What happened
Google has added three AI features to the Merchant Center dashboard: AI summary insights, AI-powered search suggestions, and a history of your previous AI chats. Search Engine Roundtable flagged the rollout on July 24. None of these arrive in a vacuum — they extend a run that already includes Merchant Advisor and the AI performance reports.
The direction is consistent across every merchant-facing tool Google ships now. The dashboard used to display feed data and leave the interpretation to you. Increasingly it answers questions, summarizes what it sees, and suggests what to do next. The chat history addition is the tell: Google expects merchants to have ongoing conversations with Merchant Center, not just check disapproval counts.
Why this matters
Merchant Center sees a layer of your store you cannot see from the storefront. Attribute-level feed problems, price mismatches between feed and landing page, disapproval patterns across the catalog — that data lives on Google's side of the pipe. An AI summary of it is genuinely useful, because most store owners only open Merchant Center when something breaks. A monthly read of what the machine thinks is wrong costs you ten minutes and occasionally catches something your own diagnostics missed.
Keep your skepticism, though. Google's suggestion layers have a history of blending real optimization advice with nudges that happen to grow Google's revenue — anyone who has watched auto-applied recommendations in Google Ads knows the pattern. An AI summary also compresses, and compression loses nuance: a flagged 'issue' may be trivial for your catalog, or the one line it buries may be the thing actually costing you impressions. The saved chat history helps here. For the first time you can go back and audit what the tool told you and whether acting on it moved anything.
What to do about it
Put a monthly Merchant Center AI review on the calendar
Open the dashboard once a month, read the AI summary insights, and compare them against your own diagnostics before touching anything. Treat it as a free second opinion from the system that runs the auction — it sees feed-level problems you won't catch from the storefront.
Verify before you act on a suggestion
When the AI flags a problem, confirm it in the Products and diagnostics reports before rewriting titles or attributes. A summary can flatten a per-SKU issue into a catalog-wide claim. Fix what the raw report confirms, skip what it doesn't.
Use the chat history as an audit trail
When you act on an AI insight, note the date and what you changed. Next month, reopen the chat history and check whether the advice actually moved clicks, impressions, or approval rates. That log is how you learn whether this tool deserves your trust — for your specific catalog, not in general.
Separate fixes from upsells
Sort each suggestion into two buckets: feed corrections that make listings more accurate, and recommendations that mostly amount to spending more. Take the first bucket seriously. Run the second through the same ROI math you'd apply to any rep pitching you budget.