Back to blog

How Many Shoppers Did AI Send You, and Did They Land on the Products You Enriched?

Sep 23, 2026

Thin trails of light arc across a deep indigo sky and land on a shelf of plain products, most of them on the few products glowing from within, and only two on the dim ones.

Every tool that promises to make your store easier for AI to read eventually hands you a score. Scores are useful, but they answer the wrong question. What you actually want to know is simpler: did ChatGPT, Perplexity, Gemini or Copilot send anyone to your store? And when they did, did those shoppers land on the products you spent time and credits improving, or on the ones you never touched?

That second question is the one we hold ourselves to. So we rebuilt the Visibility page in CortexGuard around it. Below is what the page looks like for one real store using CortexGuard, a Shopify shop with a catalog of a little over 3,000 products. We have left its name out. Every number is exactly as the page showed it on September 23, 2026.

The top of the page: this week, in one sentence

Screenshot of two cards: a green gauge reading 87 out of 100 with five AI crawler badges, and a panel showing 44 AI sessions in the last seven days with the sentence about enriched products.
The first thing you see: AI shoppers this week, and how many landed on a product you had already enriched.

Reading the screenshot from left to right:

  • The gauge on the left is the store visibility score, 87 out of 100, and the five badges under it confirm that GPTBot, ClaudeBot, PerplexityBot, Google-Extended and Bingbot can all reach the store. It is a check that the door is open, not a measure of who walked through it.
  • The big number is who walked through it: 44 AI sessions in the last seven days. A session is a visit whose first page came from an AI assistant's answer.
  • Under it, two comparisons: up 7% on the week before, and up 340% on the first week we measured. At this store's size a single week swings, so the first-week comparison is the one to trust.
  • Then the sentence the whole page exists for: 285 shoppers from AI assistants since counting started on July 21, and 186 of the 252 who landed on a product landed on one that had already been enriched.

The split, week by week

One week is a snapshot. The chart underneath shows the same split over time, so you can see whether it moves when you enrich.

Screenshot of a weekly bar chart from July 20 to September 21, each bar split into dark indigo and teal, with dashed lines marking four enrichment runs and a catalog enriched percentage under each week.
Indigo is AI shoppers who landed on an enriched product, teal is the rest. The row underneath is how much of the catalog was enriched that week.
  • Each bar is one week of AI sessions that landed on a product page.
  • The indigo part landed on a product that was already enriched on the day of the visit. The teal part landed on a product that was not enriched yet.
  • The dashed lines are enrichment runs, labelled with how many products each one touched.
  • The row under the dates is the share of the catalog that was enriched that week, so you can compare where shoppers landed with what was there to land on.

Here is what this store's chart says. From late July to the end of August, only 19% of its catalog was enriched. Over those seven weeks, 121 of the 172 AI shoppers who landed on a product, about 70%, landed on that 19%. Then in mid-September the store enriched almost everything, 3,041 products in one run, the catalog went to 97% enriched, and the teal part of the bars nearly disappears.

What that proves, and what it doesn't

Under the chart, the page turns the split into one line: enriched products drew 6.3 times more AI sessions per product than the ones not yet enriched, 19 against 3 a week per 1,000 products. It is a striking number, and the page says straight away why you should not read it as cause and effect. A store usually enriches its most important products first, and those were likely to get more visits anyway. Two sets of products compared side by side always carry some of that.

We would rather tell you that than let a big multiplier do our selling. It is also exactly why the last part of the page exists.

Each run, judged only on the products it changed

Screenshot of a list of four enrichment runs with dates and product counts, each with a verdict: two marked no before to compare, one no measurable change, and one measuring, day 8 of 14.
Every run gets its own verdict, measured on its own products over 14 days before and 14 days after.

A fairer test than comparing two sets of products is comparing the same products with themselves. For every enrichment run, the page counts AI sessions on exactly the products that run touched, in the 14 days before and the 14 days after. It never uses store-wide traffic for this, because a run that changed 47 products was never going to move a whole store, and pretending it did would flatter us.

  • No before to compare: the run happened before counting started, so there is nothing to measure against. This store's first two runs, in July, are like that.
  • No measurable change: there were visits on both sides, but too few to call. The August run touched 47 products and saw 1 AI session before and 3 after. That is honest, and it is not a result.
  • Measuring, day 8 of 14: the September run of 3,041 products is still inside its window. Its verdict arrives on its own once 14 days have passed.
  • Measured: once a run has enough visits on both sides, the page states the change in plain numbers. The sentences are written from the numbers every time, never by an AI.

Which assistants are doing the sending

Further down, the page breaks sessions out by assistant and shows where each one landed. For this store the answer is almost entirely ChatGPT: 280 of the 285 sessions since July 21. Perplexity has sent one. Gemini sent its first four this week. Copilot has not sent any yet. Of all 285 visits, 252 landed straight on a product page and 33 on the homepage, a collection or a search.

Seeing it for your own store

The Visibility page is part of CortexGuard's Starter plan and up. On the Free plan the counting still runs quietly in the background, so your history is already there the day you upgrade. A newer store starts small and fills in week by week, and the page says so rather than guessing. If you would like to know first whether AI crawlers can read your store at all, start with the free scan: cortexcommerce.ai/scan.