An LLM visibility tool tracks what large language models say about your brand, model by model.
LLM visibility is how often the large language models behind assistants such as ChatGPT, Claude and Perplexity name your brand when people ask about your category. An LLM visibility tool, also called an LLM tracker, puts your buyers' questions to those assistants again and again and records who is named, which sources the answer used, and which model version wrote it. Proofsource is one.
Is the model the same thing as the assistant your buyer uses?
No, and the difference decides what a tool is really measuring. A buyer types into ChatGPT or Claude, a product that wraps a model with instructions, web search and sometimes memory of past chats. The bare model, called directly without search, can only answer from what it learned in training.
The two can name different brands for the same question. So the first thing to ask an LLM tracker is which surface it asks: the product as a buyer sees it, or the model without its tools. Neither is wrong, but only the first matches what your buyers read.
Memory answers and search answers: why does an LLM tracker need to tell them apart?
Every answer is part memory and part lookup. The memory half is what the model learned in training, and it changes only when a new model ships. The lookup half is what the assistant fetched while answering, and it can change within days of a page being published or edited.
An answer that cites sources used the lookup half; one that cites nothing came from memory. In the public Tesla example, Claude answered "What is Tesla?" and "Is Tesla legit?" from what the model already knew, with no sources cited. Work on your pages can reach the first kind of answer quickly and the second only after the next model is trained, so a tracker that shows the sources behind each answer tells you which lever you have.
Why does the model version matter when tracking LLM visibility?
Assistants swap their underlying models from time to time. When they do, the brands they name can shift overnight with nothing changed on your side. If a tracker does not record which model wrote each answer, a model upgrade looks exactly like a win or a loss you caused.
Proofsource records the model version behind every answer and prints it under each chart. When a model changed partway through the period on screen, both versions are listed, so a jump that lines up with a new model is visible as one.
Is LLM visibility the same as AI visibility?
Nearly. AI visibility is the wider term and also covers Google AI Overviews, a search feature with no public model version to follow. LLM visibility puts the model first: what each assistant's model knows, what it looks up, and how that shifts between versions. The measurement is the same, which is why Proofsource asks the same questions on all 4 engines. The broader picture is on the AI visibility tool page.
Three things to record per answer
Who the model names
Your brand and every rival in the answer, for each question and assistant.
Sources, or none
The pages the answer cited, or the fact that it answered from training alone.
Which model wrote it
So a model upgrade is never mistaken for something you did.
How does Proofsource track LLM visibility?
Proofsource is an AI intelligence platform that shows which buyer questions name your brand in ChatGPT, Perplexity, Claude and Google AI Overviews, who gets named instead, and exactly what to fix.
- Clean sessions. Questions are asked with no personal account history, so no one user's memory shapes the answer.
- The full answer is kept. Every figure opens to the answers behind it, with the brands named and the pages cited.
- The model behind each answer, printed under every chart.
- A 95% Wilson interval on every rate. 16 named answers out of 25 is 64%, but the honest range is about 45% to 80%; 60 out of 200 narrows to about 24% to 37%.
What should an LLM tracker record?
- Which surface it asks. The assistant with its search, or the bare model.
- The model version per answer, not today's default.
- Sources or no sources, so you can tell lookup answers from memory answers.
- Whether it is you. A model can describe a different company with a similar name. Read the answer, not only the match on your name.
- Enough repeats for an interval. One answer per question is a sample of one.
How do the leading tools compare?
3 criteria, the same for every tool. Each competitor fact links to the vendor's own page.
| Tool | What it is | How often it asks | Intervals and stats |
|---|---|---|---|
| Proofsource | AI intelligence platform that shows which buyer questions name your brand, who gets named instead, and drafts the fix for each gap. | Daily by default on Custom and Managed, or hourly or any frequency you choose, set at onboarding; the trial runs 2 scans about a day apart | A 95% Wilson interval on every percentage; changes are flagged only when they clear it |
| Profoundsource | Enterprise AI search platform with answer engine insights, agent analytics and an AI Marketer agent. | Daily | Not published |
| Peec AIsource | AI search analytics for marketing teams and agencies: visibility, position, sentiment and sources. | Daily (daily or weekly on Enterprise) | Not published |
| Scrunchsource | AI search monitoring plus an Agent Experience Platform that serves AI agents an optimised version of your site. | Not published (5,000 responses a month on Core) | Not published |
| Otterly.AIsource | Self-serve AI search monitoring for ChatGPT, Google AI Overviews, Perplexity and Copilot, with GEO audits. | Daily | Not published |
| Ahrefs Brand Radarsource | AI mention tracking inside Ahrefs, built on a database of hundreds of millions of prompts plus your own. | Custom prompts refreshed daily | Not published |
The full comparison of eight tools, with what each does best, is in the best AI visibility tools.
- Profound pricing www.tryprofound.com/pricing
- Peec AI pricing peec.ai/pricing
- Scrunch pricing scrunch.com/pricing
- Otterly.AI pricing otterly.ai/pricing
- Ahrefs pricing ahrefs.com/pricing
- Ahrefs Brand Radar ahrefs.com/brand-radar
Sources read on .
What people ask about LLM visibility tools and LLM trackers
What is an LLM tracker?
Another name for an LLM visibility tool: software that repeatedly asks AI assistants built on large language models the questions your buyers ask, and records which brands they name and which sources they cite.
Which LLMs should I track?
The assistants your buyers use. Proofsource tracks ChatGPT, Perplexity, Claude and Google AI Overviews on every plan, and a Custom plan can add more engines. The table lists each tool's engines with a source.
Can I change what an LLM says about my brand?
Partly, and at two speeds. Answers that search the web can pick up a clearer page within days of it being fetched. Answers from the model's training change only when a newer model is trained on newer text. Track sources per answer to see which kind you are dealing with.
How many answers does a reliable LLM visibility figure need?
More than most people expect. At 25 answers a 64% rate has an interval of roughly 45% to 80%; at 200 answers the interval is a few points either side. Proofsource shows the interval on every figure so the sample size is never hidden.
See what the models say about your brand.
See which buyer questions name your brand, who gets named instead, and what to fix. 200 answers. 25 Prompts/Question. Top AI engines. 2 days. Free. No card.
Google AI OverviewsGoogle AI Mode
