Back to blog
eCommerce

How to check if AI recommends your store’s products

Use real shopping questions to check whether ChatGPT and Gemini recommend your product, name your store, and describe your offer accurately. Follow a sample audit record.

White headphones in a blue basket beside a recommended product card

Product, brand, store, and source are four different things

A customer might ask which headphones to choose for work or where to buy a specific model. AI could recommend a product from your brand in the first answer and direct the customer to another retailer in the second. Record product mentions, retailer recommendations, and links to your site separately.

What to checkVerification questionExample action
ProductDoes the answer name the right model and variant?Check the description and attributes on the product page.
BrandDoes the answer attribute the product to the right brand?Verify the brand and manufacturer names.
StoreDoes AI name your store as a retailer?Check the offer and purchase conditions for that market.
SourceDoes the answer link to your site or a third-party page?Open the source and check what it actually supports.

Some answers contain no sources. A citation alone does not establish that the store was recommended. Read the answer in full and in context.

Start with a category and real customer questions

Choose a category that matters to your business and identify what buyers need. Consider product discovery, model comparison, and purchase. Match questions to your offer, language, and market. If you do not sell that category in a given country, its absence from the answer may be expected.

  • Discovery: Which product suits a particular use case?
  • Comparison: How do these types or models differ?
  • Purchase: Where can I buy the product with delivery to this market?
  • Verification: Is the description of this variant accurate?

A fictional example: “Which wireless headphones work well for remote work and travel?” Review the models, selection criteria, stores, and sources in the answer. Check prices and shipping terms before using them in your own offer.

Compare your offer with relevant competitors

Choose companies and products that meet the same need. Record what AI says about your offer and its alternatives, including use, specifications, variant, price, or availability. Prices and inventory can change, so check a current source for the relevant date and market.

A competitor's appearance in an answer does not establish that its product is better or reveal why it was recommended. Check the cited sources, your product page, and whether the offer actually meets the question's requirements.

Check your brand’s AI visibility

See whether AI answers mention your brand and find questions worth investigating.

Get the free report

Which product details should you check?

Before writing a new article, compare the category page, the product page, and any available catalogue data. Look for discrepancies and information buyers need, such as variant details, dimensions, compatibility, price, and delivery terms. Add verified attributes only.

Structured data can help search engines understand an offer when it matches current information visible on the page. Adding JSON-LD alone does not guarantee citations or recommendations.

Google Search Central: Product structured data

Step 1. Define what the customer should find

Start with one category and record which products you actually sell in a given country. Choose a sample product page and category page. Decide what you want to find in the answers: the right model, an accurately described feature, or a recommendation for a particular retailer. For each need, select questions your offer can answer.

Fictional example. An audio retailer wants to check whether AI recommends its headphones for work in an open-plan office. The team prepares a question about headphones for calls, a question comparing two designs, and a question about a retailer that delivers to the chosen market. All three concern the same category, and each tests a different buying decision.

Step 2. Assess each answer using the same criteria

Record the question, language, market, model, and date. Mark whether the correct product appears, whether it was suggested as a solution, which retailer was named, and where the link leads. Keep a note if the answer specifies a different variant or store.

In the fictional example, AI recommends the correct headphones but sends the customer to another retailer. Record a product recommendation and no recommendation for your store. If another answer names your store without linking to it, record the mention and the missing link separately. This helps the team tell product recognition from retailer selection.

Check whether the source supports the specific claim in the answer. A category link may help customers find the offer, while facts about weight, microphones, or compatibility will usually be in the specification for the relevant variant. Record the claims you verified and those that need further checking.

Fill in an audit record for one product

This example uses fictional Model A headphones available in a store in the chosen market. The buying question is “Where can I buy Model A headphones with delivery to my country?” Use it to add one row to your spreadsheet and rerun the test after updating the offer.

Spreadsheet fieldExample entry
ContextRecord the country, local language, ChatGPT with search, and the actual test date.
Product and retailerModel A is recommended. The answer names a different retailer.
SourceThe link opens the retailer’s product page. Check the variant and delivery conditions at that URL.
Next stepOpen your page for the same variant. Check whether buyers can see current availability and the delivery estimate.
Acceptance criteriaThe correct product page displays current information for the local offer. Record the edit date and return to the original question.

If the model and retailer appear only in a product carousel, record where they were displayed. A note such as “store shown on a product card, no recommendation in the answer text” lets you compare the same part of the answer later.

If the store serves multiple markets, run separate trials for each offer. Check that links open in the right language and that a customer in that country can actually buy the product.

Step 3. Assign the finding to a specific page

If answers keep naming the wrong variant, compare the variant name, specifications, and URL on the product page and in the catalogue feeds you use. If they omit a use case, check whether the existing description explains who the product is for. When customers compare entire types of products, a buying guide or category page may be the right place to address the question.

Example editorial task. Use the manufacturer's specifications to add microphone compatibility with computers and phones to the selected headphones’ page. Record the URL and assign a reviewer. After the update, check that the information appears next to the right variant.

Step 4. Compare subsequent measurements

Save the baseline answers before making changes. After the update, repeat the same questions in the same language and market. Compare product presence, description accuracy, retailer recommendations, and sources separately. Record the date when you add AI models or change the questions.

If your product description is accurate and AI still directs shoppers elsewhere, review the buying need expressed in the question and whether your offer meets it. The team can then decide whether to clarify delivery terms, improve variant descriptions, or observe the same question set for longer.

How does Semly help monitor products?

Semly’s AI Shopping helps analyse product appearances in AI answers and compare offers. Brand monitoring lets you track questions, full answers, competitors, and available sources. Check the features, supported models, and limits of your plan before getting started.

Explore AI Shopping in Semly

Answers can change over time. A difference observed after a page update is worth investigating. On its own, it cannot establish what caused the change.

Frequently asked questions

Does a product recommendation also recommend my store?

Sometimes. Check separately whether the answer identifies the product, the brand, the retailer, and a link to your site.

Is one question enough to assess visibility?

One answer is a single observation. Compare relevant questions, models, and time periods, then review the answers in full.

Does an absent product mean the site is blocked?

The cause needs a separate check. The absence of a mention alone cannot establish an access problem.

Does Semly guarantee a recommendation?

Semly helps you monitor answers and decide what to improve. Recommendations and citations are controlled by the AI systems.

Further reading

How to measure brand visibility in AI

How to choose a GEO tool with actionable recommendations

Share article
LinkedInXFacebook

Find out if AI ChatGPTGeminiClaudeGoogle AI OverviewsGoogle AI Mode can see you

Enter your website address to get your AI visibility report.