AI Max for Shopping: The Product Feed Optimisation Guide for 2026
Jack Amin
Digital Marketing & AI Specialist
Quick Answer
AI Max for Shopping expands automation across targeting and creative, but the system still depends on accurate merchant data and useful landing pages. Optimise the feed by fixing diagnostics, using specific product titles, completing relevant attributes, supplying high-quality images, aligning price and availability with the website, and segmenting reporting by product economics. Test incrementally and judge results by profit or qualified revenue, not traffic alone.
Automation Does Not Repair Bad Product Data
Google's AI Max for Shopping expands automation in Shopping campaigns, including how demand is matched and how creative assets can be used. That makes the quality of the product catalogue more important, not less.
The system can only reason over the information it receives. If the feed calls an item “Classic Shirt”, omits colour and size data, uses a cluttered image and disagrees with the landing page on price, automation has weak inputs and customers have a weak experience.
The right preparation is a feed-quality programme tied to commercial measurement.
Fix Diagnostics Before Rewriting Titles
Start in Merchant Center diagnostics. Separate issues into:
- disapprovals that stop products serving
- warnings that limit quality or reach
- data inconsistencies between feed and landing page
- policy issues that require operational or legal correction
Assign each issue to a source system. Patching the exported feed every week is fragile if the product information system continues sending the wrong value.
Track the number and revenue share of affected items. One disapproved bestseller can matter more than 500 warnings on discontinued products.
Write Titles for Identification
A useful title helps a customer and platform distinguish the product quickly. The best attribute order depends on category, but often includes:
Brand + product type + defining model or feature + size/colour/variant
Examples:
- vague: “Performance Runner”
- clearer: “Brand Performance Running Shoe – Men's – Black – Size 10”
Do not stuff synonyms or promotional claims into the title. Put the most discriminating information early because some placements truncate the rest.
Maintain category-specific title rules rather than one template for the whole catalogue.
Complete the Attributes That Change the Decision
Identifiers such as brand, GTIN and manufacturer part number should be accurate where they exist. Never invent a GTIN to silence a warning.
Then complete attributes customers actually filter or compare:
- colour, size, gender and age group for apparel
- material and dimensions for furniture
- capacity and compatibility for electronics
- condition and warranty information where supported
- multipack, bundle and unit-pricing details where applicable
Use product types that reflect your own merchandising hierarchy and official categories that match the item. Consistent taxonomy makes reporting and feed rules easier to maintain.
Treat Images as Product Evidence
The primary image should show the product clearly, at useful resolution, without distracting promotional overlays. Add alternate images that answer buying questions: scale, angle, texture, packaging, included accessories and real-world context.
The landing page should use the same core product identity. A feed image showing one variant while the URL opens another creates both policy and conversion risk.
High-quality visual assets also support discovery in Google's image-rich and generative experiences. They should be accurate before they are beautiful.
Keep Price and Availability Synchronous
Price, sale price, currency and availability must match the landing page. Frequent mismatch often points to delayed exports, cached pages, variant logic or timezone mistakes around promotions.
Audit:
- feed update frequency
- structured data on product pages
- sale start and end times
- variant-specific URLs
- out-of-stock and preorder handling
- shipping and return information
A high-click campaign cannot compensate for arriving at an unavailable product.
Improve the Landing Page as Part of the Feed
The product page should confirm the promise made by the ad and feed. Include clear product name, current price, availability, variant selection, delivery information, returns, specifications and a usable purchase action.
Check the mobile path under realistic network conditions. Automated campaigns may find more demand, but slow or confusing checkout simply buys more abandonment.
Segment by Economics
Do not optimise all products to one ROAS target if their margins, repeat purchase rates or return costs differ materially.
Build labels for groups such as:
- margin band
- bestseller or long tail
- seasonal status
- stock depth
- new-customer priority
- clearance or protected range
Use these groups for analysis and, where campaign structure allows, decision-making. Document how each label is populated so it remains reliable.
Run a Controlled Rollout
Before enabling a major automation change, capture at least the recent baseline for spend, conversion value, contribution margin, average order value, new-customer share and product coverage.
Then:
- resolve critical feed issues
- improve a defined product group
- record the date AI Max settings change
- avoid overlapping promotions if possible
- allow for conversion lag
- compare against an appropriate control or prior pattern
Google's product announcements explain capabilities, not the result your store will achieve. Your test design supplies that evidence.
The Weekly Feed Review
Every week, review:
- new disapprovals and warnings
- price and availability mismatch
- products spending without useful outcomes
- products constrained by missing attributes
- image quality issues
- stock risk among high-spend items
- profit, returns and new-customer value by label
Feed optimisation is operational. The best catalogue is not the one rewritten once; it is the one whose source systems keep producing accurate data.
Official Sources
Frequently Asked Questions
Let's discuss your project
Need a feed and campaign audit built around profit rather than automation claims?
