How to track AI visibility for your ecommerce store

Shoppers now ask ChatGPT, Gemini, and Claude the same questions they used to type into Google: "best running shoes for flat feet," "sustainable swimwear brands," "durable backpack for weekend travel." If your store never shows up in those answers, you lose the sale before a shopper ever opens your site. Tracking AI visibility means treating those answers like a channel you measure on a schedule - not a curiosity you check once in a personal chat and forget.

Start with one scoreboard: mention rate

Mention rate is the number to check first. It is the share of tracked prompts where your store is named in the AI answer, out of everything you're tracking that week. A prompt like "recommend a backpack brand for weekend trips" either names your store or it doesn't - mention rate turns that yes/no into a trend line you can watch move over time.

One detail matters more than it looks: mention rate excludes branded prompts. If a shopper already typed your store's name into the question, of course the answer mentions you - that tells you nothing about discovery. The number worth watching is how often you get named when nobody asked for you by name.

Share of voice: who else is winning the same prompts

Mention rate tells you if you showed up. Share of voice tells you who you're up against. It's built from the other brands named on the same prompts as yours - so when three competitors get named on "best organic skincare for sensitive skin" and you don't, you see exactly who is filling the space you're missing.

This is where ongoing tracking earns its keep over a single screenshot from a chat: a screenshot shows one answer, share of voice shows a pattern across your whole prompt library, week over week, category by category.

Sources: where the model got its answer

Every answer an AI model gives is built from somewhere - a product page, a comparison article, a review site, a forum thread. Tracking sources means capturing the cited URLs behind each answer, so you can see which pages the models are actually pulling from when they talk about your category.

Sources turn a vague "the AI doesn't know us" into a concrete list: this comparison site got cited eleven times this month, your own product page zero times. That's a content gap you can act on, not a mystery you have to guess at.

Engines on one calendar

Different shoppers use different assistants, and each one draws from different sources - so tracking only one model gives you half the picture. The practical answer is to run your prompt library across several engines on the same schedule, so a change you notice in Gemini this week isn't forgotten by the time you happen to check ChatGPT next month.

What a free scan and a free account show you

A free scan gives you a single ChatGPT snapshot across 10 prompts - a useful first look at where you stand. A free account extends that to one manual ChatGPT scan a week, still 10 prompts. It's enough to see whether your store shows up at all, but not enough to catch a slow slide in mention rate or a new competitor quietly gaining ground.

What a trial unlocks

A 7-day trial runs 50 prompts across ChatGPT, Gemini, and AI Overviews (Claude isn't part of the trial). That's usually enough to see your mention rate, your share of voice, and your top cited sources across the engines that carry the most shopping traffic - a realistic taste of what weekly tracking looks like.

What each plan tracks weekly

Weekly tracking is where AI visibility becomes a habit instead of a one-off audit. Starter covers ChatGPT and AI Overviews. Shop adds Gemini on top. Business adds Claude, giving you the full set of engines shoppers actually use to ask for recommendations. Compare plans to see which engine mix matches where your shoppers already are.

Daily boost: a faster read between weekly scans

Weekly tracking is your scoreboard; daily boost is your early warning system. It runs a small set of pinned prompts through ChatGPT every day - 3 pins on Starter, 5 on Shop - as a separate series from your weekly library. Pin the prompts you care about most right now, a flagship product, a category you're pushing this quarter, and watch them move daily instead of waiting for the weekly refresh.

Prompt library quality: cold prompts plus your own

A tracking tool is only as good as the questions it asks. Cold shopping prompts are generated straight from your catalog and menu structure, so a store selling hiking boots gets prompts like "best hiking boots for wide feet" - pulled from what you actually sell, not a generic template that fits every store the same way.

Cold prompts give you fair, unbiased coverage of your catalog from day one. But you know your business better than any generator does - a seasonal push, a new product line, a phrase your customers actually use in reviews. You can add your own custom prompts alongside the cold ones, so the library reflects both what a typical shopper asks and what you specifically want to watch. See how prompt tracking works.

Reading the raw answer, not just the score

A mention rate percentage tells you the trend. It doesn't tell you why an answer went the way it did, or what changed between last week and this week. That's what Answer Explorer is for: it shows the full reply the model gave, not a summary, so you can read exactly how your store was described, what it was compared against, and which detail tipped the recommendation in - or out of - your favor.

Paired with a what-changed feed, you get a running log of the answers that shifted: a prompt that added a competitor, one that dropped your store, one that picked up a new cited source overnight. That's the difference between "our score moved" and knowing exactly which prompt moved it, and why.

One caution worth repeating: don't treat a single ChatGPT tab with Memory turned on as a measurement. A personalized chat reflects your own history with that account, not what a new shopper with no history sees. Real tracking runs the same prompts under the same conditions, across engines, on a fixed schedule - that consistency is what makes the numbers comparable from one week to the next.

Weekly rhythm vs one-off audits

A one-off audit is a snapshot: useful for a pitch deck, close to useless for spotting a trend. AI answers shift as models update, as competitors publish new content, and as your own pages get crawled and cited, or quietly don't. Weekly tracking catches a mention rate that's slipping before it shows up in sales, and flags a new source getting cited before you're left wondering why a competitor suddenly appears more often than you do.

Treat it like any other channel: check the scoreboard weekly, read the raw answers when something moves, and use the daily boost pins for whatever you're actively working on right now.

What "good" looks like in 30 days

Set realistic expectations from the start. In the first 30 days, "good" is not an 80% mention rate overnight - AI answers don't move that fast, and any tool promising it is measuring something else. Good looks like three concrete wins:

• One prompt where your store gets named for the first time.
• One GEO fix you can point to - a product page rewritten, a spec added, a comparison page published - based on a source gap you found.
• One content gap URL identified and closed, so the next scan has something new to cite.

Stack a few of those months in a row and the scoreboard starts moving on its own. Start tracking your store's AI visibility or see plans and engine coverage to pick the setup that matches your catalog.