What's Actually Happening When an AI Shopping Assistant Evaluates Your Product?

AI shopping assistants aren't just searching for keywords and returning links. They're running a structured evaluation before making any recommendation. Understanding that evaluation process is how you optimize for it. Here are the 8 checks that happen before your product gets recommended — or passed over.

By Steve Merrill | June 29, 2026

The 8 Pre-Recommendation Checks

1. Availability Status

First check, every time. If the product is showing as unavailable, out of stock, or has no clear availability signal — it's eliminated immediately. AI assistants don't recommend products users can't buy. Availability must be current (not stale feed data), clearly expressed in schema (InStock, OutOfStock, LimitedAvailability), and consistent between your schema and your product page. Schema.org ItemAvailability values define the accepted vocabulary.

2. Price and Total Cost Transparency

AI assistants check whether they can determine total purchase cost. Price in schema, currency specified, and estimated shipping available. Platforms like Perplexity Comet Commerce specifically filter for total cost transparency — merchants where shipping is hidden or final price is unclear are ranked below those where it's explicit. Show your price, your currency, and your estimated shipping in structured format.

3. Return Policy Access

Return policy is a trust signal in the pre-recommendation evaluation. AI assistants check whether a return policy exists in structured format (MerchantReturnPolicy schema) and whether it meets minimum thresholds for their users. A store with no structured return policy has a lower trust score than one with a clearly defined 30-day policy — even if the actual policies are similar. The structured format is what the check requires, not just the policy text.

4. Merchant Verification Status

Verified merchants get higher base trust scores on every platform. ChatGPT Commerce, Perplexity, Google AI Mode, and Microsoft Copilot all have merchant verification processes. Unverified merchants aren't excluded entirely, but they're competing at a trust disadvantage on every query. The check happens before relevance scoring — trust gates the door, relevance determines position.

5. Product Specification Match to Query

Does your product data contain the attributes needed to match the specific query? "Waterproof hiking boots for wide feet under $150" requires: waterproof attribute, boot category, width availability, price. AI assistants evaluate match quality across all specified attributes simultaneously. A partial match (waterproof boots at the right price but no width information) often scores below a complete match at a slightly different price point. Complete attribute data beats partial match on price.

6. Review Quality and Volume

Not just whether reviews exist — AI assistants evaluate review quality signals. AggregateRating schema is the primary input. A 4.7-star product with 1,200 reviews scores significantly higher on this check than a 4.9-star product with 12 reviews. Volume and recency both matter. If you have strong review data, make sure it's in structured format (AggregateRating schema). If you don't, this is a visible gap in your pre-recommendation evaluation.

7. Shipping Timeline

This is the newest check added to most major AI shopping platforms in 2026. ShippingDeliveryTime schema lets AI assistants answer "2-day delivery" queries accurately. Without it, your products are excluded from shipping-time-filtered queries — which are a growing percentage of purchase-intent searches. Add ShippingDeliveryTime schema combined with OfferShippingDetails to your product pages.

8. Data Consistency Across Sources

AI assistants cross-reference your product data across multiple sources: your product page, your schema markup, your Merchant Center feed, and any third-party product databases that have your GTINs. Inconsistencies — a price in your schema that doesn't match your Merchant Center feed, an availability status that conflicts between sources — trigger reliability flags. Consistent data across all sources is what builds the trust signal that keeps your products in the recommendation pool.

The Evaluation Is Simultaneous

These 8 checks don't happen in sequence — they happen simultaneously. A perfect score on checks 2-8 doesn't help if check 1 (availability) fails. Get the fundamentals right before optimizing the finer points. Availability, price transparency, and return policy are the table stakes. Everything else amplifies from there.

Frequently Asked Questions

Can I see which checks my products pass or fail?

Not in a single dashboard. Google Merchant Center shows product data errors that affect checks 1, 2, 5, and 8. Rich Results Test validates schema for checks 3, 4, 6, and 7. Manual AI shopping tests on ChatGPT and Perplexity show you the end result of all 8 checks combined.

Which check is hardest to fix?

Check 8 (data consistency) because it requires maintaining alignment across multiple systems: your Shopify store, Merchant Center feed, schema markup, and any third-party integrations. Most merchants have at least one inconsistency that creates a reliability flag they're not aware of.

Does my product need to pass all 8 checks to get recommended?

Not necessarily — products with strong scores on 6 of 8 checks can still appear in AI recommendations. But products that pass all 8 consistently appear in more recommendation contexts, including the high-intent purchase-ready queries where conversion rates are highest.

Optimize for the Evaluation, Not Just the Outcome

Most merchants try to "get into AI recommendations" without understanding what AI platforms are actually evaluating. Work through these 8 checks systematically against your top 20 products. The gaps are usually in checks 3 (return policy schema), 7 (shipping schema), and 8 (data consistency) — the ones most merchants haven't addressed yet. Get your store evaluated against all 8 AI pre-recommendation checks →