Does ChatGPT mention your brand? Ask what your buyers ask, and count the answers.

What analytics and rank trackers cannot see

Your analytics can show a visit that came from ChatGPT, but not what ChatGPT said about you, or about your competitors, before that visit. A rank tracker cannot fill the gap. It records where your pages sit in a list of search results, and an AI answer is not a list: it is text written for one question, it names brands in its sentences, and it can change the next time the same question is asked.

The only way to know is to ask the same questions again and again and keep every answer. We built ai-visibility-checker, a free, open-source tool, to do that through the DataForSEO API. It asks ChatGPT, Gemini and Perplexity your buyers' questions, reads Google's AI Overviews for your keywords, and reports who gets named, which sites get cited and what changed since the last run.

What the first check showed

We first ran this check as a handful of scripts for an auto-parts retailer we work with. We asked ChatGPT, Gemini and Perplexity ten buyer questions, and pulled Google's results for the retailer's searches.

3 of the 30 answers named the retailer, all three on ChatGPT. The answers and the searches together cost $1.17 in API calls.

Those numbers are the baseline. They show which assistant already names the retailer, which questions it is missing from, and how rarely Google's overviews link to its site. The tool that grew out of those scripts saves a snapshot per run, so the next check shows what moved.

How the check works

• It asks the assistants. Each question goes to ChatGPT, Gemini and Perplexity through DataForSEO's LLM Responses API, with web search on, so every answer comes back with the sources it cites.
• It reads each answer: whether it names you, whether it cites a page on your domain, which of your competitors it mentions and which other names it lists.
• It checks Google. Each keyword goes to Google through DataForSEO's SERP API, with the AI Overview loaded even when Google loads it after the page, and the tool records whether an overview appears, names you and cites your domain.
• It reports and compares: the share of answers that name and cite you on each engine, the most cited domains, the answers that name a competitor and not you, and what changed since the previous run.
• It keeps the spend in check, with an estimate before the run, a spending limit, and a cache of the day's answers so a second run the same day costs nothing.

Named and cited are two different things

An answer names you when one of your brand names or your domain appears in its text as a whole word, ignoring case and accents. The tool reads the text a reader sees, so a link address alone does not count.

An answer cites you when one of its sources is on your domain or a subdomain of it. Gemini cites through Google redirect links, so the tool uses the original address that DataForSEO returns next to each one.

The report also lists the other names each answer puts forward, read from its bold text, list items and headings, which is where assistants put the companies they recommend. That part is a heuristic, so pass your competitors by name when you want them counted exactly. If your brand is also a common word, such as Notion or Monday, use a longer form like monday.com so everyday uses of the word are not counted as mentions.

Compare runs, not single answers

AI answers vary from run to run, even for the same question on the same day. The models choose their words with some randomness, and with web search on they can read different pages each time. So read the shares across many questions, and compare runs over weeks rather than single answers.

Each run is saved as a snapshot with every answer, its sources and its cost. Run the check again a week later and the report adds what changed in between: newly named, no longer named, newly cited and lost citations.

The tool asks the models through their APIs, with web search on. The ChatGPT and Gemini apps add memory, personalization and their own choice of model, so a signed-in buyer can get a different answer. Location and language matter too: ChatGPT and Perplexity search from the country you set, DataForSEO's Gemini endpoint takes no location, and Google results follow both the location and the language.

What a check costs

Prices as published on 25 September 2026. Each AI answer costs DataForSEO's $0.0006 task fee plus what the model's provider charges for tokens and web searches, which DataForSEO passes on.

Ten questions on the three assistants are estimated at $0.71. The tool prints its estimate before it spends anything, and --max-cost, $5 by default, refuses any run estimated above it. Real costs move with the length of each answer and the number of searches a model runs, so the report also totals what DataForSEO actually charged. A second run the same day, for example with another competitor, costs nothing, because the day's raw answers are cached.

The tool is free and MIT licensed, and needs a DataForSEO account with credit. The code, the options and a sample report are on GitHub: ai-visibility-checker. We run this kind of check as part of our work for clients, described on our AI automation agency page.