Prepare Your Product Feed for Google’s New AI Shopping Report

Google is preparing a Merchant Center report that will show how products are discovered through AI Mode, AI Overviews in Search, and the Gemini app. For sellers, the useful part is not another dashboard score. It is the chance to see which product terms and attributes matter in conversational shopping, then fix the catalogue evidence that supports them.

Google’s help page, dated 27 May 2026 and still labelled “coming soon” when reviewed on 7 October 2026, lists four report areas: share of voice, shopping funnel performance, product term insights, and product attribute insights. The planned rollout covers the United States, Canada, Australia, India, and New Zealand. Google has not promised an exact account-by-account launch date, so treat the report as a feature to prepare for, not something every merchant can use today.

What the report is designed to reveal

Share of voice will compare a brand’s visibility on AI-driven shopping journeys with similar brands. Shopping funnel performance will divide activity into discovery, evaluation, and purchase stages. Product term insights will surface popular terms used in Search conversations. Product attribute insights will highlight specifications people ask about, such as colour, style, and material, alongside an attribute completeness score.

That combination matters because conversational queries are often more specific than a short keyword. A shopper may ask for “small recycled silver hoops with a secure clasp for daily wear” rather than “silver earrings.” A seller cannot control the wording of an AI answer, but they can make the product easier to understand by supplying accurate attributes and images that visibly support the claim.

Google’s current product data specification says accurate product data is a foundational input for its AI-powered formats and experiences. It also warns that missing variant attributes, low-quality images, and conflicts between a feed and the product page can restrict eligibility or produce incorrect displays. That makes the new report most useful as a catalogue-quality tool, not as a prompt generator.

Start with a baseline before changing anything

Before the report reaches your account, export or record the current state of the products that matter most. Note each item’s title, description, category, colour, material, pattern, size, primary image, additional images, and variant grouping. Also record a simple commercial baseline such as impressions, clicks, product-page views, add-to-cart rate, and sales for the same period.

This prevents a common mistake: changing titles, attributes, images, and prices at once, then assigning credit to the most exciting edit. The new report may show where visibility or funnel performance is weak, but it will not automatically prove that a single image caused the weakness.

Turn conversational terms into an evidence check

When product term insights become available, group the terms by the detail a shopper is trying to verify. A jewellery seller might see clusters around metal, finish, fastening, scale, gifting, or skin sensitivity. A homeware seller might see capacity, dimensions, compatibility, or cleaning. A fashion seller might see fit, fabric weight, occasion, or weather.

For each cluster, check three places. First, is the fact present in a structured feed attribute where Google provides one? Second, is it stated accurately on the product page? Third, can a shopper verify it in the imagery without relying on imagination? These are different jobs. The colour attribute can say “rose gold,” while the primary image must show the rose-gold variant rather than a reused yellow-gold photo. A description can mention a lobster clasp, while an additional close-up can prove the clasp shape.

Do not add an attribute simply because it appears in a popular query. Only add facts that are true for that SKU. Attribute completeness is not a licence to guess.

Use funnel stages to choose the next image

The funnel view should change which visual you make next. Discovery-stage weakness suggests the product may not be recognisable or relevant at a glance. Check whether the primary image cleanly shows the exact item, fills the frame sensibly, and matches the selected variant. Evaluation-stage weakness points toward missing proof: alternate angles, scale references, packaging, texture, fastening, or what is included. Purchase-stage weakness may be caused by price, shipping, trust, availability, or the landing page, so do not assume another lifestyle image will solve it.

For a ring, the discovery image might be a clean view of the complete design. Evaluation images could show the profile, setting height, hallmark, and the ring on a hand for scale. For earrings, show the front, side, fastening, and a worn view. For a boxed gift set, show the full contents together so buyers do not mistake a prop for an included item.

Google’s image guidance says the primary image is the first thing shoppers see and must show the exact product. It recommends images around 1500 by 1500 pixels or larger for the best performance across formats. Google has also announced that a minimum of 500 by 500 pixels will apply to all product images from 31 January 2027. Additional images can show other angles, details, staging, or the product in use.

Run small, traceable catalogue experiments

Choose one high-value term cluster and a manageable product group. Fix missing factual attributes first. Then update the images needed to prove those facts. Keep a control group of similar products if your catalogue is large enough, or compare equivalent time periods while noting promotions and stock changes.

A useful experiment could be: add accurate material and colour attributes to 20 under-specified jewellery variants, replace reused variant photos with exact images, and add one fastening close-up. Leave prices and titles unchanged for the test window. Watch whether visibility, evaluation-stage activity, clicks, and conversion move together. The report can guide prioritisation, while your own analytics test whether the work helped.

AI image tools can assist with clean backgrounds or contextual scenes, but they should not invent the evidence you are trying to add. A generated close-up is not proof of a clasp, engraving, gemstone setting, or included accessory. Use a real photograph when small construction details affect the buying decision. If an image is generated or substantially AI-edited, preserve the required AI metadata and follow the applicable disclosure rules.

What not to conclude from the report

A higher share of voice is not the same as profitable sales. A popular product term is not a recommendation to stuff that phrase into every title. An incomplete attribute score does not mean every optional field is relevant. A drop at the purchase stage does not prove the images are weak.

The report also describes aggregated AI-shopping visibility rather than a deterministic ranking recipe. Google does not say that filling one field guarantees inclusion in AI Mode, an AI Overview, or Gemini. The defensible approach is to improve truth, consistency, and visual proof across the feed and landing page, then measure the commercial result.

A practical first-week checklist

When the report appears, save the initial view before editing. Identify one product term cluster with commercial relevance. Match it to factual structured attributes. Check that the landing page says the same thing. Audit the primary image for the exact variant, then add the missing detail view that helps a buyer evaluate the claim. Change one coherent batch, document it, and compare the result with your baseline.

Google’s new report should make conversational shopping less opaque. Its best use will not be chasing every phrase that appears. It will be finding where shopper questions, structured product data, and visual evidence fail to agree, then fixing the smallest useful gap.

Sources

Google Merchant Center Help, “Insights for AI-powered shopping experiences coming soon”: https://support.google.com/merchants/answer/17117204?hl=en

Google Merchant Center Help, “Product data specification”: https://support.google.com/merchants/answer/7052112?hl=en

Google Merchant Center Help, “Image link [image_link]”: https://support.google.com/merchants/answer/6324350?hl=en

Sources reviewed 7 October 2026. Rollout timing, report fields, image requirements, and regional availability can change, so check the current Merchant Center documentation and your own account.