How to Track ChatGPT Brand Mentions

Why LLM visibility matters now

ChatGPT, Perplexity, Claude and Gemini now serve billions of answers a month. When a buyer asks "what's the best rank tracking tool for agencies", the answer they see is generated on the fly — and if your brand isn't in that answer, you're invisible for that query no matter what your Google position says.

The good news: LLM answers are highly measurable. Every response cites specific URLs, mentions specific brands, and can be replayed. That means LLM visibility is trackable in exactly the same way rank tracking has been for Google.

What to actually track

Rank tracking measures position for a keyword. LLM tracking measures four things:

1. Cited — did the answer name your brand at all?
2. Position — was your brand mentioned in the first sentence, or buried at the end?
3. Source URL — which of your URLs did the model use as its citation?
4. Sentiment — was the mention positive, neutral, or critical?

Track those four across ChatGPT, Perplexity, Claude, Gemini, Meta AI and Grok, and you have the AI-search equivalent of a Google SERP report.

Step-by-step: build a manual tracker

1. List the prompts your customers actually ask. 20–50 is enough to start. Buying-intent prompts ("best X for Y", "X alternatives", "is X worth it") matter more than definitional ones.
2. Run each prompt in a fresh browser session against each model. New session = no personalization noise.
3. Record the full answer verbatim. Store the raw text; you'll want to re-parse later as your definition of a "good" mention evolves.
4. Parse for your brand — a case-insensitive string match is enough to start.
5. Log citation URLs — most models return a list of source URLs. Track which of yours got cited.
6. Repeat weekly. LLM answers drift as training data and retrieval indices update.

How LemRank does this automatically

LemRank's LLM Visibility module runs your prompt list against 6 AI platforms on a daily schedule:

• New browser context per query — no personalization bleed.
• Full response captured, brand mentions scored (0–100) by position and prominence.
• Cited URLs logged and mapped to your project sitemap.
• Daily deltas surfaced so you can see when a competitor overtakes you.
• Prompt suggestions generated from your existing keyword list so you don't have to write them from scratch.

Common mistakes

• Tracking only branded prompts. "What is LemRank" is easy to win — non-branded buying prompts are where the money is.
• Ignoring model diversity. ChatGPT and Perplexity return very different answers to the same prompt. Track both.
• Reading a single answer as truth. LLM answers vary between calls. Look at trends over a week, not one snapshot.
• Forgetting to update prompts. Buyer language shifts. Add 5 new prompts every quarter.

What LLM tracking does not do

• It doesn't guarantee traffic — being cited is upstream of visits, not the same thing.
• It doesn't fix your content — it tells you which prompts you're missing so you can fix it.
• It doesn't replace Google rank tracking — the two answer different questions.

Related

• How to appear in Google AI Overviews
• What is share of voice