LLM Rank Tracker
Track whether ChatGPT, Perplexity, Gemini, Claude and Google AI Overviews actually name your brand — and who they name instead. Free, browser-based, no signup.
Build your prompt pack and track every citation
Enter your brand and category. The tracker writes the buyer-style prompts that decide who gets recommended, opens each one in ChatGPT, Perplexity, Google AI Mode, Claude or Copilot, and scores what comes back. Nothing is uploaded — every run is stored in your own browser.
Ten prompts, five engines, one score. No signup and no email required.
What this tool does
The LLM Rank Tracker turns an unstructured question — “do AI assistants ever recommend us?” — into a repeatable measurement. It generates a set of ten realistic buyer prompts for your category, gives you a structured place to log what each engine actually returned, and converts those observations into three numbers you can track over time: an AI visibility score, a mention rate and a citation rate.
It also tracks who wins when you lose. If Perplexity recommends three competitors and never names you, that is the most useful piece of competitive intelligence available in AI search, and no traditional rank tracker will show it to you.
How it works
- Set up your check. Enter your domain, the category you want to be recommended for, and up to three competitors. Be specific with the category — “project management software for creative agencies” produces far more useful prompts than “software”.
- Run the prompts. The tool builds ten buyer-intent prompts and gives each one a copy button. Paste them into ChatGPT, Perplexity, Gemini, Claude, or a Google search that triggers an AI Overview.
- Log what you see. For each prompt and engine, record whether you were cited with a link, mentioned without a link, or absent entirely. Tick any competitor that got named instead.
- Calculate. The tool scores every logged result, breaks visibility down by engine, and works out your share of voice against the competitors you listed. Export to CSV, then repeat monthly.
What the results actually mean
| Metric | How it is calculated | What to do with it |
|---|---|---|
| AI visibility score | A citation with a link scores 2, a mention without a link scores 1, absence scores 0 — expressed as a percentage of the maximum for everything you logged | Your headline number. Track the direction, not the absolute value |
| Mention rate | The share of logged results where you appeared at all | Measures whether the models know you exist |
| Citation rate | The share of logged results where you were linked as a source | Measures whether your pages are retrievable and quotable, not just familiar |
| Per-engine breakdown | The same score calculated separately for each engine | Shows where to focus — engines diverge sharply |
| Share of voice | Your appearances versus your tracked competitors combined | The competitive read. A rising score means little if theirs is rising faster |
A high mention rate with a low citation rate is the most common and most fixable pattern. It means the models associate your brand with the category from training data, but your live pages are not being retrieved as sources. That is a technical and structural problem — crawler access, answer-first formatting and schema — and it is exactly what the GEO Audit checks.
Why LLM rankings behave nothing like Google rankings
- There is no fixed position. An assistant names a handful of options and stops. You are either in that set or you do not exist for that prompt. There is no page two to climb from.
- The same prompt gives different answers. Model responses are non-deterministic and personalized, so a single check proves nothing. Volume and repetition produce the reliable signal — which is why ten prompts logged monthly beats one prompt checked obsessively.
- Off-site content carries unusual weight. Assistants lean heavily on review sites, comparison listicles, documentation and forum threads. Your own homepage is often not the source cited when your brand is recommended.
- Ranking first does not guarantee citation. Plenty of pages at the top of Google are never quoted in an AI Overview, because they bury the answer under an introduction. Extractability is a separate property from ranking.
Why this tool does not call the AI APIs for you
It would be straightforward to have a server run all ten prompts across five engines automatically. We have deliberately not built that into the free tier, for two honest reasons.
The first is cost. Every automated run burns paid API tokens across multiple providers, and a genuinely free tool at any meaningful traffic level costs more to operate than it returns. Tools that offer this free usually cap it hard, gate it behind an email, or quietly degrade to a single cheap model.
The second is accuracy. API responses and the consumer apps people actually use are not the same product. ChatGPT in the browser has different retrieval behaviour from the raw API, and Google AI Overviews have no public API at all. Logging what you personally see in the real interface is slower, but it is a truer measurement of what your buyers encounter.
How often to re-run this
Monthly is the right cadence for most sites. AI answers shift constantly at the level of individual responses, but the underlying picture — whether the models consider you a credible option in your category — moves slowly. Checking weekly mostly measures noise. Checking quarterly means you find out about a decline three months late.
Keep the prompt set stable between runs. The moment you change the prompts, you lose the comparison. If you need to add prompts, add them alongside the originals rather than replacing them.
Frequently asked questions
Is the LLM Rank Tracker really free?
Yes, and there is no signup. The tool runs entirely in your browser, your log is stored in your own browser local storage, and nothing is transmitted to us. Clearing your browser data clears your saved log, so export the CSV if you want a permanent record.
How many prompts do I need to log for the score to mean anything?
Log all ten prompts in at least two engines before you read the score seriously. Below roughly twenty logged results, a single lucky or unlucky response swings the number too far. The tool will calculate a score from fewer, but treat it as directional only.
Why do ChatGPT and Perplexity give me such different results?
Because they retrieve differently. Perplexity is search-first and cites live web sources aggressively. ChatGPT blends training knowledge with browsing depending on the prompt and model. Gemini leans on Google infrastructure. Divergence between engines is normal and useful — it tells you which retrieval path is failing you.
Can I track more than three competitors?
The interface caps the list at three, deliberately. Share of voice becomes hard to read past that, and logging fifty checkboxes per run is the fastest way to abandon a tracking habit. If you need broader coverage, run a second pass with a different competitor set and compare the exports.
Does being cited matter more than being mentioned?
For traffic, yes — a citation carries a link a user can click. For brand and consideration, a mention still counts, because the assistant has named you as an option. The scoring weights citations at double a mention to reflect that difference without dismissing mentions entirely.
What should I do if my score is zero?
Start with crawler access, since blocked bots make citation impossible regardless of content quality. Then check whether your key pages answer questions in the first eighty words. Then work on off-site presence in the review sites and comparison articles that assistants lean on. The LLM SEO guide covers the full sequence.
Related tools and reading
Pair this with the AI Visibility Checker for a broader brand-level baseline, the GEO Audit to score a specific page for citation readiness, and the AI Overview Checker to see whether your keywords trigger an AI Overview at all. All of them are listed on the free tools hub.