A ChatGPT visibility tracker asks ChatGPT the questions your buyers ask and records whether each answer names your brand. Depra captures the answer a person sees in the ChatGPT web app with web search switched on, asked from inside your country once a day, and stores it with your position, the tone of the mention and the websites ChatGPT cited.
Some tools sell this as a ChatGPT rank tracker. The job is the same: show how often ChatGPT names you, where you come in its list, and which competitors it names instead. Depra runs the same questions on Gemini, Perplexity and Google AI Overviews, so you can compare engines side by side. When they disagree, why AI engines disagree about your brand covers the usual causes.
How does Depra capture ChatGPT answers?
Depra captures ChatGPT from chatgpt.com, the consumer web app, with web search switched on. In that mode ChatGPT searches the web before it answers and links to the websites it used, so the answers Depra stores come with the websites ChatGPT cited.
Each question is asked from inside the country you track, such as India, the UAE or the United Kingdom. You can pick any of 40 markets and up to 10 per project. Location is set at country level: Depra does not target cities, add personas or simulate a chat history.
Checks run once a day for every tracked question and market, and Depra keeps the full answer text. If your plan's monthly check limit is reached, scheduled checks drop to weekly instead of stopping.
| Engine | What Depra records | How often |
|---|---|---|
| ChatGPT | The answer in the chatgpt.com web app, with web search on | Daily |
| Gemini | The answer in the gemini.google.com web app | Daily |
| Perplexity | The Sonar model answer with live web search, through the API | Weekly |
| Google AI Overviews | The AI Overview on a real Google results page, and whether one appeared | Daily |
What you see for ChatGPT
Depra reads every ChatGPT answer and records the same numbers for you and for each competitor you track, from 3 to 20 per project depending on plan. The scoring rules are on how Depra measures AI visibility.
- Visibility: the percentage of ChatGPT answers to your buyers' questions that name your brand, for any date range you pick.
- Position: your place in ChatGPT's list of recommendations. Position 1 means ChatGPT named you before any other brand.
- Sentiment: warm, neutral or critical, with the exact ChatGPT sentence that mentions you.
- Cited websites: the pages ChatGPT's web search linked to, labelled as editorial, forum, marketplace, company site and more, and whether those answers named you.
- Change over time: each number compared with the previous period, last month or last year, plus a daily "what changed" page listing what moved since the previous scan day.
- Every answer stored: read the full text of what ChatGPT said in your category, filter it, and export it as CSV.
Your ChatGPT ranking against competitors
A ranking table puts you and every tracked competitor in order on visibility, sentiment and position. Depra also lists brands the engines keep naming that you do not track, and you can add each one as a competitor in one click.
Questions that name your brand are left out
Depra does not count questions that already contain your brand name, because an answer to a question that names you says little about whether new buyers find you. Suggested questions come as 5 topics with 8 buying questions each, none naming your brand. Your report shows a confidence range next to each engine, so you can see how many answers each number rests on and how far it could move by chance.
Does ChatGPT answer differently in Hinglish?
Barely. ChatGPT gives a slightly different brand list each time you ask the same question, so Depra's study compared the change between languages with that normal rerun change. Hinglish means Hindi words typed in English letters and mixed with English, the way most Indians ask ('sabse accha protein powder kaunsa hai').
The study counted the brands two answers to the same question have in common, as a percentage of all brands either answer named. Across 480 answers collected from inside India on 14 August 2026, two ChatGPT answers in the same language shared 54.7% of their brands, and an English and a Hinglish answer shared 52.5%. That 2.2 percentage point gap is the study's weakest result: 4 of the 10 buying questions moved the other way, and the whole effect came from the skincare questions.
ChatGPT cited websites in all 160 of its answers, about 2.5 sites per answer, and replied in the language it was asked in. On the same measure the gap was 7.8 points for Gemini and 23.4 for Perplexity, so the language of the question matters far more on those engines. Depra suggests Hinglish versions of your questions when India is one of your markets, and your report splits visibility by English and Hinglish. Read the English vs Hinglish AI shopping study for the method and the brand-by-brand tables.
Brands shared by two ChatGPT answers to the same question: 54.7% in the same language, 52.5% across English and Hinglish. (Depra Hinglish study, 480 answers, India, 14 Aug 2026)
Can OAI-SearchBot read your site?
OAI-SearchBot is the crawler OpenAI uses to show websites in ChatGPT search answers. OpenAI's crawler documentation says sites that opt out of it will not be shown in ChatGPT search answers, though they can still appear as navigation links. One line in your robots.txt, the file that tells bots which pages they may visit, can remove your site from the answers Depra tracks.
OpenAI documents three other agents. GPTBot collects pages that may be used to train its models, which is a separate choice from search. ChatGPT-User fetches a page when a person asks ChatGPT about it, and OpenAI notes that robots.txt rules may not apply to it. OAI-AdsBot checks the landing pages of ads submitted to ChatGPT.
Depra's AI crawler checker reads your robots.txt and your competitors', checks 41 AI crawlers from 26 vendors, and shows the exact rule that allows or blocks each one. For what every bot does, see the guide to AI crawlers. A blocked crawler is one of the causes in why ChatGPT doesn't mention your brand.
How to check ChatGPT by hand
You can run a small version of this yourself before you pay for a tool. The spreadsheet method for tracking brand mentions in AI answers has the full columns and scoring rules.
- Write 10 buying questions. Use the words a buyer types before choosing, with no brand names, such as 'best sunscreen for oily skin under ₹500'.
- Start a new chat with search on. Open chatgpt.com, start a fresh chat for each question, and check that the answer shows its sources.
- Ask each question 3 times. ChatGPT returns a slightly different list on each run, so one answer is a sample of one.
- Log every answer. Record named or not, position, tone and the cited websites in a sheet, then repeat each week.
Where the manual method stops working
Your own account's memory and location can change what ChatGPT shows you. 10 questions asked 3 times a week is 120 answers a month to log, before you add competitors or a second language. Depra asks from inside your chosen country every day and keeps every answer.