# How to Track ChatGPT Mentions of Your Brand (The Right Way) Learn how to track ChatGPT mentions of your brand: the five metrics to watch, the denominator brands get wrong, and what visibility tracking really takes. **Published:** September 14, 2026 **Author:** Kate Starr --- Your CEO asks a simple question: "does ChatGPT recommend us?" If you type your category into ChatGPT and read the answer, you won't learn much. That's because a single prompt is only a sample of one, so it tells you very little about consistent visibility. A useful visibility number needs more than a single prompt check. You need a consistent way to measure mentions, citations, share of voice, position, and sentiment, all while keeping the denominator and sampling method consistent. In this article, you'll learn what these metrics actually tell you and why the math behind them matters. We'll also cover the four requirements your data must meet before you bring it to your CEO. ## How do you track ChatGPT mentions? Tracking ChatGPT mentions means measuring how often your brand appears across a defined set of buyer questions over time. It's one part of [AI visibility tracking](https://www.searchable.com/blog/ai-visibility-tracking), which covers every AI engine your buyers use. Because ChatGPT can return different answers to the same question, tracking relies on structured statistical measurement rather than point-in-time checks. This article covers what AI visibility metrics reveal about brand presence. It details how shifting AI prompt sets and run counts distort visibility percentages. It also outlines the exact methodological requirements your data must satisfy before you can use it in reporting. ## Why you can't just ask ChatGPT whether it mentions your brand A simple ChatGPT query won't reliably tell you whether your brand is mentioned, for three main reasons: First, **answers can change between runs.** If you ask ChatGPT the same question twice, you may get different answers with different brands recommended. One response tells you what happened on that run, but it doesn't tell you how consistently your brand appears. **Your own session is contaminated.** When you use ChatGPT with your own account, your chat history, memories, and custom instructions can affect the response you get. So, your result may differ significantly from what another user gets. At minimum, use a logged-out or temporary chat when you check brand mentions to avoid a biased response. And finally, **you can't see real user prompts.** Unlike Google Search Console, ChatGPT doesn't show you the list of queries people used to find your site. That means you have to decide which questions to measure yourself, and those choices shape the visibility number you get. If your prompt set doesn't reflect real search intent, the results won't be reliable. ## What's the difference between a ChatGPT mention and a citation? A mention means an AI answer names your brand without linking to it, which signals that you're recognized in your niche. Like this: ![ChatGPT answer to "What's the best CRM for startups?" with the brand names Attio, HubSpot, Pipedrive, Close, folk, and Salesforce boxed in its comparison table](https://www.searchable.com/blog/track-chatgpt-mentions-chatgpt-mention.webp) A citation means the answer includes an in-text link to a source, which signals [where ChatGPT gets its data](https://www.searchable.com/blog/where-does-chatgpt-get-its-data) about you: ![ChatGPT answer recommending AI tools for LinkedIn carousels, with the in-text source links after each tool boxed](https://www.searchable.com/blog/track-chatgpt-mentions-chatgpt-citation.webp) This in-line link can lead either to your website or a third-party page about your brand, like a review site, news article, directory, or blog. | | Mention | Citation | | :---- | :---- | :---- | | **What happens** | Names your brand | Links to a source | | **What it tells you** | Whether your brand is recognized in your category | Where ChatGPT gets its information about you | | **Where it leads** | Nowhere | Your site or a third-party page | Since these are two different metrics, your mention and citation counts will likely differ as well: ![Searchable Mentions and Citations cards side by side for a bike brand, with the 9,518 total mentions and 8,396 total citations boxed](https://www.searchable.com/blog/track-chatgpt-mentions-mentions-citations.webp) Reading either alone leads you to the wrong strategy. For example, high mentions with low citations means that ChatGPT knows your brand, but relies on other sources to support its claims. If this is the case, publishing more content on your own site might not help. Instead, focus on digital PR and earned coverage, because earned media accounts for 84% of AI citations, according to [Earned media still drives 84% of AI citations](https://muckrack.com/blog/what-is-ai-reading-may-2026). ## Which ChatGPT metrics should you track? When tracking your ChatGPT visibility, pay attention to five key metrics: **1\. Visibility score** is the percentage of tracked AI answers in which your brand appears. It tells you how consistently AI considers your brand relevant to the questions your audience asks. **2\. Share of voice (SoV)** measures your brand's share of total industry mentions in AI responses. It shows you how frequently AI recommends or names your brand compared to competitors. **3\. Average position** is the average rank or position your brand holds when AI lists multiple brands in an answer. It indicates whether AI sees your brand as a leading option or not. ![Searchable Share of Voice and Avg Position cards for a bike brand, with its 28.5% share of voice and #2.1 average position boxed](https://www.searchable.com/blog/track-chatgpt-mentions-share-of-voice.webp) **4\. Mention-to-citation ratio** is the relationship between how often AI mentions your brand and how often it cites your content as a source. It tells you whether AI recognizes your brand or considers your content authoritative enough to support its answer. **5\. Sentiment** is the overall tone AI uses when describing your brand, which can be positive, negative, or neutral. It shows whether rising visibility is building trust or damaging your brand, because appearing in a hesitant or even negative response isn't the same win as an outright recommendation. ![Searchable Sentiment Score chart with positive, neutral, and negative bars over seven days, next to a Head-to-Head card with the attributes AI associates with the brand boxed, including premium pricing as a negative](https://www.searchable.com/blog/track-chatgpt-mentions-sentiment.webp) These five metrics give you a clear picture of your AI visibility, but only if the math behind them is right. Each number depends on how you define the prompt set, sample the answers, and calculate the denominator. The math is simple, but most teams get it wrong. ### The denominator most brands get wrong To calculate AI visibility accurately, you need to include every single prompt you used in a test set, even the ones that name no brands at all. So, if you run 100 prompts and your brand shows up in 20, your visibility score is 20%. Calculating it as 20 out of the 40 prompts that mentioned a brand would incorrectly inflate your score to 50%. You need to get this baseline right in order to diagnose your visibility and decide how to improve it. ### Understanding the metrics and what to do with them To interpret the metrics, think of your performance across two key axes: **visibility vs. share of voice** and **visibility vs. average position**. | Pattern | What it means | What to do | | :---- | :---- | :---- | | Low visibility + low SoV | Invisible to AI | Build baseline presence and get indexed | | High visibility + low average position | Frequently mentioned, but buried at the bottom | Improve messaging and positioning to get a higher rank | | Low visibility + high SoV | Strongly recommended, but only on a few niche prompts | Expand topic coverage to trigger more prompts | | High on both axes | Consistently recommended as a top choice | Maintain brand authority and defend your position | Identifying where your brand falls on these axes reveals your blind spots. Once you know whether your issue is reach, rank, or relevance, you can target your optimization efforts instead of guessing what to fix. ## Why does the same question produce different brand lists? AI visibility fluctuates for two main reasons. First, the same prompt can return different answers each time you run it (**run-to-run variance**), and second, different ways of asking the same core question can trigger entirely different results (**phrasing variance**). Because of phrasing variance, your test set needs to include a wide range of prompt variations to capture how real people search. Take three users searching for the same product: * One asks for the "best CRM for startups" * Another asks "what CRM a small team should use" * And a third looks for an "affordable CRM for a 10-person company" Even though all these prompts have the same intent, an AI model can generate completely different answers for each. Your brand might consistently show up for one variation and not for the rest. That makes measurement challenging. If your prompt set mostly contains phrasings you already rank for, your visibility score creates a false sense of security. A strong score may simply mean you're winning one specific version of the question. This matters because prompt-set design decides whether your visibility score means anything. Running N prompts R times isn't enough. What you actually need is **buying intents × phrasings per intent × runs**. As a best practice, track at least three different phrasings for every buying question that matters. If your brand appears for one phrasing but not for others, that isn't true coverage. Your visibility is simply tied to a specific way of asking the question. Buyers also tend to ask full, conversational questions rather than clipped keyword phrases, so write your phrasings the way a person would actually ask them. ## What makes a ChatGPT visibility number defensible? A defensible ChatGPT visibility number needs a consistent prompt set, repeated sampling, and a stable baseline. Without these conditions, a changing score is difficult to interpret. It could mean your visibility shifted, or it could just mean your measurement methodology changed. Your methodology must at least meet these four criteria: * **A prompt set built from buyer intent**, weighted toward the bottom of the funnel and kept strictly unbranded. Branded prompts flatter performance and skew the baseline since the user has already put your name into the question. * **Several phrasings per buying question**, because the same buying question can produce different answers when the wording changes. * **Repeated sampling**, because each prompt needs at least three runs in a clean session. One response is a snapshot, while repeated runs give you a better picture of how consistently your brand appears. * **An unchanged prompt set over time**, because a comparison only means something if the prompts, metrics, and run order are identical to the baseline. If the prompt set drifts, it invalidates everything you've already measured. These requirements make rigorous tracking more demanding than a quick manual check, but they're ultimately what allow you to distinguish a real change in visibility from a change in how you measured it. For example, take 30 buying questions, three phrasings per question, and three runs each week. That's 270 ChatGPT queries a week, or more than 1,000 queries a month for one category. If your brand has six categories, you're already past 6,000 queries a month. ## How Searchable tracks ChatGPT mentions Searchable tracks ChatGPT, Perplexity, Claude, Gemini, Copilot, Google AI Overviews, Google AI Mode, Grok, and DeepSeek in one place. We run the same prompt set across multiple AI engines at the same time, and show the results in one [AI visibility analytics](https://www.searchable.com/features/aeo-insights/visibility) dashboard. Those engines source their answers differently, so the brands they name are often not the same. Comparing the results side by side means you don't have to repeat the same analysis for each engine. The tracking itself is split across a few features, each answering a different question about your AI visibility: **1\. Prompt Research → Generate Prompts & Topics** [Prompt Research](https://www.searchable.com/features/prompts) builds the prompt set you'll use to track your brand's AI visibility. You can generate prompts from your brand, a seed keyword, or competitors. Then group them by topic, funnel stage, and branded vs. unbranded queries. ![Searchable prompts table for a bike brand, with prompts grouped by topic and columns for visibility score, average position, sentiment, topic, and branded status, and one topic group boxed](https://www.searchable.com/blog/track-chatgpt-mentions-prompts.webp) **2\. AEO Insights → Visibility** AEO Insights shows how visible your brand is across the tracked AI engines. You get a visibility score, share of voice, average position, and rankings with filters for engine, topic, location, date range, and branded vs. unbranded queries: ![Searchable Visibility card with a 35.1% visibility score boxed, next to a Rankings table comparing eight bike brands on visibility, share of voice, sentiment, and average position, with the brand's own row boxed](https://www.searchable.com/blog/track-chatgpt-mentions-visibility.webp) **3\. AEO Insights → Mentions & Citations** Here, you'll see mentions and citations side by side. You can see how often your brand is mentioned in AI answers alongside how often AI cites your site as a source, which funnel stage those mentions come from, and where your brand lands in each answer: ![Searchable Decision Journey funnel showing mention rate by funnel stage, next to a Position Distribution table with the brand's own row boxed](https://www.searchable.com/blog/track-chatgpt-mentions-position-distribution.webp) **4\. AEO Insights → Sources** Sources show which domains AI engines are using to support their answers: ![Searchable Sources view with a Top Domains citation trend chart for five domains, with youtube.com and reddit.com boxed, next to a breakdown of cited content types](https://www.searchable.com/blog/track-chatgpt-mentions-sources.webp) Exclude your own domain and you can see which sites are earning citations for the same topics where competitors are appearing. **5\. Query Fan-Out** Query Fan-Out shows the sub-questions AI considers when answering a prompt. For example, a question like "What is the best mountain bike for aggressive trail riding?" leads the engine to search smaller, related questions. It does that to collect information before providing you with an answer. Searchable lets you see those sub-queries for each prompt you're tracking: ![Searchable Query Fanouts by Prompt table listing six tracked bike prompts with the number of sub-queries each one generated, with the mountain bike prompt from the example boxed](https://www.searchable.com/blog/track-chatgpt-mentions-query-fanout.webp) **6\. View as Competitor** View as Competitor lets you look at the same visibility data from a competitor's perspective. Instead of only asking where your brand appears, you can see where your competitor is showing up and where they aren't. **7\. Reports** Reports give you a dated, shareable snapshot of your AI visibility. This helps you keep your reruns consistent and share the results with leadership, clients, or internal stakeholders. ### The half of tracking most tools miss Most ChatGPT tracking tools tell you what happens inside the AI answer, including whether your brand was mentioned, where it appeared, and which sources were cited. Searchable also looks at what happens on your own website. Our [LLM Analytics](https://www.searchable.com/features/llm-analytics) shows which AI bots crawl your site, which pages they access, and how much traffic arrives when someone clicks your site from an AI answer: ![Searchable LLM Analytics Crawl to Click Attribution table comparing AI crawls with referral clicks by platform, with the OpenAI row boxed, next to a Platforms chart of AI crawl share](https://www.searchable.com/blog/track-chatgpt-mentions-crawl-attribution.webp) That gives you three parts of the journey: **crawl → cite → traffic.** **Crawl** tells you whether AI systems can access your content. If AI crawlers can't fetch your pages, you'll be absent from answers entirely. [69% of AI crawlers can't execute JavaScript](https://www.searchviu.com/en/ai-crawlers-javascript-rendering/), so client-side-only content is invisible regardless of quality. **Citations** tell you whether they are using it in their answers. And **traffic** tells you whether those answers are sending people to your site. ## What to do once you can see your mentions Once you can see where and how your brand appears in AI answers, the next step is figuring out why you're seeing that pattern. If you're not retrieved at all, check if your pages are [readable to AI crawlers](https://www.searchable.com/blog/we-built-a-dual-served-content-system-for-ai-crawlers). If you're retrieved but not cited, look at how useful your pages are as sources. Adding statistics, quotes, and cited sources improved visibility in AI answers by up to 40% in the [Princeton GEO study](https://collaborate.princeton.edu/en/publications/geo-generative-engine-optimization/), and our guide on how to [increase AI search citations](https://www.searchable.com/blog/increase-ai-search-citations) walks through what to fix first. If you're cited but not moving up, build content clusters with relevant topics. AI visibility is often about how well you cover the topic across your site, not how strong just one article is. Finally, if your brand appears in AI-generated answers but is described negatively, focus on PR. Work on getting reputable publications to mention your brand, rather than publishing more content on your own site. ## Frequently asked questions ## Start measuring your brand visibility across AI search The number only means something when you know exactly what's included in its measurement and can confidently defend it. You need to know why your prompt sets reflect buyer intent and have a consistent baseline for comparison over time. The same approach works beyond ChatGPT. Point that prompt set at the other AI engines your buyers use to see where your brand appears across AI search. Once you have a consistent prompt set, repeated sampling, and a stable baseline, you can start treating changes in visibility as something worth investigating rather than assuming every percentage change represents a real shift. Want to know where your brand shows up in AI search? [Start tracking your AI visibility](https://www.searchable.com/free-visibility-report) with Searchable. --- [Back to Blog](https://www.searchable.com/blog) | [Searchable Homepage](https://www.searchable.com)