All research / Research / Product data quality
Why can't AI see smart pet products clearly?
A field-level review of 11 brands' product pages
AI shopping assistants need product information to answer shoppers' questions. Alongside descriptions, specifications, identifiers, delivery details, and after-sales policies may affect how AI interprets a product.
We selected public product pages from 11 smart pet product brands, examined their structured data, and used OpenAI's published product feed specification as a reference for discussing relevant information issues.
YuSeen Research Team · · Estimated reading time: 6 minutes
AI is becoming another way for shoppers to access product information.
According to Adobe Analytics, referrals from generative AI tools to US retail websites increased by approximately 693% year over year during the 2025 holiday shopping season.
The trend indicates that AI is already playing a role in product discovery and comparison. Source: Adobe's 2025 holiday shopping report.
For brands and merchants, this raises a practical question:
When shoppers ask AI about specifications, suitability, delivery, or after-sales terms, what reliable information is available to answer them?
Accurate, complete product data helps AI interpret products
Shoppers form impressions from images, page design, and marketing copy.
AI systems extract and organize information from sources they can access, including product pages, structured markup, platform data, and other public materials.
OpenAI publishes a specification for merchant product feeds. It covers basic information such as item identifiers, titles, descriptions, links, brands, sellers, images, availability, and prices.
The specification also covers additional information such as product attributes, GTINs, shipping, and return policies. Whether a field is required, conditionally required, or optional depends on the applicable specification and use case.
This specification applies to a particular product feed mechanism. It is not a universal structured data standard that every website must implement.
On-page structured markup and the product data a merchant submits to a platform should be checked separately.
For cross-border e-commerce businesses, the question is not simply whether a field has been filled in. It is whether important information can be retrieved accurately through suitable channels and remains consistent across sources.
Reference: OpenAI's product data specification.
Public product data on 11 smart pet product pages
The sample covered products such as automatic feeders, pet water fountains, and smart litter boxes.
These products commonly involve detailed specifications, including capacity, dimensions, power supply, suitable pets, and connectivity. Shoppers may ask specific questions before buying.
We selected one product page from each of 11 brands' official websites and examined product-related structured information in the public pages.
Checks and resultsPages containing the relevant markup
- GTIN product identifier3 / 11
- Product-level shipping information1 / 11
- Return policy markup3 / 11
- Rating markup5 / 11
Other observations
The review also identified information quality issues, including:
- A zero weight recorded for some product variants.
- Missing SKU identifiers.
- Very short descriptions, with key specifications provided mainly in images.
- Chinese punctuation appearing in some English content.
- Inconsistent basic brand information across pages.
These issues do not necessarily prevent AI recommendations, but may make information harder to identify, connect, and verify.
Research limitations
The sample was limited to one public product page per brand.
We did not examine the brands' complete submissions to Google Merchant Center, other marketplaces, or dedicated product data interfaces.
The findings therefore describe information issues observed in this sample. They cannot establish a brand's overall data management quality or its actual performance in AI recommendations.
Specific shopper questions need specific supporting information
A traditional keyword search may begin with a category name. Questions asked during AI-assisted shopping often include more specific needs.
For example:
“I have two cats and need to control their portions separately. Which automatic feeders would work?”
“Will the feeder keep working during a power outage?”
“Who pays for return shipping if the product is not suitable?”
Accurate answers require clear information about features, technical specifications, and after-sales policies.
Broad marketing terms such as “high quality,” “smart,” or “best” do not provide that information.
If a feature is not documented on the product page, AI may be unable to confirm whether the product meets the shopper's needs.
Conflicting sources can also lead to inaccurate descriptions.
These discrepancies matter because shoppers may form purchase expectations based on an AI answer.
When product information and the actual experience differ, the result may be additional support requests, returns, or loss of trust.
Start with five checks to improve product information
01Check access to public information
Confirm that product pages, important policy information, and applicable data interfaces are accessible.
For an independent website, review relevant crawler rules and security settings for unintended access restrictions.
02Clarify delivery and after-sales policies
State shipping costs, delivery coverage, estimated delivery times, return windows, return costs, and warranty terms clearly.
Keep this information consistent across pages and applicable data channels.
03Complete key specifications
Prioritize the specifications, features, and suitability criteria shoppers use most often to compare products in the category.
Important specifications should be available as readable text or data fields, as well as in images where appropriate.
04Check identifiers and variant data
Verify SKUs, model numbers, GTINs, and variant attributes.
Where a GTIN is required, use a legitimately assigned identifier that matches the actual product. Do not generate arbitrary or false codes.
05Review descriptions against real customer questions
Check that the page adequately explains who the product is for, how it can be used, its main limitations, and the relevant after-sales terms.
Descriptions should help shoppers make decisions, with specific information rather than a collection of marketing adjectives.
Product information can be improved. AI recommendations cannot be guaranteed.
Generative AI recommendations depend on many factors, including the shopper's question, model capabilities, available data, platform mechanisms, and changes over time.
No single test establishes that a product will continue to be recommended in the future.
A more dependable approach is to define repeatable test conditions, identify specific information issues, make improvements, and observe what changes.
This is the basis of YuSeen's GEO assessment and optimization services.
Our service focuses on verifiable information quality, observed test results, and ongoing improvement. It does not promise guaranteed recommendations.
