Google Ads

Product Feed Optimization: Visibility in Shopping and AI Search

Your product feed no longer powers just Shopping ads; it feeds AI-driven shopping answers too. This guide shows how to optimize your catalog for the 2026 rules.

Product feed optimization is the practice of structuring the product data you send to Merchant Center (titles, descriptions, GTINs, prices, stock, attributes) so it earns more visibility in shopping ads and AI-driven search surfaces. In 2026 this matters twice over: Performance Max takes most of its bidding signals from the feed, and Google's AI Mode and Gemini shopping answers don't read your website, they read the Shopping Graph your feed populates. A weak feed means you overpay for ad traffic and stay invisible in AI search at the same time.

Why is the product feed the foundation of ad performance?

The feed is nearly everything Google knows about your product. In Shopping and Performance Max you can't bid on keywords; Google matches queries to products using feed data. A missing attribute, weak title or wrong category directly translates into fewer matches, lower impressions and pricier clicks. In practice, feed quality comes before campaign settings: you cannot build a good campaign on a bad feed.

  • Matching: Google pairs search queries with titles, descriptions, categories and attributes; there is no keyword list.
  • AI visibility: AI Mode and Gemini shopping answers draw from the Shopping Graph; a product without a Merchant Center feed never enters it.
  • Disapproval risk: policy violations (promo language in titles, inconsistent prices) get products disapproved and silently invisible.
  • Bidding signal: custom labels carry margin and stock data into campaign structure, steering smart bidding toward profit.
~70
visible title characters (despite the 150 limit)
20% → 80%
a small share of products drives most revenue (typical split)
20-40%
share of products with zero impressions in a typical catalog

How do you optimize product titles?

The title is the single most powerful field in the feed. The technical limit is 150 characters, but users see only the first ~70 on most surfaces, so the important information gets front-loaded. The pattern that typically works best: Brand + Product Type + Key Differentiator + Variant (size, color, material). Keep the attribute order identical across sibling variants so the catalog looks consistent to both users and the algorithm.

  1. Define a title template per major category (e.g. apparel: 'Brand + Type + Fabric + Color + Size').
  2. Place naturally searched terms within the first 70 characters; leave the rest for long-tail detail.
  3. Keep size/color/material order fixed across variants; refresh templates with real phrases from query reports.
  4. For large catalogs, use feed rules to build titles automatically from brand, type and color fields instead of editing thousands of SKUs by hand.
Title optimization: before / afterBeforeAfter0,91,2CTR (%)3446Clicks2,12,3Conversion rate3,23,9ROAS (x)Illustrative example scenario
Typical shift after title templating and attribute completion (illustrative scenario).

How do you segment by margin with custom labels?

Custom labels 0-4 are five free-form fields in Merchant Center; Google doesn't read them, it carries them for your campaign segmentation. The highest-leverage use is assigning every product a score derived from margin percentage, stock velocity and price competitiveness, then writing that score into a label. That way 'high-margin fast movers' and 'slow stock to clear' never share the same ROAS target. You can go deeper on the profit side with the POAS approach.

  • custom_label_0: margin band (high / mid / low)
  • custom_label_1: performance bucket (best seller / standard / zombie)
  • custom_label_2: season or collection
  • custom_label_3: price band (compare with the Merchant Center price competitiveness report)
  • custom_label_4: stock depth (deep stock / last units)
How ad revenue spreads across the catalogTOPLAM100 %Top 20% of products%78Middle 30%%17Tail 50% (incl. zombies)%5
Revenue distribution across the product tail in a typical catalog: a small group carries revenue while a wide tail stays invisible (illustrative).

How do you find and revive zombie products?

A zombie product lives in the feed but gets zero impressions and clicks. Industry analyses put the typical share at 20-40% of the catalog, reaching 80% in some accounts. Because Performance Max piles spend onto proven products, the tail never revives on its own; it takes deliberate intervention.

  1. List products with zero impressions over the last 30 days in a product-level report; that is your zombie inventory.
  2. Blame the feed before the product: missing GTINs, weak titles, wrong categories and single images are the four most common causes.
  3. Split enriched zombies out with a custom label and give them test exposure in a separate low-target campaign.
  4. Consider removing products still dead after 30-45 days; catalog quality lifts the average.

Your feed won't tell you which products it hides

Ads Sensor uses AI to analyze where spend and revenue concentrate across product groups in your Google Ads data, and proposes reasoned actions.

Join the Beta →

How are AI Mode and agentic shopping changing the feed? (2026)

In January 2026 Google announced dozens of new Merchant Center data attributes for the conversational commerce era: answers to common product questions, compatible accessories, substitute products and more. The same announcement introduced UCP (Universal Commerce Protocol), an open standard co-developed with Shopify, Etsy, Wayfair, Target and Walmart and endorsed by 20+ companies; it will power agentic checkout directly on eligible listings in AI Mode and the Gemini app. The practical takeaway is blunt: AI agents don't browse your website, they read your feed. A clean, rich feed is the entry ticket to the agent era.

How do you monitor feed health continuously?

Feed optimization is a process, not a project: price updates, stock changes and policy scans can produce new disapprovals and anomalies every day. A healthy loop watches approvals, price consistency and product-level performance together; a sudden impression drop usually starts in the feed before it shows in the campaign. The feed is also the first place to look when opening up Performance Max's black box. Join the Ads Sensor beta to unify Google Ads, Meta, TikTok and Criteo in one panel with 24/7 anomaly monitoring.

The feed optimization loop1Raw catalogE-commerceplatform / ERP2EnrichmentTitles, GTINs,attributes,images3ScoringMargin + stock +price, customlabels4SegmentationCampaign andbudget groups
Sustainable feed operations: enrichment, scoring, segmentation and monitoring in one loop.

Frequently asked questions

What is product feed optimization?
It is structuring the product data sent to Merchant Center (titles, descriptions, GTINs, attributes, prices, stock) to earn more and better-matched impressions in shopping ads and AI search surfaces. It is independent of campaign settings and comes before them.
How many characters should a product title use?
The technical limit is 150 characters, but most surfaces show only the first ~70. Put the critical information (brand, product type, key differentiator) in the first 70 characters and use the rest for long-tail detail.
What are custom labels for?
Five free-form fields (custom_label_0-4) let you tag products by your own criteria: margin, performance, season, price band. Google doesn't read them; you use them for campaign and bid segmentation. Labeling margin bands is the most practical way to steer smart bidding toward profit.
What is a zombie product and why does it happen?
A product that sits in the feed but receives no impressions. The typical share is 20-40% of a catalog. The most common causes are missing GTINs, weak titles, wrong categories and the algorithm concentrating spend on proven products. Feed enrichment plus a separate low-target campaign can revive them.
Does the feed really affect AI Mode visibility?
Yes. Google's AI shopping answers read the Shopping Graph populated by Merchant Center feeds, not your website. A product without a feed, or with a weak one, does not appear in AI Mode or Gemini shopping results. The conversational attributes and UCP standard announced in 2026 tightened that link further.

Make the whole catalog sell, not just 20% of it

Ads Sensor unifies four ad platforms and GA4 in one panel; AI catches feed-driven performance issues early and proposes reasoned actions.

Join the Beta →