Data

When AI closes the sale, who may claim the credit

The IAB's AI-Era Attribution Blueprint, published 12 November 2026, sorts every sale into three pathways and names who may claim credit.

The IAB published its AI-Era Attribution Blueprint on 12 November 2026, sorting every sale into one of three pathways, Human-Owned, AI-Assisted or AI-Executed, and setting who is allowed to claim credit for each. It is a United States framework meant to run alongside existing attribution, not replace it, and it is not UK law. It still forces one decision on UK leaders before the next reporting cycle: keep accepting each platform's own conversion count, or treat those counts as claims to be checked against your own books.

The short answer: in the Blueprint's AI-Executed pathway, one platform can run the auction, execute the buyer's purchase through its own assistant, and write the conversion row that credits itself, three steps owned by the same party. That is marking your own homework with a marker you also sell. Read the pathway closely and you reach a conclusion the IAB stops short of stating: the number that platform hands you is a bid for credit, not a neutral count, and UK advertisers should stop treating platform self-attribution as the revenue their board counts before any standard forces the question.

When AI closes the sale, the platform's conversion number becomes a claim, not a count

The sale you can defend is the one your own systems record, counted once. The double count is already mechanical: Google reports the sale, Meta reports the same sale, your affiliate network reports it a third time, so three systems log three conversions for one real order. You can measure your own exposure this month: add up what every platform claims it drove, then compare that total to the paid orders in your ledger. The platform total will be the larger number, and the gap is pure double count. The AI-Executed pathway widens it, because the party that executed the purchase is now a claimant too, and often it is the same platform producing the conversion figure. When the party that benefits from the credit is the party reporting the number, the number cannot be trusted on its own.

Plenty of performance agencies will tell you to keep trusting platform-side conversion reporting, because that data trains Google Smart Bidding and starving it hurts results. They are right about the algorithm and wrong about the ledger. You can still feed the platforms the signal their bidding needs without treating the figure they return as the revenue your board counts. The two jobs are different, and the AI-Executed pathway is the moment they come apart.

The Blueprint sorts every sale into three named pathways

The framework places each sale into one of three categories, defined in plain terms:

  1. Human-Owned, where a person did the deciding.
  2. AI-Assisted, where an AI helped the person decide.
  3. AI-Executed, where an AI made the purchase on the buyer's behalf.

That third category is the one to watch. As buyers hand more of the choosing and the buying to assistants and agents, the company that owns the AI gains a seat at the credit table it never had.

A claim is the right to be evaluated for credit, not the credit itself

The Blueprint says who may claim credit and what evidence each party must bring, and that is narrower than it sounds. A claim is the right to be evaluated, not credit handed over. It does not split a conversion into percentages, and whoever evaluates the sale still decides how much each party earns. It works more like a register of who may make the case, and what they must show to be heard, than any automatic payout.

The evidence the Blueprint relies on is evidence almost nobody keeps yet

Much of the data the Blueprint depends on is either not collected today or sits inside the AI platforms. There is the practical catch. Part of the proof a claimant must bring is data you do not currently capture, and part of it is locked inside the likes of OpenAI and Google, which you cannot see. The business that waits will be asked for evidence it never kept. Start now: for every order, capture a unique transaction ID, a timestamp, the channel that claimed it and a customer identifier. Those four fields are what let you match any future claim, from a platform or an agent, back to a real sale. Capture them at the point the invoice is paid, not reconstructed from reports later, because a report is the thing you are trying to check.

Instrumenting before adoption costs four fields; adjudicating after it does not

Whether IAB UK or other markets adopt the Blueprint is not yet known, so treat it as a direction to prepare for, not a rule that changes your reporting the moment you read this. Nothing in it automatically alters what a UK advertiser reports. The reason to move early is cost. Before claims arrive, readiness is a one-time instrumentation job: stamp a transaction ID and timestamp on each order. After they arrive, you are trying to trace credit for sales your systems never recorded, which cannot be done retrospectively. The cheap window is the one where no claim is yet in your inbox.

Reconcile every claim against your order record before the next reporting cycle

Nominate one system as the single ledger: the order database, ERP or CRM where the paid invoice actually lands. Every platform, affiliate and AI claim then gets matched to a row in that ledger on the transaction ID, and where two systems claim the same ID, one real sale stands. Before the next reporting cycle, instrument three things in order:

  1. Name the one system that holds the sale that counts, and tell the board that is the number.
  2. Stamp every order with a unique transaction ID and timestamp at the point of sale.
  3. Match each incoming claim to that ID, deduplicate, and report two figures deliberately: the reconciled revenue to your board, and the platform count to the optimiser that needs it.

Do that and the Blueprint becomes something you referee. Leave it and it referees you.

Editorial standards and corrections

Ask this article

A question on this piece?

Ask, and you get an answer drawn only from this article, with nothing invented.

Answers come only from this article. For anything beyond it, speak with Nic.

The Senior Mind

Written by the practitioner
who does the work.

Judged on your profit: a film from The Crane Consultancy
Nic Crane · Founder

A decade engineering profit at scale.

Led paid media for the LUSH Cosmetics global digital transformation across sixteen markets, recognised by Welocalize as Best Global Client Team. A power user of HTML, CSS, JavaScript and Google Tag Manager, now applying that technical depth to AI search: structured data, entity signals and the machine readability that decides visibility in Google AI Overviews and answer engines. Google Partner. Stape Partner. Amazon Ads Partner.

The senior mind that wins the work is the senior mind that does the work.

Start the conversation.

Tools and guides

Put it to work.

The Crane Standard

The senior read,
once a week.

Most marketing newsletters are noise on a schedule. Once a week I send a short, considered read on the developments that actually change what you should do, across the AI shift, PPC, technical SEO and analytics. Filtered, weighted and signed off by me before it goes out. The judgement is mine, and now you can hear every edition in my own voice.

Nic CraneFounder, The Crane Consultancy

Join senior marketers and founders reading The Crane Standard.

Topics you want

All three land every Tuesday. Choose a selection and you get only those, as and when they publish, always on a Tuesday.

One considered email a week, edited and signed off by Nic. No filler, never sold, unsubscribe in a click. By subscribing you agree to our Privacy Policy.

Prefer LinkedIn? Subscribe on LinkedIn instead, the same edition, delivered by push and email.