ARF research across 13 major buyers found five have cut or dropped retail media investment over measurement problems, two of them naming it the primary reason. It landed in the week GroupM put the category at $101 billion, 18% of digital advertising and 11% of all ad spend. For three years the measurement gap has been a request: buyers asked for incrementality and kept spending while they waited. This week it came out of the budget instead.

It is not only retail media. The IAB convened its first Creatorfronts and used it to unveil an initiative mapping where creator deals break down. In AI search, Digiday found marketers fighting over whether chatbots mention them at all, with no standard to settle it. And marketers at State Farm, Mercedes and Hoka are citing AI search behavior to argue for bigger upper-funnel budgets — an argument that works precisely because nobody can price the alternative.

The inventory kept getting harder to see anyway. Google put an open-standard Universal Cart across Search, YouTube and Gmail. Criteo became ChatGPT Ads’ commerce layer on 4 billion product SKUs, and Thrad opened a self-serve platform for buying inside model conversations. Each of those ends inside a surface the seller owns, logs and reports. The clearest case is below.

What measurement did ship was faster, not more independent. Nielsen moved streaming ratings to daily delivery, cutting delivery time by up to 80%, in the same week it appealed a court order to make Cumulus a standalone ratings offer. FreeWheel bought attention and emotion signals from three vendors to package inventory by. The one study that compared the methods found MVPD deals held 61% sustained accurate reach over 90 days against 24% for programmatic CTV.

What it costs you: a measurement gap stops being a talking point the moment it becomes a line item, and five buyers turned it into one this week. Find out what your own number would be before the renewal, not during it.

Nine reports in twelve days, across three story keys the desk has not yet merged, all pointing one way. On September 5 AdExchanger reported that ChatGPT ad buyers had only limited controls over what their ads ran beside, and that roughly 14% of ads served inside ChatGPT had no relevance to the user’s query. Four days later four advertisers told Adweek they wanted prompt-level visibility, the context around each placement, and some evidence that exposure drove a sale — asked as OpenAI’s ad business passed $1 billion in annualized revenue.

The product shipped anyway, and it shipped in the opposite direction. On September 14 Amazon DSP became a managed buy path into ChatGPT with Delta Vacations among the first pilot brands, OpenAI keeping control of delivery and placement. On the 15th Digiday reported Sponsored Agents, where the click opens a branded chat with an AI agent rather than a website, with Wayfair testing at limited scale. By the 17th OpenAI had launched conversational ads with CRM and ecommerce integrations, and EMARKETER was reading the format as the template for AI-native advertising. Buyers asked for visibility in week one. In week three they were sold a format with less of it: no site visit, no referrer, no third party in the room.