Call · 30 min
MarketingMattia EspositoUpdated 10 November 202613 min read

GEO tools to track ChatGPT. The free ones first, then what the paid ones claim and what they cost.

No tool sees inside ChatGPT: they all ask questions and count the answers, and each one picks different questions. Here is what each measures, what it costs, and an Excel sheet to compare them and test by hand.

In brief

To track ChatGPT and the other assistants, start with the free tools from Google and Microsoft, and pay for a third-party tool only if you need to repeat many questions every day. Search Console, Bing Webmaster Tools, Clarity and GA4 give data from the engine or from your site; for ChatGPT, Claude and Perplexity there is the manual test.

Paid tools repeat a set of questions and count mentions and citations. Google warns that no third-party tool has access to its internal systems.

On 10 November 2026, list prices for entry plans ran from €29 to €190 a month, with Profound publishing no price. The page lists them alphabetically: no ranking.

The Excel comparison sheet, with a three-round manual test, can be downloaded below, with no email required.

Written by: Mattia Esposito, AI and marketing consultant, founder of Itria AI. From Puglia, based in Turin since 2023, he works remotely in Italy and abroad. LinkedIn profile.

The GEO tools comparison sheet, to download

An Excel file with three sheets: a comparison of eleven tools, a manual test with ten questions in three rounds, and eight questions to ask a vendor before paying. The PDF carries the comparison table, to keep beside you during a demo. Presence in the manual test is calculated automatically.

The comparison table of tools for tracking ChatGPT and AI answers, with the free ones at the top and the list prices of the paid tools
FileContentsLink
Excelto fill in, in Italian

The Comparison, Manual test and Vendor questions sheets, with blank rows for a tool you are evaluating.

strumenti-geo-confronto-foglio.xlsx

PDFto read, in Italian

The comparison table and the questions to ask a vendor.

strumenti-geo-confronto-confronto.pdf

On 13 September I opened by hand three searches where my site sat fifth, sixth and ninth. All three had an AI Overview at the top, and itria.io was cited in none of them: the body that writes the rule was there, and in one, a courier. No tool had told me. Opening the page did.

In mid-October I counted again on Perplexity, still by hand, and put it in a post: “Perplexity cites me in 4 answers out of 19, and in one I am the only source.” Nineteen questions, one by one, with the spreadsheet open beside me. The method in the sheet above works the same way, and it takes an hour a month. The full post.

Paid tools do the same thing at a larger scale, every day, across more engines. What you need to understand before paying is how they choose the questions and how they read the answers, because every figure they show depends on it. This page sets out what each one states in its documentation.

How to track brand mentions in ChatGPT

By repeating the same questions over time and counting how many answers feature the company, by hand or with a tool. OpenAI offers no dashboard for publishers, unlike Search Console for Google and Bing Webmaster Tools for Copilot. From the outside, what you can see of ChatGPT are its answers and the visits it sends to your site.

These are two different measures. Answers tell you whether the brand gets named, and next to whom; visits tell you whether someone clicked. Visits are counted for free in your site analytics, as the guide on how to measure GEO explains. Answers need questions asked on purpose, and this is where the tools come in.

Free GEO tools first: what Google and Microsoft measure

The free tools give first-hand data, but only about the publisher’s own engine. They cover Google’s AI Overviews and AI Mode, Copilot and Bing’s AI summaries, and the visits assistants send to your site. Each report is covered in detail in the measurement guide; here you need the overall picture, to see what is left uncovered.

Free toolWhat it countsWhat it leaves out
Search Console, generative AI performance report

Impressions in AI Overviews and AI Mode, by page, country, date and device.

Queries and clicks in the report

GA4, AI Assistant channel

Visits from ChatGPT, Gemini, Copilot, Claude, Deepseek and Grok.

AI Overviews, AI Mode, visits without a referrer

Bing Webmaster Tools, AI Performance

Citations in Copilot, in Bing’s AI summaries and with some partners, with the grounding queries.

Traffic and competitors

Microsoft Clarity, Citations

Citation Rate, Queries Cited, Share of Authority.

Very low volumes: it is a sample

Server logs

Visits from bots that identify themselves, such as OAI-SearchBot and ChatGPT-User.

Clicks by real people

Manual check

What the assistants answer to the questions you choose.

The average across different users

What the free tools lack is ChatGPT, Claude and Perplexity seen from the answer side, and competitors: no official report shows who gets cited in your place. This is the gap the paid tools promise to fill. For Google, how the answers work is in the guide to AI Overviews and AI Mode, and how many clicks they bring in the one on AI Overviews and traffic.

How paid GEO tools work

They all carry out the same operation: take a set of questions, put them to one or more assistants at fixed intervals and count which answers mention the brand. Five choices vary, and each one changes the final figure. Google, in its guide to generative AI features, reminds anyone selling inside data: “No third-party tool has access to our internal ranking or AI systems.”

How an AI answer tracking tool works: it picks the questions, puts them to the engines through the interface or the API, collects the answers and counts mentions and citations The questionsyours or generated The engineswhich and how many The accessinterface or API The roundsdaily or weekly The countmentions and citations question 1question 2question 2 question 3questions 5-6
The five choices that make a figure what it is. Below each, the question from the “Vendor questions” sheet that brings it into the open.

The first choice is where the questions come from. Peec AI and Otterly.AI work on prompts you choose, in bundles per plan. Semrush and Ahrefs generate questions from keyword databases: Ahrefs from Google’s “People Also Ask” corpus, Semrush from browsing data on AI search and from AI Overview keywords. Questions generated this way are plausible, but they are not your customers’ questions.

The other four are which engines, through what access, how often and what gets counted. A mention without a link and a citation with a link are different things, and a report adding them together should be read with that in mind. The difference between the two, for a brand, is explained in the guide on how to get cited by ChatGPT.

Profound, Peec AI, Otterly, Semrush, Ahrefs: what they claim and what they cost

This table shows what each vendor writes in its documentation and the list price seen on 10 November 2026, in alphabetical order. These are vendor claims, not independent evidence. Prices can change by country and by monthly or annual billing: viewed from Italy, Peec and the Ahrefs product page show dollars, while Semrush, Otterly and the Italian Ahrefs pricing page show euros.

ToolHow it collects answers, according to the vendorList price, 10 Nov 2026
Ahrefs, Brand Radar

Questions from People Also Ask and the keyword database, run on ChatGPT, Perplexity, Gemini, Copilot, AI Overviews and AI Mode.

Prompt Tracking from $50 a month; AI Visibility Index from $199, from €179 on the Italian pricing page

Otterly.AI

Chosen prompts, across 7 engines, 4 per plan; credit-based. The collection method is not described on the pricing page.

€29, €189, €489 a month in the default view

Peec AI

Reads the interface, the way a user would; some Enterprise models via API. Chosen conversational prompts.

$95, $245, $495 a month

Profound

Structured prompts: citations, sentiment, position, competitors. In the trial the prompts are suggested and cannot be edited.

Free 7-day trial; paid plan on request

Semrush, AI Visibility Toolkit

Answers captured from real requests, not through APIs; questions from browsing data and keywords, grouped into topics.

€190 a month for the Starter SEO + AI Search bundle, €158.40 with annual billing

Peec’s entry plan includes 50 prompts on 3 models, checked daily: Peec counts one answer per prompt, per model, per day, so 4,500 answers a month. Otterly starts at 16 prompts, Semrush at 50 prompts a day, and Ahrefs includes 5 prompts in its Lite plan. Otterly’s page has a monthly/annual switch with a 15% discount: before paying, check which figure you are looking at.

Some vendor figures do not match from one of their pages to the next. The size of the Semrush database changes depending on the page you read, and the same goes for Ahrefs’ prompt count: so the table quotes none of them. Ahrefs also spells out its limits: “Chatbot usage is highly personalized, and the number of possible AI queries is effectively infinite.”

UI or API: why two tools give different figures

A tool can read the answer a user sees in the app, or request it from the model through the API, and the two answers may not match. Peec says it reads the interface “exactly as real users do”, and marks with “(API)” the few models it queries the other way. Semrush says it captures answers from real requests and not through APIs.

The difference matters because in the app, Peec writes, the platform chooses the model for most platforms, while the API answers with the model you specify. Two honest tools with the same questions can therefore give different percentages. A comparison only holds within the same tool, month on month, with the same questions.

Ahrefs adds that links invented by chatbots are not filtered out and that the tool does not estimate real usage by country. None of the five publishes the size of its Italian-language sample, and Ahrefs states its coverage is best in English: for a small business selling in Italy, this is the first question to ask.

Why your position in ChatGPT does not hold

Assistant answers change from one time to the next, and a ranking taken from a single answer does not repeat. Rand Fishkin of SparkToro measured it with Patrick O’Donnell of Gumshoe.ai, a company selling tracking, in January 2026: 600 volunteers, 12 questions, 2,961 answers from ChatGPT, Claude and Google’s AI.

What was measuredResultSource
Same list of brands in two answers

Less than a 1 in 100 chance, for ChatGPT and Google’s AI.

SparkToro and Gumshoe.ai, 27 Jan 2026

Same list in the same order

About 1 in 1,000.

SparkToro and Gumshoe.ai

Presence of the leading brands

Between 55% and 77% of 994 answers about headphones, for Bose, Sony, Sennheiser and Apple.

SparkToro and Gumshoe.ai

Chatbot usage

Highly personalised, with an endless number of possible questions.

Ahrefs, Brand Radar methodology

The useful conclusion comes down to one point: counting how often you appear makes sense, average position does not. This is why the sheet asks each question three times and works out the share of rounds where the company shows up. A tool showing a position in ChatGPT, without saying how many answers it is based on, is showing a figure the study calls unstable.

What I measure on itria.io, with the free tools

Bing and Google. Between 11 October and 7 November 2026, Bing’s AI Performance report counted 268 citations of itria.io pages in Copilot, and Search Console 3,550 impressions in AI Overviews and AI Mode.

Perplexity, by hand. In mid-October, 4 answers out of 19 cited itria.io, and in one it was the only source.

Claude. On 25 September 2026 Claude’s web search knew a single page of the site in the morning and five in the evening, after I asked Brave to recrawl seven: the case is in Brave and Claude.

The fault no report showed. The email on the contact page vanished from the HTML, and to an assistant the contact did not exist. The case is the email AI assistants can’t see.

GEO tools for small business: what is enough

For a small business the free tools and an hour a month of manual testing are nearly always enough; a paid tool starts to earn its keep when you follow dozens of questions or more than one market. The table matches each situation to the tool that covers it.

SituationWhat is enoughCost
You want to know whether Google shows you in AI answers

Search Console, generative AI performance report.

Free

You want to know whether Copilot cites you

Bing Webmaster Tools, AI Performance.

Free

You want to count visits from assistants

GA4 or your site analytics.

Free

You want to know what ChatGPT says on 10-20 questions

The manual test in the sheet, every month.

An hour a month

You follow dozens of questions, several engines or several countries

A paid tool, after the vendor questions.

From €29 a month, list price 10 Nov 2026

You want to compare yourself with competitors every day

A paid tool that states it does so, with export of the raw answers.

Depends on the plan

Buying a tool does not change what the answers say. If the company does not appear, the work is on the pages: opening the doors to crawlers, writing facts with their sources, getting named by the sources assistants read. Where to start is in the guide on how to show up in ChatGPT, and in the glossary entry on GEO.

The questions to ask before paying for a tool

Before a demo, ask where the questions come from, on which engines and through what access, how often they are repeated and how many are in Italian. All eight questions are in the third sheet of the file, with room for the answer. If the answer is vague, so is the figure the tool will show you.

Two questions matter more than the rest for an Italian business. First: how many questions in Italian and for Italy as a country are covered, given that none of the vendors read publishes its Italian sample. Second: can you export the raw answers, with date and engine, to check them by hand? A tool that allows it can be verified; one showing only a score cannot.

How to build this presence for a real business, with the right questions to own, is covered on the page about presence in search engines and AI assistants. The other guides in the series are in the SME marketing index, and the acronyms are compared in the one on the differences between GEO, SEO and AEO.

Questions and answers

How do you track what ChatGPT says about your brand?

By asking ChatGPT, at fixed intervals, the questions a customer would ask, and counting how many answers name the company or cite it with a link. You can do it by hand, with a spreadsheet, or with a paid tool that repeats the questions every day.

For visits arriving from ChatGPT, your site analytics are enough.

Are there free GEO tools?

Yes. Search Console counts impressions in AI Overviews and AI Mode, Bing Webmaster Tools counts citations in Copilot, Microsoft Clarity gives a citation rate and GA4 counts visits from assistants. They cover Google and Microsoft; for ChatGPT, Claude and Perplexity you are left with the manual test.

How much does a tool for tracking ChatGPT cost?

At the list prices seen on 10 November 2026, from €29 a month for Otterly.AI to €190 a month for the Semrush bundle with AI search; Peec AI starts at $95, Ahrefs at $50 for prompt tracking alone. Profound does not publish the price of its paid plan. Prices change by country and by billing period.

Can GEO tools see ChatGPT's or Google's internal data?

No. Google says so for its own systems: no third-party tool has access to its internal ranking or AI systems. Paid tools put questions to the assistants, through the interface or the API, and count the answers.

Is it worth tracking your position in ChatGPT?

Hardly. In a SparkToro study with Gumshoe.ai, across 2,961 answers, the chance of getting the same list of brands in two answers was below 1%, and the same order about one in a thousand. How often a brand appeared was far steadier: the figure to follow.

Notes on sources

  1. Google Search Central, Optimizing your website for generative AI features on Google Search, updated 10 July 2026, CC BY 4.0: the first quotation. Search Console Help on the generative AI performance report, available to all sites since 31 August 2026.
  2. Google Analytics Help on default channel groups, entry of 13 May 2026 on the AI Assistant channel.
  3. Bing, AI Performance, 10 February 2026, and the additions of 16 June 2026; Microsoft Clarity, AI citations, 17 February 2026. OpenAI, Overview of OpenAI Crawlers.
  4. Ahrefs, Brand Radar Methodology, published 29 October 2025 and revised 16 June 2026, source of the second quotation; product page and pricing. Vendor.
  5. Otterly.AI, pricing; Peec AI, pricing and method, source of the third quotation; Profound, pricing; Semrush, data sources and pricing. Vendors, read on 10 November 2026 from an Italian address.
  6. Rand Fishkin with Patrick O’Donnell, AIs are highly inconsistent when recommending brands or products, SparkToro, 27 January 2026: a study carried out with a tracking vendor, conflict of interest declared by the author.
  7. The itria.io figures come from Search Console (generative AI report) and Bing Webmaster Tools, window 11 October to 7 November 2026; the Perplexity count from the post of 14 October 2026.
·Who does it for you

A tool tells you whether you appear. What to write so they choose you is the work after that. Positioning is built starting from the market.

Mattia Esposito, who wrote this guide, also works with Itria on positioning: in-depth market analysis to find where competitors are exposed and take that ground. He did it on his own site first. In August, asked who had founded itria.io, ChatGPT couldn’t find him. From 11 October to 7 November 2026 itria.io pages appeared 13,803 times in Google and Bing results and 3,818 times in AI answers: Google’s AI Overviews and AI Mode, and Microsoft Copilot.