OpenAI added conversion-optimized CPC, AppsFlyer and Adjust integrations, automatic advanced matching and a bulk API, and as of July 27 a "daily budget" is a rolling seven-day average, so single-day spend now overshoots by design. Anyone who opened an account last week needs to reset their pacing alarms.
Why it matters: The self-serve launch was the door. Conversion-optimized bidding plus real mobile-measurement attribution is what makes a test defensible when you have to report on it.
Nielsen turned Ad Intel into a conversational engine across 5.5M brands and 4.6M advertisers in 90+ markets, and exposed it over MCP so customer-built agents can pull competitor spend and creative straight into their own workflow.
Why it matters: Once the data is addressable by anyone's agent, the advantage moves to the agent that knows which question to ask.
Product feed use in Demand Gen hit 35%, up from 16% in 2024, and Smarter Ecommerce warns Demand Gen has drifted down-funnel into "an uncomfortably high degree of overlap" with PMax. Run Uplift Experiments before November or pay twice for the same conversion.
Why it matters: Cannibalization between two black boxes is invisible on a dashboard and provable with an experiment. Run it before peak season, not after.
Passkeys replace password-plus-TOTP for generating new OAuth refresh tokens, hitting anyone standing up Ads Editor, Scripts, BigQuery transfers or a fresh agent integration. Existing tokens and service accounts keep working, and a new passkey sits in a seven-day trust delay, so create them before you need them.
Why it matters: A week-long auth delay lands squarely on the timeline for any agent build wired into a Google Ads account.
The "AI entity footprint audit" scores six dimensions (identity, differentiation, evidence, consistency, relationships, specialization) from 0 to 5, starting by asking ChatGPT, Gemini and Claude to explain the business and watching where confidence drops. It treats the whole evidence base as the unit of work rather than individual pages.
Why it matters: A public, scoreable method for AI-search readiness. Worth reading for gaps in whatever audit you already run.