CapstonAI · Guide
A practical guide to seeing whether ChatGPT names and cites your brand — what to measure, how to run prompts, and how to monitor your share of voice over time.
ChatGPT is, for many buyers, the new first step before a search. If it recommends three brands in your category and you’re not one of them, you’ve lost the shortlist invisibly — no impression, no click, nothing in your analytics. Tracking mentions turns that blind spot into something you can measure and act on. It’s the ChatGPT-specific slice of broader AI visibility.
What makes this blind spot dangerous is that it leaves no trace in your usual reports. A lost organic ranking shows up as falling impressions in Search Console; a lost ChatGPT recommendation shows up nowhere at all. The buyer simply never hears your name, forms a shortlist without you, and you have no record that the conversation happened. Deliberate tracking is the only way to make that invisible loss visible — which is the first step to doing anything about it.
For the full definitions and how to calculate each, see AI visibility metrics.
A prompt set that only asks “what is [your brand]” tells you nothing useful — of course the model knows your name if you ask directly. The prompts that matter are the ones where the buyer hasn’t decided yet and your brand has to earn its way in. Cover these intents:
When ChatGPT browses and cites, it surfaces the pages it pulled from — and that list is a gift. Note which domains it cites for your category prompts: those are the sources the model currently trusts. If the same third-party listicles or competitor pages appear repeatedly, your route to being mentioned often runs through them (earning a place in those sources) as much as through your own site. Record the cited URLs alongside your mention/citation data so you can see not just whether you appear, but who the gatekeepers are.
The manual method above works for a one-time read. It breaks down at scale: dozens of prompts, multiple engines, repeated on a schedule, with competitor tracking, is a lot of spreadsheet work — and ChatGPT’s variability makes single runs unreliable. The AI visibility tool automates the capture and scoring, and connects the gaps to fixes through agents for WordPress, Shopify, Drupal and Chrome. CapstonAI measures and helps you improve — we are not an agency.
ChatGPT is the natural place to start because it’s where most buyers experiment first — but it’s one surface, and the engines disagree. A brand strong in ChatGPT can be invisible in Perplexity, which leans harder on live citations, or absent from Google AI Overviews, which sits in front of your existing search traffic. Each builds and cites answers on its own logic, so a single-engine read gives you a partial and sometimes misleading picture. Once your ChatGPT tracking is running, extend the same prompt set to Perplexity, Gemini and AI Overviews and compare. The pattern across engines is usually more instructive than any one of them: it tells you whether your gap is universal (a structural problem on your side) or surface-specific (a problem with how one engine sources answers). Treat ChatGPT tracking as the entry point to full AI visibility, not the whole job.
When you log results, separate two numbers that are easy to blur. Mention rate is how often ChatGPT names your brand in an answer; citation rate is how often it links your site as a source. They move independently and they have different fixes. A brand can be named frequently because the model learned it during training, yet rarely cited because no current page answers the query cleanly — that’s a content and structure problem on your live site. The reverse also happens: a page gets cited as a source without the brand being recommended by name, which is a positioning problem. Recording both, per prompt, tells you whether to work on being known or on being retrievable — and stops a healthy mention rate from hiding the fact that your own pages aren’t earning the citation.
Run a fixed set of category prompts, record whether ChatGPT names and cites your brand versus competitors, calculate share of voice, and repeat on a schedule.
Yes — when ChatGPT links to sources, you can record whether your domain is among them. Citation is separate from being mentioned by name.
Generative models can return different answers to the same prompt over time. That’s why tracking across multiple runs is more reliable than a single check.
The proportion of relevant answers where your brand appears, relative to all brands named for the same prompts. It’s the competitive view of your ChatGPT presence.
Enough to represent how buyers actually ask about your category — often 10–30 to start — covering different intents and competitors.
Treat it as a diagnosis. The usual causes are structural — unparseable content, weak entity signals, no llms.txt — which you can fix to improve future answers.
Run a free scan and find out whether ChatGPT names and cites your brand.
