Google AI-Powered Shopping Ads cheat sheet

Google AI-powered Shopping campaigns replace manual product group bidding with automated bidding (typically Target ROAS or Maximize conversion value) combined with AI asset and query matching.

Product feeds remain the single source of truth: AI-generated creatives and query matching draw attributes (title, description, product type, GTIN, brand, price, availability) directly from the Merchant Center feed, so feed quality caps performance.

AI-powered Shopping ads can serve on Search, Shopping tab, YouTube, Discover, and Gmail through asset expansion; the same campaign may generate text and image assets from feed data.

Asset generation from feeds is opt-in per campaign; advertisers can review and remove auto-generated assets, and removed assets stop serving but do not block the campaign.

Conversion tracking must be accurate for AI bidding to work: use Google Ads conversion tracking or GA4 imports with deduplication; missing or double-counted conversions distort Target ROAS learning.

Target ROAS bidding uses a 15-day (or longer) learning period after major changes; changes to budget, target, or feed structure can reset learning and cause short-term volatility.

Final URL expansion and auto-generated assets require the landing page to support the advertised product; mismatched landing pages trigger policy review and can suppress serving.

Campaign priority in Shopping uses a single campaign per country/account structure with inventory filtering via feed labels rather than legacy campaign priority tiers.

Required feed attributes for Shopping ads: id, title, description, link, image_link, availability, price, condition, brand, GTIN (or MPN + brand), and Google product category.

Product disapprovals in Merchant Center block ads entirely; common causes are missing GTIN, price mismatch between feed and landing page, and unavailable products marked in stock.

Feed updates are processed asynchronously; allow up to 24-72 hours for changes to propagate to ads, and use Merchant Center diagnostics to confirm processing status.

Use supplemental feeds or feed rules to correct attributes without rebuilding the primary feed; rules apply at processing time and are visible in feed diagnostics.

Custom labels (0-4) are free-form attributes used for segmentation, bid adjustments, and reporting; they are not shown to users and do not affect policy.

Automatic item updates can correct price and availability mismatches, but persistent mismatches (more than a few percent) can lead to item-level disapprovals or account suspension.

Shipping and tax settings in Merchant Center must match the website; mismatches cause price accuracy disapprovals and can trigger account-level warnings.

Merchant Center account linking to Google Ads is required for Shopping campaigns; unlinked accounts cannot serve product listings.

Target ROAS = conversion value / cost; set the target from historical performance, typically 10-20% below current actual ROAS to allow the algorithm room to scale.

Maximize conversion value with an optional Target ROAS is the recommended strategy for AI-powered Shopping campaigns; Maximize clicks is only appropriate for new campaigns without conversion history.

Budget changes above roughly 20% can reset the learning period; scale budgets gradually (10-20% increments) to preserve stable bidding.

Use the 'Bidding strategy' report and 'Change history' to distinguish algorithmic shifts from feed or competitive changes when diagnosing performance swings.

Segment performance by product type, brand, custom label, and device; AI bidding optimizes at the campaign level, so segment-level ROAS reveals where value is concentrated.

New customer acquisition goals can be layered on Shopping campaigns; incremental conversion value from new customers is often lower ROAS than returning customers, so set separate targets.

Conversion value rules (e.g., location, audience, device) adjust reported value and influence AI bidding; they must be documented because they change the optimization signal.

Benchmark against the 'Auction insights' and 'Competitive visibility' metrics; impression share lost to budget indicates a budget constraint rather than a bidding problem.

Prohibited content in Shopping ads includes counterfeit goods, unsafe products, and restricted categories (e.g., healthcare claims, weapons); violations can suspend the Merchant Center account.

Price accuracy policy: the landing page price must match the feed price at crawl time; sale prices must be reflected in both feed and page during the sale window.

Image policy: primary image must show the product on a plain background without watermarks, borders, or promotional text; violations cause item disapproval.

Title optimization: front-load brand, product name, and key attributes (size, color, model); titles over 150 characters are truncated and reduce match quality.

Product-level reporting in Google Ads shows clicks, impressions, and conversions per item; use it to identify high-impression/low-conversion items for feed edits.

Negative keywords and account-level negative lists reduce wasted spend on irrelevant queries; AI query matching still respects negatives, so maintain lists for branded and competitor terms.

Listing groups (product groups) in AI-powered campaigns are used for reporting and exclusion, not manual bidding; exclusions remove items from serving.

Seasonal and promotional annotations in Merchant Center (e.g., sale price, promotion) can increase click-through rate; promotions require approval and must match the landing page.