Google's AI Search Overhaul: GML, AI Mode Data, May Core Update

Install Clarity on at least one high-value content section this week. The grounding queries it exposes only cover Microsoft surfaces (Bing, Copilot), but the intent-decomposition pattern is generalizable: what you learn about how Copilot fans out a question is a useful proxy for how ChatGPT and Perplexity likely behave on the same topic.

Don't pilot ChatGPT ads with brand budgets yet — the inventory is too thin and measurement isn't there. But get on the waitlist; first-mover data on a new answer-engine ad surface is worth the seat.

Use these numbers to size GEO investment against actual reach — 1B MAU is large enough to fund a parallel workstream, not a sidebar experiment. Track follow-up query rate and multimodal share for your category before assuming the topology generalizes.

Ship llms.txt anyway — it's a flat file with near-zero maintenance cost and clear upside for non-Google agents (Perplexity, ChatGPT, Claude). Don't expect a Google ranking impact, but don't wait for Google to bless it either.

Don't make site changes mid-rollout. Mark the start date in GSC annotations and segment your traffic analysis by AI Overview / AI Mode appearance so you can separate core update impact from AI surface cannibalization later.

Read this before any vendor GEO pitch. Most paid GEO playbooks repackage Google's stated guidance with markup added — knowing the source doc lets you challenge consultants' incremental value.

Keep schema implementations tied to specific SERP features (FAQ, Product, Review) where the payoff is documented. Don't sell schema as an AI-visibility tactic to clients — there isn't evidence behind it yet.

Ask your Google rep, in writing, whether your existing Search and Performance Max campaigns will auto-enroll into AI Mode placements when they ship. If the answer is yes or unclear, plan to spot-check placement reporting the day the formats go live so you can opt out before spend leaks into a surface you can't measure.

Treat Ask Advisor recommendations like any other auto-applied suggestion: review-only until you've validated three cycles of its proposals against your own analysis. The bias is toward increasing spend on Google surfaces.

If you run ecommerce, get your merchant team watching the UCP onboarding docs as Google publishes them. The launch partner list (Nike, Target) signals where the inventory floor will be set; sites that don't ship a UCP-ready feed when the spec opens will be invisible to whichever agents adopt the protocol first.

On upgraded sites, audit which WP roles can add or rotate keys in the new Connectors screen and rotate any test credentials used during the upgrade dry-run. Pin the core version in your update workflow so 7.0 doesn't fan out to every client at once before you've watched a representative site in production for a week.

Before this auto-enables in your account, document your current conversion baseline. Qualified Future Conversions will inflate reported value relative to historical numbers; if you don't have a clean before-line, you'll spend Q3 explaining variances you can't isolate.