Track four numbers for your shopping questions, not one score: whether your brand is named, whether it's recommended rather than just listed, whether a specific product URL is cited, and whether AI sends the shopper to that product page instead of a category page or a competitor. Most ecommerce teams only check the first one.
This is the ecommerce layer on top of getting recommended by ChatGPT, Claude and Perplexity: the same groundwork applies, but a store has to measure it at the product level, not just the brand level.
It means four different outcomes for the same shopping question, and a store can win one while losing the other three. Named means your brand appears somewhere in the generated answer. Recommended means the answer tells the buyer to pick you specifically, not just lists you alongside two competitors. Cited means a specific URL from your site is shown as a source. Right-page means that URL is the product or category page actually built to answer the question, not your homepage or an old blog post.
Why it matters: a store can be "named" in half its shopping answers and still send every one of those shoppers to a competitor's product, if the citation and the recommendation both go elsewhere.
Named rate, recommendation rate, citation accuracy and competitor substitution rate, each measured against the shopping questions your own buyers actually ask. The table below is the order to build them in.
| Metric | What it shows | Where to see it |
|---|---|---|
| Named rate | How often your brand appears at all in answers to your shopping questions | Ask ChatGPT, Claude and Perplexity directly |
| Recommendation rate | How often you're named as the pick, not just listed alongside others | Read the raw answer text, not a single score |
| Citation accuracy | Whether the URL cited is the specific product page, the category page, or nothing of yours | What AI actually sees on the cited page vs. the page that should answer |
| Competitor substitution rate | Which competitor, and which of their products, gets named instead of yours | The same questions asked by hand, three times each |
Do this first: write down the 10–15 shopping questions your own buyers ask most ("best X for Y", "X vs Y", "where to buy X"), and run all four metrics against that same list every week, not a fresh list each time.
Ask the exact shopping question in each engine, note the URL it cites, then compare that against the page you'd actually want a buyer to land on. If the engine cites your category page for a question that should be answered by one specific product, that is a citation-accuracy miss even though the brand got named.
Shortcut: run the free What AI actually sees check on both pages. It shows the raw HTML a fetcher reads, which is usually where the answer becomes obvious.
Usually because the category page states the comparison in plain, self-contained text, while the product page's price, availability or variant selector is rendered by JavaScript the fetcher never runs. A category page that already reads like "the best running shoe for flat feet is…" is easy to quote. A product page where the price only appears after a script loads is not, regardless of how good the product is.
See AI can't read my site. Is it the JavaScript? and why AI gets my pricing wrong for the fix on each.
Fix: make sure price, stock and the key spec sit in the page's raw HTML, and give each product page one self-contained sentence that answers the comparison question the category page currently wins.
Ask the same shopping questions with a competitor's name in place of yours, and log which of their specific product URLs gets cited. Shoppers increasingly research this way: 77% of US consumers say they used AI to help with a shopping decision in the past six months, and 64% of US online shoppers now use at least one AI-assisted shopping tool (Exploding Topics, 1,009 US consumers, 2026; Ryder 2026 E-commerce Consumer Study, 1,160 US online shoppers, March 2026), so a shopping question answered by a competitor's product is a buyer you never see in any funnel.
Do this: treat the competitor's cited product page as the benchmark for what yours needs to match, not just a line in a report.
None of this requires buying anything, and each step takes a few minutes:
Bottom line: a store that tracks only "are we named" is reading one quarter of the picture; the other three quarters are where the shopper actually ends up.
There isn't one. Named, recommended, cited and right-page each fail for a different reason, so a store that tracks only one number, usually "named", can look fine while still losing every shopper AI sends to a competitor's product instead.
Weekly. Engine answers vary run to run and change after every catalog update, price change or new review, so a single monthly snapshot misses the swings that matter for a live store.
No. An AI answer that names your product without a click-through leaves no session in Analytics, so most of what a shopper sees from ChatGPT, Claude or Perplexity never appears in your existing reporting at all.
Yes. The same shopping question that fails to name you usually names someone else, and seeing which competitor and which of their products wins that slot is what tells you what to fix first.
Ask ChatGPT, Claude and Perplexity your own shopping questions and write down the four counts yourself. The free check at dashboard.reflexaagent.com/audit covers the other half: whether the pages those answers would have to quote can be reached, read and attributed at all.
The free check reads your product pages the way ChatGPT, Claude and Perplexity read them, and shows the exact URL each one cites. 3 minutes, evidence included.
Let the engines in, with one click: the robots.txt switch for AI crawlers · install Reflexa Agent · på svenska