Intelligence Artificielle
AI visibility: how a small business can tell if ChatGPT recommends it
To find out whether ChatGPT, Gemini, or Perplexity recommend your business, you need to ask them your customers' questions on a regular basis and write down what they answer, then cross-check that log with the visits AI crawlers make to your site, as recorded in your server logs. Your usual analytics tool won't tell you on its own, because much of this activity leaves no trace in a standard dashboard. A small business can set up this tracking in half a day with a spreadsheet and a few settings.
Why can't your dashboard see AI?
Your analytics count human visits, while AI often acts before the visit, or instead of it. When an assistant compares three stores for a shopper, it reads your pages without a single session showing up in your reports.
This is not a fringe behavior. According to an Odoxa survey for Fevad, the French e-commerce federation, published on February 11, 2026, 31% of French online shoppers already use generative AI when they shop online, mostly before buying, to compare products or make a first selection. The consulting firm Converteo, in a study published on January 27, 2026, reports that traffic generated by large language models in retail grew more than a hundredfold in less than a year. Yet the crawlers that read your pages to answer a shopper usually don't show up in your analytics tool, and your server logs are where you'll find them.
If you judge your site on visit counts alone, you only see part of what assistants do with your pages.
How do you know whether ChatGPT mentions you, and gets it right?
The simplest approach is to build a panel of about twenty questions and run it through the main assistants every month. It's a manual check, and it becomes reliable once you repeat it.
Use the questions your customers actually ask by phone, by email, or in store, such as "which product should I pick for…", "where can I buy… near…", or "which French brand makes…". Ask them in ChatGPT, Gemini, Perplexity, and Claude, and for each one record three things: whether your business is mentioned, whether the information is accurate (prices, delivery times, opening hours, return terms), and which sources the assistant cites.
Ask each question several times, because answers vary from one session to the next. The software company Custplace did this at scale. According to an interview published by Visionary Marketing on July 8, 2026, it compared up-to-date information for 500 locations with what ChatGPT, Claude, and Perplexity returned, repeating the test ten times over three months. It found a 41% gap, mostly on opening hours. Its founder, Nicolas Marette, explains that these tools "look for what they were able to retrieve earlier" rather than the most recent data.
Your log should therefore check the accuracy of the answers as closely as your presence. A wrong delivery time or opening hour can cost you a sale, and you'll never know unless you ask.
Which metrics should you track alongside traffic?
Four metrics are enough to start, and most of them require no paid tool.
| Metric | What it measures | Where to find it |
|---|---|---|
| Presence rate | The share of panel questions where you are mentioned | Your monthly log |
| Cited sources | The pages (yours or other sites') the AI relies on | The links shown in the answers |
| AI crawler visits | The number of pages read by assistants' crawlers | Your server or hosting logs |
| AI-referred visits and sales | Sessions sent by chatgpt.com, perplexity.ai, or gemini.google.com, and their conversion rate | The referral report in your analytics tool |
For crawlers, AI companies publish the names they use. OpenAI documents OAI-SearchBot, the crawler used to surface websites in ChatGPT's search results, and states that a site blocking it in robots.txt will not appear in those answers. Perplexity does the same for PerplexityBot. Checking that your robots.txt file doesn't block these crawlers takes five minutes, and it's often the first thing I look at during a diagnosis.
What makes an AI recommend you or ignore you?
Assistants recommend what they can understand without ambiguity and what looks trustworthy to them. The content of your pages, their structure, and the consistency of your reviews matter more than your ad budget.
GEO (generative engine optimization) is the work that makes your pages readable and citable by these tools. Mirakl measured the size of the job: in a press release dated April 28, 2026, the company says it analyzed 427 product pages across 35 countries. The average score was 48 out of 100, while Mirakl sets the threshold at which a product becomes visible to AI agents at 61, and fewer than 1% of pages scored above 80. According to the same analysis, 43% of pages show no reviews, ratings, or Q&A, 86% have images that AI reads poorly, and only 9% provide structured data that clearly states price, size, and availability. These figures come from a vendor that sells a fix for the problem, but they match what I see in the field.
Custplace makes a similar point about reputation. According to Nicolas Marette, AI tools cross-check several review platforms, and a business rated consistently across several sites earns more of their trust than one rated excellent on a single platform.
Should you invest in a specialized measurement tool?
For a small business, it isn't necessary right away. A manual log, server logs, and the referral report are enough for the first few months to know where you stand.
The tools are arriving fast, though. E-marketing.fr reported on October 15, 2026 that Contentsquare opened a beta on October 13, 2026 of a platform that separates human browsing from the activity of language models and agents, and reassesses their contribution to sales with multi-touch attribution. This kind of offering targets large brands first. Above all, it shows that measuring AI is becoming a routine part of the job, just as SEO tracking did fifteen years ago.
Other tools focus on what assistants say about you, such as vurto.ai, the platform I'm building. It regularly runs your questions through ChatGPT, Gemini, Claude, or Perplexity, measures the share of answers where your brand appears and your rank against competitors, and tells a passing mention apart from a real recommendation. It also shows the sources each engine cites, and therefore the sites where it pays to be present.
My advice is still to start by hand for the first few months. You'll learn which questions matter for your business, and when the log gets too heavy, you'll know what to expect from a tool.
My prescription for the coming month
- This week, write down twenty questions your customers really ask, and run them through ChatGPT, Gemini, Perplexity, and Claude. Log your presence, the accuracy of the answers, and the cited sources in a spreadsheet.
- Right after that, open your robots.txt file and check that it blocks neither OAI-SearchBot nor PerplexityBot, unless you made that choice deliberately.
- Within two weeks, fix wrong information at the source: your product pages, your shipping and returns pages, your business listing, and your profiles on review sites.
- Every month, rerun the same panel and track your presence rate, along with the visits and sales sent by assistants.
- Before paying for a tool, wait until you have three monthly logs, so you know exactly what you want to measure.
If you'd like us to look together at what AI assistants say about your business, it's one of the diagnoses I run at Nexperio.
Frequently asked questions
How can I tell if ChatGPT recommends my business?
Write about twenty questions your customers really ask, run them through ChatGPT, Gemini, Perplexity, and Claude, and record whether you're mentioned, whether the information is accurate, and which sources are cited. Ask each question several times, since answers change from one session to the next, and repeat the same log every month to track progress.
Is Google Analytics enough to measure AI traffic?
It shows visits sent by chatgpt.com or perplexity.ai in the referral report, which is useful. It doesn't see the crawlers that read your pages to answer a shopper who never clicks. These crawlers show up in your server logs under the names their publishers document, such as OAI-SearchBot for OpenAI or PerplexityBot for Perplexity.
What is GEO?
GEO, short for generative engine optimization, covers the work that makes a website readable and citable by answer engines such as ChatGPT, Gemini, or Perplexity. It relies on complete and accurate pages, structured data that clearly states price and availability, consistent reviews across several platforms, and a robots.txt file that lets these assistants' crawlers in.
Why does AI give wrong information about my business?
Assistants often rely on what they collected earlier rather than on your latest data. Custplace measured a 41% gap between up-to-date information for 500 locations and what ChatGPT, Claude, and Perplexity returned, mostly on opening hours. Start by fixing your own pages, your business listing, and your profiles on review sites.
Do I need a paid tool to track my AI visibility?
You don't need one at first. A spreadsheet for the monthly log, your hosting logs, and the referral report are enough for a small business during the first few months. Once the panel grows or you want to track competitors, a platform such as vurto.ai automates these logs and shows the sources assistants cite. Contentsquare also opened a beta offering on October 13, 2026, but it targets large brands first.
Sources
- Les Français sont de plus en plus nombreux à adopter l'IA pour acheter sur internet · Fevad (étude Odoxa), 2026-02-11
- [Étude] IA & commerce agentique : le guide pour transformer votre expérience client · Converteo, 2026-01-27
- Visibilité entreprise LLM, comment maximiser sa présence dans l'IA · Visionary Marketing, 2026-07-08
- Moins de 1 % des pages produits compatibles avec les LLM : Mirakl lance Agentic Activation · Mirakl, 2026-04-28
- SEO et commerce agentique : Contentsquare intègre l'analyse des LLM · E-marketing.fr, 2026-10-15
- Overview of OpenAI Crawlers · OpenAI, 2026-10-19
- Perplexity Crawlers · Perplexity, 2026-10-19
- Vurto.ai : mesurer et améliorer sa visibilité dans les réponses des assistants IA · Vurto.ai, 2026-10-19