OpenAI told EEA and Swiss users on Friday that Free and Go tiers get ads later this month, selected on current conversation topic, general location and device type only. Advertisers get total views and clicks and nothing else, so the first European ChatGPT buys are contextual placements with no audience lever and no path to attribution.
Why it matters: Set the measurement expectation before anyone buys. This channel ships without the reporting most teams will assume is already there.
Robert Simpkins, after a year building agentic systems for Google Ads, argues most teams have not exhausted off-the-shelf: export account data to ChatGPT or Claude for structure audits and wasted-spend analysis, then wire live data through pre-built MCP connectors to Ads, Analytics and Merchant Center. Custom builds only earn their keep when you need margin, stock and CRM data fused, or continuous monitoring instead of prompted runs.
Why it matters: The MCP-connector tier is the fastest win available and the honest starting point. It also tells you whether you actually need a custom build.
Last week's media buyer index has Meta at 55.16% of budget share returning +0.01% ROAS, three straight weeks of share without a return signal, while Snapchat posted +11.55% ROAS and +28.18% CvR on 0.77% share and AppLovin hit +8.49% under 2%. The recommendation is a 10-15% incremental shift into both with CAC ceilings, before the auction reprices the efficiency.
Why it matters: Cross-platform reallocation math is a weekly deliverable an agent can assemble, and it is the easiest place to show value fast.
OtterlyAI's Agent Analytics reads server logs to surface ChatGPT-User, Claude-Bot, Perplexity-User and Google-Agent visits, which pages they hit, and how that maps to appearing in AI answers. Standard JS-based analytics miss all of it by design, which means most AEO reporting to date has been inference on top of a blind spot.
Why it matters: Server-log agent traffic is the missing input to AI-visibility work, and it is a baseline nobody can argue with.
DeepLumen says 1,000+ merchants are live on Agentic Page with AI-driven orders attributed back to brands, and it is commercially rolling out a cost-per-sale model. Charging on attributed AI orders means somebody believes the attribution holds, which is a further claim than most of this category will make.
Why it matters: Pressure-test the attribution logic before betting on it. CPS pricing on AI-driven orders is the model buyers will start asking about.
Goodway Group is feeding buyer agents real client media briefs and grading what audiences and inventory the seller agents come back with, while restricting them to pre-approved paths. Andrea Kwiatek's concern is the old one in new clothing: a seller agent curating inventory that optimizes for the seller, not the advertiser.
Why it matters: Every agentic buying pitch deserves the same question: who is the agent on the other side of the trade optimizing for?