Can you track AI visibility for individual products or SKUs?

Product-level tracking means writing prompts about a specific product, running them on each AI engine, and recording whether the product is named, linked, and described correctly. A prompt like "best waterproof trail shoe under $150" tests a category. A prompt like "is the [product name] waterproof" tests a single SKU.

No engine hands you this data per product. Google's AI report in Search Console exposes impressions only, with no clicks and no position. ChatGPT, Claude and Perplexity publish no per-product report at all. So every SKU-level number you see comes from a tool asking questions and reading the answers.

That has two consequences. Cost scales with the number of products you track. And the method only covers the products you wrote prompts for. The rest of your catalog needs a different kind of check, which is where crawl checks come in. For the wider picture of how this works for online stores, see AI visibility for ecommerce and DTC brands.

The arithmetic is simple and it adds up fast. Say you run a store with 200 SKUs. You write 3 prompts per product and run them on 5 engines. That is 3,000 answers per run, before you repeat runs to smooth out answer-to-answer variation.

Most tools price by tracked prompts or keywords, so the bill follows that multiplication. For a small business, tracking every SKU is rarely worth it. A better split:

• Track categories ("best espresso grinder for beginners") to see which brands engines name.
• Track hero SKUs: the 10 to 20 products that drive revenue or get asked about by name.
• Crawl-check everything else, because one product template usually serves hundreds of pages.

If you want a repeatable way to write and run those prompts, the method in ChatGPT prompt tracking applies to product questions too.

A tracked prompt tells you what one engine said to one question at one moment. It's useful, and it's narrower than it looks.

• The real query is hidden. OpenAI says ChatGPT search rewrites your question into one or more targeted queries. Google may fan a query out across subtopics. Your prompt is a proxy for an input you cannot observe.
• Retrieval is discretionary. As of October 2026, ChatGPT decides per query whether to search, and Claude.ai search is a user toggle. An answer given without search comes from training data with a cutoff, so a price in it can be months old.
• There is no rank. Google exposes no position data for AI answers, so a "product rank" has nothing to be checked against. Why there's no AI rank to track covers this.
• Engines differ. In Otterly's 2,000-post experiment, Microsoft Copilot produced 69% of one site's tracked citations while ChatGPT and Google's AI surfaces barely cited it. Score each engine separately, as one dashboard for all AI engines explains.

The most useful product signal is often accuracy, not presence. If an engine names your product with the wrong price or a discontinued feature, that is a specific, fixable problem.

These checks are the part of AI visibility with the strongest evidence, and they scale to your whole catalog because product pages share templates.

Prices and availability in the raw HTML

Vercel found that ChatGPT and Claude crawlers fetch JavaScript files but do not execute them, so they cannot read client-side rendered content. If your price, stock status or variant options load through JavaScript, those crawlers may see an empty product page. Googlebot renders JavaScript, so this gap mostly affects non-Google engines. Fetch a product URL with JavaScript off and confirm the price is in the response. A stated price on a fetchable page gives an engine a sentence it can quote.

Product URLs AI crawlers can reach

robots.txt rules are per bot. OAI-SearchBot controls ChatGPT search, while GPTBot governs training only, and OpenAI says each setting is independent. A blanket AI block can remove your products from ChatGPT search without anyone noticing.

User-agent: OAI-SearchBot
Allow: /

robots.txt is not the whole gate. OpenAI's help page says the host or CDN must also allow traffic from its published search bot IP addresses. A firewall can return 403 to a crawler that robots.txt allows. The AI Crawler Checker requests a page as each AI crawler and shows which ones get turned away.

Index and snippet eligibility

Google states that a page must be indexed and eligible for a snippet to appear as a supporting link in AI Overviews or AI Mode. Check product templates for stray noindex, nosnippet, max-snippet:0 or data-nosnippet on the price block. The ecommerce product data check walks through each of these.

Product catalogs create problems that brand-level tracking never meets.

• Variants. If size or color options share one URL and switch by JavaScript, crawlers that don't execute scripts see only the default state.
• Stock changes. An engine that cached a page last month may still describe a sold-out item as available. Recheck answers for high-turnover products more often.
• Training-data answers. As of October 2026, OpenAI lists a February 16, 2026 knowledge cutoff for its GPT-5.6 models and August 31, 2025 for the Instant path. Products launched after those dates cannot appear in answers given without search.
• Platform defaults. Hosted store platforms and CDNs can change crawler rules without you editing anything. If you sell on a hosted platform, see store-level checks for Shopify.

1. Pick 10 to 20 hero SKUs and a handful of category prompts. Track those, per engine.
2. Fetch one product page per template with JavaScript off. Confirm price, availability and key specs are in the HTML.
3. Read your robots.txt bot by bot, then test whether each AI crawler actually gets a 200 response.
4. Check Google Search Console and Bing Webmaster Tools separately for index status.
5. Review tracked answers for wrong prices and outdated details, and fix the source pages that state them. For how fast that can be caught, see detecting misinformation in AI answers.

None of this guarantees a product gets named. It removes the blocks that stop engines from reading it.