# See when ChatGPT recommends your store, and who it names instead

> Convrail tracks the buying questions your shoppers ask. Every week it asks them to an OpenAI model with web search, three times each, and records whether your store is named, at what position, and which shops are named instead. Once ten measurements exist, you see how often you are recommended, country by country.

Canonical: https://convrail.com/ai-visibility/

- Up to 25 buying questions
- Checked every week, per country
- No scraping: OpenAI’s own API

## What changes for your store

### Know where you stand

For each question a shopper would ask, how often ChatGPT-style answers name your store, and your best position.

### Know who you are up against

The shops named instead of you, and how many times, so you see the real competition in AI answers.

### Know if your work pays off

A better feed, clearer product pages, more reviews: week after week, you see whether your store is named more often.

## Everything visibility tracking needs, included

- **Questions in the shopper’s words**: Track up to 25 buying questions, written the way a buyer speaks, not as keywords.
- **Per buyer country**: Each question is checked for a country, by default your store’s, because answers differ from one market to another.
- **Weekly checks, repeated**: Every question is asked three times per weekly check, because a model’s answer varies from one call to the next.
- **How often you are named**: A percentage, shown only from ten measurements: below that, the app says it does not know yet.
- **Your best position**: Where your store appears when it is named, so you can tell a mention from a first recommendation.
- **Who is named instead**: The shops the answers recommend, ranked by how often they appear.
- **Compliant by design**: Measured through OpenAI’s API with web search. Convrail never scrapes ChatGPT.

## Live in three steps

1. **Choose the Scale plan**: Visibility tracking is included in Scale, with a 14-day free trial.
2. **Add your buying questions**: Write the questions your shoppers really ask, and pick the buyer country for each one.
3. **Read the weekly results**: After the first check, see who is named. After four weeks, see how often you are.

## How the measurement works

What is measured, what is not, and why the app refuses to show a percentage it cannot stand behind.

## What is measured, exactly

Every week, Convrail asks each tracked question to an OpenAI model with web search, through OpenAI’s API, three times. For each answer it records which shops are recommended, whether your store is one of them, and at what position.

*What the Visibility screen shows for each tracked question*

| You see | What it means | When it appears |
| --- | --- | --- |
| You are named | The share of answers that name your store | From ten measurements, about four weekly checks |
| Best position | Your highest place when you are named | From the first check |
| Who gets named instead | The other shops recommended, and how many times | From the first check |
| Last checked | The date of the latest weekly check | From the first check |

## Why there is no percentage in the first weeks

A model does not give the same answer twice. On a yes or no measurement, ten samples still leave a wide margin of error, and three would suggest a precision nobody has. So below ten measurements the app shows “Not enough measurements yet” instead of a number you might act on.

## What this does not measure

- The answers ChatGPT gives a specific person. Those can depend on their history, settings and location; the measurement uses OpenAI’s API with web search, an approximation of what a shopper sees.
- Your place in the ChatGPT product carousel. That comes from your product feed: see how OpenAI receives it on the [product feed page](/product-feed/).
- Ads. Sponsored placements are bought and reported in ChatGPT Ads Manager; [attribution](/attribution/) measures the orders they bring.

## Frequently asked questions

### Does Convrail scrape ChatGPT?

No. The questions are asked to an OpenAI model with web search through OpenAI’s API. Scraping ChatGPT would break OpenAI’s terms of use.

### Why do I need four weeks before seeing a percentage?

Each weekly check asks every question three times, and a percentage is shown only from ten measurements. Before that, a figure would look precise without being reliable.

### Which plan includes visibility tracking?

Scale, which tracks up to 25 buying questions and comes with a 14-day free trial. The other plans show what the screen does and how to unlock it.

### How do I get named more often?

Start with the data AI assistants rely on: a valid product feed, complete product pages, identifiers, reviews and clear policies. The guide on getting products recommended by ChatGPT goes through each lever.