Google AI Max for Shopping: Feed Readiness Checklist

AI Max for Shopping is Google’s most significant change to Shopping campaigns since Performance Max. It turns your Merchant Center feed from a static product list into a deep attribute database that the AI uses to match conversational and long-tail queries. The beta is opt-in via your Google Ads rep, and the most reliable signal we’ve seen for acceptance is feed quality. This is the checklist.

What changed

Standard Shopping campaigns match queries against product titles and a limited set of structured attributes. AI Max for Shopping does more: Google’s AI extracts attributes from your full feed (descriptions, custom labels, category, materials, use cases) and uses them to answer queries the product title doesn’t mention.

The official framing from Google’s blog: “AI Max for Shopping campaigns let retailers meet shoppers’ intent before they even search for specific product details, using your Merchant Center feeds to transform product data into dynamic Shopping ads that answer conversational queries.”

In practice: if a shopper searches “running shoes for flat feet that work on trails”, AI Max can match a product whose title is “Brooks Adrenaline GTS 24 Stability Running Shoe”, provided the description and custom attributes contain enough signal that the AI can connect “flat feet” to “stability” and “trails” to the shoe’s listed use cases.

That’s a huge shift. It means feed depth matters more than feed breadth, and product attribute completeness becomes the single biggest determinant of campaign performance.

Why the beta acceptance gate exists

Google’s AI Max gating isn’t arbitrary. The campaign type only performs well on feeds with rich attribute data. Without depth in the feed, AI Max has nothing to match against, and bad early-beta results would damage adoption. So Google checks: low disapproval rate (signals operational maturity), high attribute completeness (signals data depth), specific attribute population (signals AI-matchable detail).

The threshold criteria we’ve seen confirmed by beta participants:

• Disapproval rate < 2% over the previous 30 days
• Image link, GTIN (or identifier_exists), brand, google_product_category, description populated for at least 95% of active SKUs
• Average description length > 100 characters (with longer being better)
• Use of at least one of: custom_label_0-4, product_highlight, material, size_type, age_group, target_gender

These aren’t published officially. If you don’t meet them, the Google Ads rep will usually decline the beta request without telling you exactly why.

The 8-point feed readiness checklist

1. Disapproval rate under 2%

The most important single metric. Pull Diagnostics, calculate disapproved / total_active. Above 2% means run a remediation pass first. See our 30-minute disapproval recovery playbook.

Why it matters: a feed with 5%+ disapproval rate signals operational issues that will compound under AI Max. The beta won’t accept you, and frankly the campaign type wouldn’t perform well if it did.

2. Attribute completeness > 95%

For each of these fields, count how many active SKUs have a valid (non-empty, spec-conforming) value:

• image_link
• gtin or identifier_exists=no
• brand
• google_product_category (at leaf level)
• description

If any drop below 95%, that’s the field to fix first.

3. Descriptions are actually descriptive

AI Max extracts more signal from descriptions than from titles. A 30-word description that says “Comfortable running shoe in black” is barely useful. A 250-word description that covers fit (true to size, narrow heel), materials (engineered mesh upper, EVA midsole), use case (road running, daily training, long-distance comfort), and care (machine washable cold, air dry) gives the AI dozens of matchable signals.

Audit description length. Target 200+ words. For SKUs below 100, prioritize rewriting.

4. Leaf-level google_product_category

The Google product taxonomy has thousands of categories, often 4-6 levels deep. Mapping a running shoe to Apparel & Accessories > Shoes (level 2) loses you category-specific surfaces. Mapping it to Apparel & Accessories > Shoes > Athletic Shoes > Running Shoes (level 4) opens them up.

Use the Google product taxonomy and map every SKU to the deepest applicable leaf.

5. Custom attributes for material, age, gender, size

These are the attributes that drive conversational query matching:

• material for “leather running shoe”, “cotton t-shirt”
• age_group for “kids’ bike”, “adult vitamins”
• target_gender for “women’s running jacket”
• size_type for “regular”, “petite”, “plus”, “tall”
• is_bundle for multipack matching
• product_highlight for prominent feature callouts

Not every attribute applies to every category. Populate the ones that do.

6. Title formatting

AI Max favours clean titles. The pattern: Brand + Product Type + Defining Attributes. Avoid:

• ALL-CAPS (signals spam)
• Promotional language (“BEST”, “HOT”, “DEAL”)
• Repeated punctuation
• Generic openers (“Buy”, “Shop”, “Get”)

Most catalogs benefit from an AI title rewrite pass before opting into AI Max. AI Shopping Feeds and equivalent feed managers handle this at scale; doing it manually on a 5,000-SKU catalog takes weeks.

7. Image quality, no overlays

The standard image policy applies. Under AI Max it’s enforced more strictly because the system generates dynamic ad copy around the image, a misleading image (overlay text, lifestyle context as primary, wrong product) breaks the ad more visibly.

Minimum 800x800, neutral background, no text overlay, single product per primary image. See our image validation playbook for the full check.

8. Daily refresh minimum

AI Max prefers fresh inventory signals. A feed that fetches once a week will work but will underperform a daily fetch. For high-velocity catalogs (fast fashion, electronics with frequent price changes), use the Merchant API for real-time updates on price and availability.

After acceptance: enabling AI Max settings

Once Google approves your beta access, two settings unlock the AI Max behaviour:

Text customization. Google rewrites your product titles in-ad to align with the shopper’s query. Required for the full AI Max experience. Without it, you get the matching benefit but not the dynamic ad-copy benefit.

Final URL expansion (FUE). Google can route shoppers to deeper landing pages on your site than the link URL in your feed. Requires text customization. Use with care, if your category pages are uneven in quality, FUE will surface the bad ones.

Both can be toggled off if early results disappoint. Most beta participants report needing 2-3 weeks of optimization before the system stabilizes; don’t toggle off in the first week.

What’s likely coming

Google has signalled that more AI Max controls will arrive over the next two quarters. The most-requested by beta participants:

• Brand inclusion/exclusion controls (so you can prevent AI Max from suggesting competing brands’ products in your dynamic ads)
• Attribute weighting (signal which attributes matter most for your category)
• Custom audiences integration (AI Max is currently audience-light)

For now, the way to influence AI Max performance is upstream: in the feed itself. The teams that win the AI Max beta are the ones whose feeds were already in good shape.

Pre-flight checklist (printable)

• Disapproval rate < 2% over last 30 days
• image_link, gtin/identifier_exists, brand, google_product_category, description populated for > 95% of SKUs
• Description average > 200 words
• google_product_category at leaf level
• material, age_group, target_gender populated where applicable
• Titles follow Brand + Type + Attribute pattern, no ALL-CAPS, no promotional language
• Image minimum 800x800, neutral background, no overlays
• Feed refresh daily minimum

Sources

• Google blog, AI Max for Shopping announcement
• Google blog, Steer performance with new AI Max features
• Google Ads Help, About AI Max for Shopping campaigns (beta)
• Related reads: Top Google Shopping feed errors · Image validation for Google Shopping