How to Measure AI Visibility for E-commerce Products and Categories

We checked the shopping documentation from OpenAI, Google, Perplexity and Shopify, and ran our own shopping prompts through ChatGPT and Claude, on 14 October 2026.

How do you measure AI visibility for e-commerce products and categories?

Measure AI visibility per category and per product, alongside the brand-level figure. A brand-level figure says how often AI names your brand. An e-commerce team needs to know which categories and products AI recommends, and where it sends the buyer to check out.

Shoppers ask for a moisturizer or a pet-hair vacuum, and the answer names products. A brand can lead in serums and be absent from sunscreen, and one brand-wide figure blends the two.

ChatGPT's help pages explain how it picks products. The factors it considers include structured metadata such as price and description from first-party and third-party providers, plus other third-party content. When more than one merchant sells an item, it ranks them on factors like availability, price, quality and whether the merchant is the maker or primary seller.

Your product can win the recommendation while another merchant wins the click. Results can be personal too: ChatGPT considers the query and context such as Memory when picking carousel products, and Perplexity uses past searches, interactions and saved preferences to refine recommendations over time. A screenshot from your own account proves little.

For how engines pick products, read how AI shopping agents decide which products to recommend.

Which shopping prompts should you track for products and categories?

Build the prompt set from your category tree and best sellers, written the way shoppers type. Cover these types for every category you sell in:

• Category prompts. best trail running shoes for wide feet.
• Named-product prompts. is the Shark Stratos good for long pet hair.
• Comparison prompts. Brooks Adrenaline or ASICS Kayano for flat feet.
• Price-capped prompts. best cordless vacuum on a budget.
• Gift and replacement prompts. what should replace my old Dyson.

Shoppers word the same need in many different ways. In SparkToro's research, human-written prompts with the same intent, buying headphones, had an average semantic similarity of only 0.081. Track a handful of phrasings per intent.

How do you track product visibility in ChatGPT, Google AI, Perplexity and Shopify?

Track each surface separately: each one reads different product data and reports results in a different place.

Google made AI performance insights in Merchant Center generally available in September 2026 in Australia, Canada, India, New Zealand and the US.

What metrics measure AI share of voice for an e-commerce brand?

Start with category share of voice, then add product, accuracy and revenue metrics, each read by category.

Read median position as a trend across runs (see the run-count section).

How do you check that AI describes your products correctly?

Once a month, run prompts for your top products and compare each price, stock status and claim with your product page and feed:

1. Compare the quoted price. OpenAI says price and shipping changes can take time to show in ChatGPT, and its first shopping response usually shows the first merchant's price, which may not be the lowest.
2. Check stock and discounts. OpenAI warns that in shopping research, prices, stock and discounts may not always match the retailer's page. Its product feed spec says to update availability when an item becomes available or sells out.
3. Read the feed's Upload History. It reports missing required columns and invalid values, and a malformed row can be rejected while valid rows keep processing.
4. Let AI reach your pages. When a retailer blocks automated access, ChatGPT shopping research skips it or relies on other sites with similar products.
5. Serve product markup in the HTML. Google warns that JavaScript-generated Product markup can make Shopping crawls less frequent and less reliable, which can be an issue for fast-changing price and availability.
6. Confirm markdowns reach agents. Shopify's changelog says it now shares compare-at prices with AI shopping agents through Shopify Catalog; before, agents reading Catalog saw only the current price.

For the work that gets products recommended, see how DTC brands win customers from AI search.

How many times should you run each shopping prompt?

SparkToro advises running each prompt usually at least 60 to 100 times and averaging the results. Report the share of runs that name you. SparkToro's test had 600 volunteers put 12 prompts to ChatGPT, Claude and Google's AI, for 2,961 runs in all.

SparkToro found less than a 1 in 100 chance that ChatGPT or Google's AI returns the same brand list in any two of 100 responses, and roughly 1 in 1,000 runs before two lists match in order.

Aggregates still hold. Across 994 responses to the differently worded headphone prompts, Bose, Sony, Sennheiser and Apple appeared 55% to 77% of the time. Sample sizes and scoring rules are in our guide to running a brand AI visibility measurement.

How do you find AI-referred sessions in GA4, Search Console and Shopify?

Use GA4, Search Console and Shopify together, since none sees everything.

• GA4. OpenAI's publisher FAQ says ChatGPT adds a UTM source parameter set to chatgpt.com to referral URLs. Put AI referrers into one custom channel group, split by landing page and category.
• Search Console. Google says a site's appearances in AI Overviews and AI Mode count toward its overall Search Console traffic, inside the Performance report's Web search type. Its generative AI performance reports show impressions, pages, countries and devices, for all websites worldwide as of 31 August 2026, by hour, day, week or month.
• Shopify. Read Meta's results in agentic settings (see above) against orders by category.

OnlyAEO's AI crawler and referral tracking captures AI-referred visits that GA4 misses and logs which pages AI bots read. Our guide to tracking AI referral traffic in GA4 has the setup.

How do you measure revenue per AI session?

Divide revenue from AI-referred sessions by the number of AI-referred sessions for each category, and compare it with that category's site average. Start in the categories where AI visits are worth more.

Adobe's August 2026 AI traffic trends report found AI-referred retail visits earned 53% more revenue per visit than non-AI traffic in July 2026, versus a year earlier when non-AI visits were worth 128% more; its September 2026 holiday forecast put the lead at 43% in August.

That forecast also found AI visitors converted at a rate 48% higher than non-AI traffic in August 2026. Adobe's figures are for US retail as a whole, so test them on your own categories with our post on the conversion rate of AI-referred traffic.

How do you count assisted conversions from AI?

Count assisted conversions in GA4's attribution paths report, which shows the channels that initiate, assist and close key events such as purchases, and add a post-purchase question for shoppers it cannot see. A shopper can read an AI answer, leave, and buy days later through search.

For each path, the report also gives purchase revenue, days to key event and touchpoints. Google suggests watching conversions and time on site in tools such as Google Analytics, and reports that, in its experience, clicks from results pages that carry AI Overviews are higher quality.

Word the question where did you hear about us, with AI assistants as options, and treat answers as a direction signal. For buyers with no referrer, see how to prove AEO drove sales with no referrer.

What did measuring a real retailer's AI visibility show?

Measuring by buyer question and cited source showed a New York edibles retailer where it was missing from AI answers. The audit captured the questions AI models were being asked in the category, across every buyer type, and each became a content target structured so a model could lift a direct answer.

The audit also showed which sources AI cited most, and the program was engineered so the client's own website, not a directory or review site, became the page AI models cite.

The retailer's AI visibility rose from 1.5% (#50) to 25%, placing it #4 of 374 brands. Its own website ended up as the source AI cited most in that category, ranking above the official New York State site. The New York edibles retailer results have the detail.

How we tested shopping prompts in ChatGPT and Claude

We asked ChatGPT and Claude Sonnet, with web search on, about a price-capped moisturizer for sensitive skin, running shoes for flat feet, and Dyson or Shark for pet hair, and logged each engine's searches and cited domains.

• ChatGPT fanned out. It ran many searches: product-by-name price checks at Target, Walmart and Ulta, model-year and version checks for stability shoes, and dermatology and eczema association guidance. It cited retailer sites, brand sites such as Brooks, and health or review sites.
• Claude ran a single search per prompt. It searched the question nearly as written and cited editorial, review and retailer sites such as Healthline, Fleet Feet and Walmart.
• Both leaned on review sites for the vacuum. Each cited RTINGS, and ChatGPT added Good Housekeeping.

Log cited domains per category and engine, because they differ. Check retailer pages as well as your own: a wrong price on a Walmart listing can reach the shopper even when your site is right.

How often should you measure, and when should you expect movement?

Report monthly and watch products weekly. OnlyAEO's monthly report tracks mention rate, LLM recall frequency and citation share. Commerce shows gainers and losers week to week, before the change shows up in revenue.

Keep the prompt set fixed, or the trend line ends up comparing different questions. In our client work, the first measurable results usually need at least 60 days. We give no guarantee of results, and we make no promise of specific rankings or citations.

How do you choose a provider to measure AI shopping visibility?

Choose a provider that scores by category and product, reruns prompts and shows raw answers. Ask:

• Does it score by SKU? The first section shows why a brand-wide score falls short.
• Does it separate recommendations from mentions? A product named seventh may not win the shopper.
• Does it show cited domains and merchants? Retailer links belong in the report.
• Can you reproduce answers? Ask for each exact prompt and answer.

Google's AI optimization guide says: "No third-party tool has access to our internal ranking or AI systems." Every provider, us included, measures from outside.

OnlyAEO is an AEO and GEO agency. Our Commerce product for AI shopping visibility runs a fixed set of category and buying-intent prompts and scores your share of the answer against top competitors, with a per-product score and category leaderboard. It tracks product visibility share, competitor benchmarks, retailer recommendations and the traits AI compares on.

The AI assistants OnlyAEO works on are ChatGPT, Gemini, Claude, Google AI Overviews, Perplexity, Microsoft Copilot, DeepSeek, Grok, Meta AI (Muse) and Instinct. Its AI visibility platform can run ChatGPT, Claude, Gemini, DeepSeek and Grok.

More questions about measuring AI visibility for e-commerce

Is shopping in ChatGPT available outside the US?

OpenAI says shopping in ChatGPT is currently live for users in the US. Outside the US, check which other surfaces in this guide are live in your market, and note that Merchant Center's AI performance insights cover only some countries (see the surfaces section).

Should you measure more often during the holiday season?

Adobe's September 2026 holiday forecast says AI-driven traffic to retail sites grew 127% year over year in August 2026, and it expects 130% growth over the 2026 holiday season and 141% on Thanksgiving. On those figures, start weekly runs on gift and deal prompts in October.

Should your accuracy checks include Merchant Center attributes?

Yes. In Google's testing with lululemon, conversational attributes the brand submitted in Merchant Center were used 50% of the time in relevant AI Mode product recommendations. Check that the attributes AI repeats match what you submitted.