether recommendations are paid or organic, disclose sources, and consistently label AI-origin traffic. It also invites brands and agencies to apply standards for claims and evidence, and publishers to identify the necessary evidence to support the influence of their content. A typical case describes an AI response relying on a publisher's review: without a click on the citation, the sale is often counted as search or direct traffic by the retailer. The use of content for training is not traceable from the outside, grounding and citation are observable but insufficient to prove the extent of influence, and the link between sourcing signals and downstream conversions is lacking to fully support a contributory claim.

Why attribution falters: AI journeys without clicks and insufficient tools

AI-driven journeys often lack impressions, clicks, and benchmark data, causing sales from recommendations to be attributed to other channels. Measurement officials note that current tools struggle to connect the influence of AI search to conversions. Some measurement platforms can track AI visibility, citations, and brand presence in responses, but do not reliably link these interactions to purchases. A browser session analysis conducted by Profound showed that one in three consumers had a relevant AI conversation in the month leading up to their purchase that remained unattributed, a result established without including Google’s AI mode or its previews. This phenomenon is described as a tricky problem at this stage.

What stakeholders say: missing receipts, constrained budgets, and support needs

Publisher and measurement officials observe that media provide responses but lack the evidence to demonstrate their role in conversion. The framework helps compare offerings in an environment where proprietary solutions are multiplying, using identical terms for different definitions. On the advertiser side, the absence of first-hand data from AI platforms limits the extension of budgets towards LLMs. Several executives emphasize the need to better understand how models respond and how to attribute the observed effect, as CTRs are not measured in this context. They call for support similar to that offered by major social or video platforms. From the IAB's perspective, officials describe the initiative as a call to action in the face of platform opacity and missing evidence. Finally, publishers mention agreements with AI companies providing compensation for content and references; at the Washington Post, 95% of AI references received are said to come from OpenAI, according to its executive editor.