How to Track Brand Mentions in ChatGPT and AI Answers

Originally published on depra.ai on July 21, 2026 (updated September 15, 2026).

A practical, spreadsheet-based method for monitoring brand mentions across ChatGPT, Gemini, Perplexity and Google AI Overviews. You create 10-20 buying questions that do not name your brand, ask them multiple times per engine, and log the results in a structured sheet.

The 8-column tracking sheet

Column	Purpose	Example
Date	When the question was asked	15 Sep 2026
Engine	Platform used	ChatGPT
Language	English or Hinglish	Hinglish
Prompt	The exact question	India mein oily skin ke liye...
Named	Yes / No / No Overview	Yes
Position	Brand ranking in the response	2
Sentiment	Positive, neutral or negative, with a quote	Neutral: "YourBrand is a budget pick"
Cited sites	Source domains	Three domains

What counts as a brand mention

• Variants count: brand name variations, misspellings, domain names and product lines.
• Skip look-alikes: words that match your name in ordinary usage do not count.
• Position by first appearance: number brands in the order they appear.
• Sentiment by sentence: judge tone from the specific sentence that mentions you.
• Citations are separate: track source links independently from mentions.
• "No Overview" is a valid result: record it when Google does not show an AI Overview.

Setup: half a day spread over 3 days

Day 1, question writing

• Write 10-20 questions that buyers typically ask.
• Take the wording from reviews, support chats, Q&A sections and sales calls.
• Leave your brand name out, or results will be inflated.

Day 1, Hinglish versions

• Write Hinglish versions of your 5 most important questions.
• Use English letters with everyday Hindi words mixed in.
• Pair English and Hinglish versions with matching row IDs.

Day 2, running the questions

• Ask each question 3 times per chat engine, each in a separate new chat.
• Use private windows for Google searches.
• Log out or disable memory between runs.
• 15 questions x 3 engines x 3 runs is roughly 2 hours of work.

Day 3, roll-up

• Apply formulas to calculate mention share per engine.
• Work out average position and the most-cited sources.
• That gives you a baseline.

Ongoing

• Re-run the same questions weekly or monthly, on the same weekday.
• Keep old row IDs when questions are unchanged.
• Start a new trend line when you reword a prompt.

Why multiple answers matter

AI answers vary a lot from run to run. In Depra's data, two ChatGPT answers to the same question, asked within one hour, shared only 54.7% of the brands they named. Two Perplexity answers shared 68.3%.

Because of that, a single run can show a false positive or a false negative. Asking 15 questions 3 times gives you 45 answers per engine per round, which smooths out much of the randomness.

Tracking brand performance over time

• Freeze your question set.
• Ask on a fixed schedule.
• Compare mention-share percentages between periods.
• Always report the sample size next to the percentage.

95% confidence ranges

The range shows how much variation happens by chance alone.

Answer count	Mention share	95% range
20	50%	30% to 70%
50	50%	37% to 63%
100	50%	40% to 60%
135	50%	42% to 58%
400	50%	45% to 55%

Comparison rules

• Keep questions, engines and runs consistent across periods.
• Compare month to month, or in 4-week intervals.
• Count a change as real only when the confidence ranges do not overlap.
• Track competitors at the same time to spot platform-wide trends.
• Keep a Notes tab recording what changed and when.

Roll-up formulas

These work in Google Sheets and Excel.

Mention share on ChatGPT

=COUNTIFS(B:B, "ChatGPT", E:E, "Yes") / COUNTIF(B:B, "ChatGPT")

Mention share in AI Overviews

=COUNTIFS(B:B, "Google AI Overview", E:E, "Yes") / (COUNTIF(B:B, "Google AI Overview") - COUNTIFS(B:B, "Google AI Overview", E:E, "No Overview"))

Average position on ChatGPT

=AVERAGEIFS(F:F, B:B, "ChatGPT", E:E, "Yes")

Hinglish share on Perplexity

=COUNTIFS(B:B, "Perplexity", C:C, "Hinglish", E:E, "Yes") / COUNTIFS(B:B, "Perplexity", C:C, "Hinglish")

Share of voice: your mention count divided by the total mentions of all tracked brands.

Tracking English vs Hinglish

• Give English and Hinglish versions matching row IDs.
• Report the two languages side by side rather than blended.
• Expect big differences between engines.

Findings from the Depra AI study run on 14 August 2026 (full write-up):

• ChatGPT: 2.2 percentage point difference in brand overlap between English and Hinglish.
• Gemini: a 7.8 point gap, and sources were cited in 90% of English answers against 41.3% of Hinglish answers.
• Perplexity: a 23.4 point gap.

Writing Hinglish prompts well

• Use English letters and everyday vocabulary.
• Example: "India mein oily skin ke liye 500 rupees ke andar konsa face wash kharidna chahiye?"
• Avoid word-for-word Hindi translations.
• Have native speakers check each question pair.

When to switch to automated tools

Manual tracking stops being practical when:

• Filling the sheet takes longer than acting on the results.
• Daily tracking of 20 questions across 3 engines means 1,800+ answers a month.
• You keep skipping runs.
• Several people score answers differently.
• You need daily reporting.

About Depra AI

Depra AI automates this process:

• Tracks ChatGPT (with web search), Gemini and Google AI Overviews daily, and Perplexity weekly.
• Runs from the countries you choose, in English and Hinglish.
• Records mentions, position, sentiment with the quoted sentence, and cited pages.
• Includes competitor tracking.
• Excludes questions that contain your brand name by default.
• Exports data as CSV.

The Starter plan is ₹1,999/month plus GST and includes 16 prompt slots (8 English + 8 Hinglish questions) with up to 1,568 scheduled checks a month. There is a 7-day free trial.

Advantages of the spreadsheet approach

• No cost beyond your time.
• You can track any engine, including Claude and Microsoft Copilot.
• You can test logged-in answers with personal memory switched on.
• You have full transparency and control.

FAQ

How do I track brand mentions in ChatGPT for free?
Use a spreadsheet with 10-20 buying questions. Ask each one 3 times in separate chats, log whether you are named plus position, tone and sources, calculate mention share, and repeat monthly.

How many times should I ask each question?
At least 3 times per engine per round, in new chats, then pool the results across questions. At a 50% share, 20 answers give a 30-70% range and 100 answers give a 40-60% range.

What is AI brand monitoring?
Checking how ChatGPT, Gemini, Perplexity and similar engines mention your brand when users ask category questions: how often, in what competitive context, with what tone and from which sources. Also called LLM brand monitoring.

Should I track questions that include my brand name?
Keep them separate. Questions that name your brand almost always get a mention, which inflates share. Use them to see how AI describes you, and report share from unbranded questions.

Can I use Google Sheets to track this?
Yes. The 8-column format and formulas work in Google Sheets and Excel. Depra has no direct Sheets integration, but data exports as CSV.

Which engines should I track in India?
Start with ChatGPT, Gemini, Perplexity and Google AI Overviews, in English and Hinglish. Perplexity and Gemini changed brand recommendations the most between Hinglish and English in our study.

What is a good mention rate?
There is no standard benchmark. Compare your trend over time and against competitors on identical questions.

Does Depra do this automatically?
Yes. Depra runs buying questions daily on ChatGPT with web search, Gemini and Google AI Overviews, and weekly on Perplexity, from your chosen country in English and Hinglish. It stores full answers with scoring for you and your competitors.

Depra AI is AI visibility tracking software for brands in India. It checks how ChatGPT, Gemini, Perplexity and Google AI Overviews mention your brand, in English and Hinglish, from ₹1,999 a month. Based in India, working worldwide.