The Merchant Center Feed Checklist for AI Mode Ads

Google is not writing your ads anymore. In AI Mode, Gemini pulls straight from your Merchant Center feed and generates the explainer text a shopper sees, which means the attribute you forgot to fill in last quarter might be the reason your product never shows up. That is a hard shift for anyone who spent years optimizing ad copy instead of product data.

The useful way to think about your feed right now is as two separate problems stacked on top of each other. The first is compliance: attributes that, if missing or malformed, get your product excluded from consideration entirely. The second is quality: attributes that are technically present but too thin for Gemini to write anything specific with.

Chasing the new ad formats — Conversational Discovery Ads, Highlighted Answers, AI-powered Shopping Ads — without fixing both tiers is like buying billboard space for a sign you haven't designed yet. Here's how to audit each dimension in priority order.

Identifiers and Titles: Getting In the Room vs. Getting Chosen

Compliance-tier problems here are binary. A missing or invalid GTIN, a blank brand field, or an MPN that does not match the manufacturer's actual part number will get a product disapproved or silently deprioritized. Run your feed through Merchant Center's diagnostics tab first, and treat every GTIN and identifier warning as a blocker, not a suggestion.

This is the single most common reason mid-size catalogs underperform in AI-generated ad formats: a large share of SKUs are simply not eligible to be pulled in the first place.

Once identifiers are clean, the quality-tier gap shows up in titles. A title like "Blue Shirt" is technically valid but useless to a language model trying to match it against a query like "a breathable button-down for a summer wedding." A title with fabric, fit, and use case — something closer to "Men's Blue Oxford Slim Fit Shirt, 100% Cotton" — gives Gemini real material to work with. We cover related ground in how to rewrite old content so it ranks in ai search explained.

The compliance fix gets you into the pool. The quality fix decides whether you get picked out of it.

Images and Structured Data: Eligibility vs. Selection Signal

Image requirements are where most disqualifications happen quietly. Merchant Center rejects images below minimum resolution, images with watermarks or promotional text overlays, and images that do not match the product's actual color or variant. If you're running a Shopping feed today, pull your disapproved-image report first — those products stay invisible no matter how good the rest of your data is.

Structured data on your site forms the quality layer behind the feed. Google's crawlers and Gemini both benefit from schema.org Product markup that echoes what's in your feed: price, availability, review counts, and aggregate ratings. When your on-page structured data and your Merchant Center feed disagree, you create ambiguity that a generative system resolves by simply not featuring the product.

Consistency between the two sources matters more here than in traditional Shopping ads, where a mismatch might just cost you a policy warning.

Availability and Pricing: Real-Time Accuracy vs. Long-Term Trust Signal

Availability is the most time-sensitive compliance attribute in the entire feed. If a product shows as in stock in Merchant Center but is actually sold out on your site, Google can suspend the offer — or the whole account, depending on severity and frequency. Set your feed refresh cadence to match your actual inventory turnover rather than an arbitrary daily cron job, and prioritize real-time inventory API integration over batch uploads if your catalog moves fast.

The quality-tier version of this problem is pricing and promotional accuracy over time. A feed that matches the landing page price at the moment of crawl but drifts during flash sales or regional pricing tests erodes the trust signal Google uses to decide which merchants get surfaced repeatedly.

Because AI Mode sessions tend to run longer than a typical search, shoppers do more comparison work before they ever see your ad. A feed with a track record of accurate pricing earns more consistent placement than one that's only occasionally right.

Description Depth and Attribute Completeness: The Line Between Disqualified and Ignored

A missing product_type or google_product_category attribute is a compliance issue — Google cannot classify what it cannot categorize, and unclassified products get excluded from many of the newer conversational formats outright. Fill every applicable category-specific attribute Merchant Center offers for your vertical, including size, material, pattern, and age group where relevant. These aren't optional extras anymore; they're the taxonomy Gemini uses to match your product to a specific, longer query.

Description length and specificity sit on the quality side, and this is where most catalogs quietly fail without ever getting flagged. A single-line description that reads "high quality for every need" gives a generative model nothing to answer with when a shopper asks for "a neutral running shoe with extra cushion for high mileage."

Write longer, attribute-rich descriptions that name material, fit, use case, and explicit exclusions — what the product is not right for. That specificity is what lets Gemini write a custom explainer against a detailed query instead of a generic one.

If you're running a catalog under a few thousand SKUs with a merchandising team that already updates data weekly, start with the compliance audit. Clear every GTIN, image, and availability flag in Merchant Center diagnostics, and you'll likely see improvement within one feed refresh cycle.

If your catalog is larger or spans multiple regional feeds, treat the quality-tier work — richer titles, deeper descriptions, complete category attributes — as its own project with its own timeline. That's the work that determines whether Gemini ever writes something specific about your product instead of skipping it for a competitor's cleaner data.

Either way, fix the feed before you touch bidding or format selection. In AI Mode, the feed is the ad.