1 ChatGPT got a conversion objective — the wall between chat and performance fell
OpenAI launched conversion-optimised campaigns in ChatGPT Ads Manager today, with early access for any advertiser who set up the JavaScript Pixel or the Conversions API by June 1. Until now, ChatGPT campaigns optimised for reach and clicks; today they optimise for purchases, sign-ups, and any pixel-fired event. The pixel tracks website activity after ad interactions; the CAPI streams first-party data back into OpenAI's auction.
This is the structural change performance marketers have been waiting for. ChatGPT ads have been a brand bet because there was no way to bid against revenue. Today, you can. Expect ChatGPT spend to move out of the "AI experimentation" budget and into the performance team's testing queue alongside Google Performance Max and Meta Advantage+. The window of soft CPCs ($3–$5) and $25 CPMs is officially closing — get a test live before the rest of your competitive set does.
2 US programmatic CPMs jumped 34% the same week
PPC Land's pricing data showed US programmatic CPMs up 34% as ChatGPT conversion campaigns came online. The two facts are connected: a new buyer with deep pockets (and a need to fill ChatGPT inventory and external exchange placements) is now competing with Meta, Google, and DSP-based buyers for the same impressions.
Brand-side teams should re-model their media plans for the next quarter on the assumption that CPMs hold the new level, not revert. The cleanest defense is incrementality testing: more efficient channels look more efficient on a relative basis when the whole market clears higher. The teams that don't re-baseline will misread their entire performance dashboard for the rest of 2026.
3 AdRoll exposed its data to Claude and ChatGPT via MCP
AdRoll launched a Model Context Protocol server connecting its advertising data repository directly to external AI applications — ChatGPT, Claude, and any MCP-capable client. Users prompt the LLM they already work in; AdRoll returns performance reports, audience data, and campaign management actions. No dashboard hop, no separate login.
Read alongside ChatGPT's new pixel and CAPI, this is what the new analytics shape looks like: the data sits in the vendor; the interface sits in the chat. The scoring criterion for renewals just got simpler — does your DSP, attribution tool, or ESP expose data via MCP, or are you still paying for someone else's UI? If the answer is the UI, push them on a roadmap or budget time to switch.
4 Netcore took SQL out of segmentation
Netcore launched Audience Agent, a system that interprets plain-language descriptions to build customer segments — "bought baby gear in the last 90 days, no login in 30" returns a list, no SQL, no data-team ticket. The agent reasons over schemas, audiences and product catalogs in real time.
This category has been promised for five years and undershipped for four. Netcore's release is the cleanest version we've seen at scale — but the prerequisite hasn't changed. The agent is only as good as the schema underneath. If your CDP names the same event three different ways, the chat-based segmentation will be confidently wrong. Spend the week on data taxonomy hygiene; the natural-language layer rewards it for years.
5 Bloomreach put an agent on retention — with the budget to act
Bloomreach shipped an AI agent that autonomously detects at-risk customers, picks an intervention, and dispatches a targeted SMS or email discount to re-engage. No manual campaign setup, no rules editor. The agent owns identification, selection, and execution.
Marketing automation runs the rules you wrote; an agent makes the call. The honest risk: an agent's idea of "at-risk" may be miscalibrated against your real margin model, and "save rate" rewards over-discounting. Pilot inside a contained segment with a hold-out group, and measure incrementality (purchases the agent caused) not saves (purchases that may have happened anyway). The wrong KPI is how good autonomy goes financially bad.
6 InMobi and Scope3 shipped a sell-side agent for autonomous buying
InMobi Advertising — partnered with Scope3 — launched a programmatic ad system that opens mobile and TV inventory to autonomous AI buyers. Sell-side agent talks to buy-side agent; humans handle exceptions. The infrastructure step that turns "agents help with bidding" into "agents do the bidding" is now live.
The IAB and Ad Context Protocol working groups are racing to build the vocabulary for this — what's an inclusion list when the seller's agent and your agent are negotiating? Until that standard lands, demand transparency from your DSPs and SSPs on what their agents are agreeing to on your behalf. Specifically: get the agent's brand-safety rules and price ceilings in writing, because the agent is now a contractual party.
7 DoorDash relabelled its ad business commerce media
DoorDash announced the transformation of its advertising arm into a global commerce media platform, surfacing targeted ads across the app and external surfaces. The rename matters: DoorDash now wants to be on RFPs next to Amazon Ads, Walmart Connect, and Instacart Ads — not labelled as sponsored placements.
For food, beverage, and household CPG, DoorDash carries a signal no other retail media network has: time-of-day consumption. Walmart and Amazon know what you bought; DoorDash knows when you ate it. That's a different and more leading-edge dataset for category planning. Test it in the next planning cycle before pricing tightens to match.
8 ANA Masters of B2B closed with the 2026 B2 Awards
The ANA Masters of B2B Marketing Conference wrapped its June 3–5 run at the Chicago Marriott Magnificent Mile today, with the 2026 B2 Awards celebrating the year's best B2B work. The conference's underlying message was that AI-enabled ABM, demand-gen, and analytics now look more like B2C personalisation than the old B2B template.
The B2 Awards are the only US awards judged entirely on B2B outcomes — pipeline, ABM efficiency, lifetime value. The winners' list is the most credible single-document snapshot of which agencies and in-house teams are actually moving B2B numbers in 2026. Use it to short-list partners; the categories are tuned for buyer outcomes, not gallery aesthetics.
9 Google's Preferred badges are nudging clicks — users 2x more likely to click
Google's new AI Search feature letting users mark sites "Preferred" continues to roll out, and Google said users are twice as likely to click links carrying the badge. "Highly Cited" labels and original-reporting carousels are also active inside AI Overviews and AI Mode answers.
For brand and publisher marketers, this is a new earned-media job. The optimisation isn't toward the algorithm — it's toward the reader, asking them to mark you as a trusted source inside Google. Send the prompt with your newsletter, embed it on owned pages, train support to mention it. SEO spent two decades chasing the crawler. The next playbook lives a step closer to community.
10 The Friday read: the AI ad market just hit performance-maturity week
Pull this week's launches together — ChatGPT conversion campaigns and CAPI, AdRoll MCP, Netcore natural-language segments, Bloomreach autonomous retention, InMobi/Scope3 sell-side agent, DoorDash commerce media, Google Preferred badges, programmatic CPMs up 34%. None of them is the headline. Together they're the story: AI-marketing infrastructure crossed from experiment to performance-ready in a single week.
The CMO's planning conversation just shifted. The right question for the second half of 2026 isn't "where do we test AI?" It's "which percentage of the marketing budget already runs through agentic infrastructure, and how do we govern it?" The teams treating the answer as "build the test plan" are about to be outrun by the ones treating it as "build the policy." Performance is here. Governance has to catch up.