IAB's attribution blueprint: AI now wants credit for your sales

You know the pattern. Revenue goes up, and the line that grows fastest in analytics is "Direct". Nobody knows where those people came from. That used to mean word of mouth, bookmarks, luck. Increasingly it means ChatGPT, Gemini or Perplexity did the shortlisting, and you saw none of it.

That is the gap the Interactive Advertising Bureau is going after with the AI-Era Attribution Blueprint, published on November 12, 2026. The core, per the IAB and PYMNTS:

• Every conversion gets one of three labels: Human-Owned, AI-Assisted or AI-Executed.
• The guiding questions: who or what created the intent, who made the decision, who completed the purchase?
• There are two kinds of claim, a decision claim and a contributory claim, each with defined evidence.
• It is meant to sit alongside existing attribution, not replace it.

The IAB is blunt about why: clicks and referrals are shrinking, and AI-influenced sales often show up as search or direct traffic. So the next dollar goes to the wrong channel. CEO David Cohen even calls the framework honest about what the industry can't yet substantiate. Humility from a trade body. Mark the date.

This isn't a measurement story. It's a value transfer.

It sounds like a job for the analytics team. It isn't. Attribution decides who gets paid. And the blueprint explicitly addresses AI platforms as an audience. For the first time there is industry vocabulary that lets an AI assistant say: I made that decision.

Now put that next to the ad side. OpenAI is already selling ChatGPT Ads through an exclusive reseller in 44 countries. To me that is the real story: whoever can prove the decision claim walks into the next budget meeting with the strongest argument. My long-held view is that platform terms are a value transfer. Read them, then scale. That now applies to measurement rules too.

Why this gets urgent fast: a PYMNTS Intelligence report from October found that 23% of US consumers, nearly 61 million people, now start online retail research mainly with AI. Those are US numbers. Anyone in Europe who thinks it stays over there missed the smartphone moment.

The catch

The three labels are clean. The evidence isn't. What an AI assistant weighed internally is invisible to the brand. The platform can see it. So when someone proves that the AI made the call, it will most likely be the platform itself. The referee is also on the field. And the full text sits behind a (free) IAB account, so the exact definitions have barely been debated in public.

What I would do now

• Create a dedicated analytics channel for AI referrers (chatgpt.com, perplexity.ai, gemini.google.com, copilot.microsoft.com) so they stop disappearing into "Referral" or "Direct".
• Add "How did you hear about us?" to your contact or checkout form, with "AI assistant (e.g. ChatGPT)" as an answer. Your own data beats someone else's claim.
• My rule of thumb on the AI side: ask several AI systems five real customer questions without your brand name, and log the date and the answer. If you don't know how the AI describes you, you can't audit anyone's claim.

Bottom lineIf you don't measure what AI contributes to your revenue, the AI will soon send you its own number, with an invoice attached.