Product Feed Optimisation for Google and AI Shopping

Product feed optimisation starts by fixing missing or incorrect data, then adding useful attributes and keeping stock and prices current across every channel. For merchandising managers, the first priority is a feed that describes the exact variant a customer can buy.

Fix disapprovals before expanding descriptions. A product excluded from Shopping ads cannot benefit from better matching until its eligibility problem is resolved.

How a Product Feed Works

A product feed is a structured export of the items you sell. Channels read its records to build listings, match shopper requests and display offers. An XML product feed uses named tags. A CSV feed puts attributes in columns and products in rows.

Your site catalogue is the source customers browse. A product data feed is a channel-specific representation of that catalogue, with its own attribute names and accepted values. An export can omit fields that exist on the site or retain values after the site changes.

The shared product feed components are identifiers, titles, descriptions, prices, availability, images and category data. Product links connect the record to the purchasable variant. Prioritise identity and matching attributes alongside price and availability when auditing a feed. There is no universal order in which all shopping agents read them.

Product feed optimization is the US spelling of the same work. Both describe improving the accuracy and usefulness of an ecommerce product data feed.

Running a Product Feed Audit

Before auditing, get access to your source catalogue, its channel export and Merchant Center. You also need permission to change the catalogue or its mapping rules. Keep a copy of the existing export so you can compare records after a correction.

1. Open Products in Merchant Center and select Needs attention. Record affected products and the issue attached to each one. Check account setup issues as well.
2. Inspect required attributes against Google’s Google specification. Check identifiers and variant fields where the specification makes them conditional.
3. Compare an affected record with its selected product page and checkout. Look for a different price, currency, availability or variant image.
4. Open product and image URLs without signing in. Fix inaccessible pages, blocked images and links that lead to a category page.
5. Correct the source field or mapping responsible for the error. Resubmit the data and confirm the affected record contains the new value.

Merchant Center’s diagnostics are the first free product feed audit tool to use. Needs attention identifies warnings, disapprovals and data mismatches. Google explains these statuses in its Merchant Center issue guidance.

Feedonomics also advertises a free product feed audit. Its requested audit returns a scorecard covering more than 50 common feed issues and prioritised recommendations. It requires submitting a form, so it is a service assessment rather than an immediate self-service product feed grader.

A clean issue list confirms detected errors have cleared. It does not prove every description is useful or every variant is mapped correctly. Inspect records from each category, including a sold-out variant and a product on sale.

Product Feed Enrichment

Product feed enrichment adds factual attributes that help a channel match an item to a shopper’s requirements. Populate colour and size at variant level. Add material, pattern, dimensions and other supported attributes from the product record.

Consider this hypothetical coat record. Its original description says “winter coat”, while the store record also contains navy colour, wool material and a knee-length cut. Export those facts into supported attributes and include them naturally in the description. A shopper asking for a navy wool coat now has explicit facts for the channel to match.

Product feed SEO means using the words shoppers use in factual titles and descriptions. Avoid repeating search phrases or adding claims the product cannot satisfy. Keep titles within this feed-wide audit. The guide to product title optimisation covers title formulas and testing.

Richer data can improve relevance in performance marketing campaigns that use the feed, including Shopping campaigns. The benefit comes from more accurate matching and fewer missing details. A feed edit does not guarantee more clicks or sales. The guide to product attributes explains which details help shoppers distinguish similar items.

Attribute Mapping Across Channels

Attribute mapping connects each store field to the destination’s accepted field. Keep the store’s original data consistent, then transform names and values for each channel.

Store information	Google field	Mapping check
Variant identifier	id	Keep the same identifier for the same variant across updates.
Product page	link	Select the variant described in the record.
Main image	image_link	Show the exported colour or configuration.
Stock status	availability	Convert store status to an accepted channel value.
Material	material	Export the actual composition from the catalogue.
Variant family	item_group_id	Group variants of the same product.

For Shopify product feed optimisation, fix reusable product facts in Shopify first. The Google & YouTube channel imports store data and exposes missing product data for correction. Other Shopify feed apps can transform fields for several destinations. Choose one place to own each transformation so the next sync does not overwrite your correction.

Merchant Center’s attribute rules can modify incoming fields or combine data sources. Google’s attribute-rule instructions explain the required advanced data source management add-on. Use these rules for Google-specific mappings, while keeping shared facts in your catalogue.

If a corrected value returns to its old value after syncing, inspect the source and rule that wrote it last. Editing a channel record alone leaves the upstream error intact. Use the feed management software guide for tool choices and the Google product feed guide for initial setup.

Keeping Stock Current With an Inventory Feed

An inventory feed updates availability without requiring every description to change. Match each stock update to the same product identifier used in the main feed. Update the exact variant that sold out, while leaving available sizes purchasable.

Sync stock whenever it changes if your integration supports event-based updates. Otherwise, schedule updates frequently enough to stay within your stock-change rate. A daily update can suit slow-changing inventory. Fast-selling or low-stock products need shorter intervals. This is an operating recommendation, not one schedule imposed by every channel.

Measure the delay from a stock change to the channel’s processed record. If products sell out inside that delay, shorten the interval or use a faster update method. Google can disapprove products when availability conflicts with the site. An agent using a stale record can also recommend something you cannot ship.

During stockouts, the agent must explain that the requested variant is unavailable and offer suitable in-stock alternatives. Give a restock date only when the connected feed carries one. Never add a sold-out item to the cart. Recheck availability before a cart action because an earlier search result can become stale during a conversation.

Promotion Feeds and Sale Prices

A sale price changes the price of an individual offer. A promotion feed describes an offer with separate conditions, such as a discount code or minimum purchase requirement.

For Google, submit sale_price alongside the regular price. Use sale_price_effective_date to define the sale period when needed. The active price must match the landing page and checkout.

A Google promotion uses promotion_id to connect eligible products with its terms. Set its effective period and redemption conditions to match the offer customers receive. Google reviews promotions before displaying them, as its promotions specification explains.

Check the start and end times in the same time zone as your store promotion. When an offer ends, confirm the channel has processed the regular price again. A feed displaying an expired discount creates a mismatch even when the original submission was correct.

Multipack and Bundle Attributes

Use multipack for identical products you, the retailer, group and sell together. Submit the number of manufacturer-defined units grouped into that offer. Google’s multipack guidance distinguishes retailer-made packs from a pack already created by the manufacturer.

For example, a hypothetical retailer groups four identical single mugs into one pack. The multipack value is 4. If the manufacturer supplies a boxed four-pack that you sell unchanged, do not use this attribute to describe its internal mug count.

Use is_bundle for a retailer-created package with a main product and different complementary items. A camera sold with a separately added case is an example. Google’s bundle guidance requires the attribute for qualifying retailer-created bundles in Australia. Manufacturer-created bundles follow different treatment.

Submit the price of the whole pack or bundle actually sold. Its description and image must also explain the complete offer. Pricing a four-mug pack as one mug sends shoppers to a purchase they did not expect.

Feeds for AI Shopping Agents

AI shopping tools need clear product identity, factual descriptions, variant-specific images and current purchase information. Attributes help them answer requests such as “navy, wool and available in my size” without guessing from a vague description.

The destination still decides the schema. OpenAI’s supported discovery feed requires nine basic fields, including item_id, seller_name and image_url. Its optional fields cover variants and further product details. These names differ from Google’s mapping, so a Google export is not automatically an accepted AI product feed.

OpenAI also documents a Google-compatible upload path that requires confirmation for the registered feed. Follow the supported schema for your integration. Product discovery and checkout have separate requirements. Instant Checkout is available to approved partners and requires a separately enabled checkout integration.

For an agent on your own store, supply the retailer’s catalogue and policy documents through its supported connection. Check its actual field schema before assuming it uses every attribute in a Google feed. Keep delivery and returns information consistent with what the store promises.

Tracking the Results

Track impressions, clicks and disapproval rates before and after a feed change. Compare the same product group over comparable periods. Keep a record of changes to bids, budgets and promotions because they also affect performance.

Calculate the disapproval rate using disapproved products divided by submitted products for the same destination. A hypothetical feed with 20 disapproved products out of 1,000 has a 2% rate. If the next export removes products, a lower count alone does not prove the original issues were fixed.

Confirm processing success before judging campaign results. Then compare relevant impressions and clicks for the enriched group with an unchanged group where feasible. Avoid treating every extra impression as better matching. Inspect the requests or searches behind the traffic when the channel exposes them.

For a connected shopping agent, ask for the corrected variant and an unavailable variant. The expected result is accurate product information and an in-stock alternative when needed. Record whether the answer reflects the processed feed, rather than judging it only by how naturally it reads.

Removing a Recalled Product Fast

Remove a recalled product from every sales feed and recommendation source the same day you identify the affected items. A feed omission alone can leave an earlier record active, so use each channel’s supported removal procedure.

1. Identify the affected product identifiers, variants and batches from the recall notice. Block their purchase in the store.
2. Remove the affected offers from channel feeds and send explicit deletion or removal updates where required. Stop any exporter from recreating them.
3. Update the store documents the agent uses. Replace recommendation passages with the recall notice and customer instructions. Keep safety information accessible to existing buyers.
4. Refresh the agent’s catalogue and document connection. Confirm the affected items disappear from recommendations and cannot enter a cart.
5. Inspect each channel’s processed records and test the affected product requests again. Escalate a retained listing to the channel while the store purchase block stays active.

This feed procedure is one part of a recall. Follow the ACCC’s supplier recall guidance for the wider response. Finish the removal check by confirming the agent can give the recall instructions without suggesting the affected product as a replacement.