Are There Open-Source LLM Rank Trackers?

Which open-source trackers exist?

As listed on GitHub:

Ansvisor can be self-hosted for free, or used as a managed cloud service. Self-hosting needs Node.js, a Supabase project (whose free tier works), at least one AI provider key, and a scraping service for AI platform responses. It lists ChatGPT, AI Overviews, AI Mode, Gemini, Claude, Perplexity, Copilot, Grok and ChatGPT Shopping among supported platforms.

The AWS sample is explicit about its purpose. It sends your keywords to each engine with web search enabled, stores every answer and citation, extracts the brands named, and crawls the cited pages. It warns that it is sample code for demonstration and education, and needs security review before production use.

Open-source AI visibility audit tools: what they check and recommend

Audit tools answer a different question: can AI engines reach, read and cite your site? Three active examples:

• geo-score (MIT): gives a free 0 to 100 readiness score in about 20 seconds against an open, versioned rubric. With your own API keys, it can also ask engines a question, or watch a question set weekly. Its authors keep citation results separate from the readiness score.
• GEO Optimizer (MIT): a Python toolkit that scores a site 0 to 100 on whether ChatGPT, Perplexity, Gemini, Claude and Google AI Overviews can crawl, understand and cite it, then lists fixes.
• open-geo (MIT): a command-line audit suite, described as "like Lighthouse" for GEO. It checks robots.txt, llms.txt, Schema.org markup and AI crawlability, runs locally, and offers an MCP server.

These tools typically check:

• robots.txt rules for AI crawlers such as GPTBot, OAI-SearchBot, ClaudeBot and PerplexityBot.
• Rendering: whether key content is in the HTML or depends on JavaScript.
• Structured data and metadata.
• llms.txt presence.
• Content structure: headings, answer-first paragraphs and clear facts.

Treat their scores with care. A readiness score measures preconditions, not citations. Some rubrics give weight to items with little evidence behind them. Google says it does not use llms.txt, and tests show assistants ignore JSON-LD when fetching pages. The robots.txt and rendering checks are the most valuable parts, because a page that cannot be fetched or read is much less likely to be cited.

What does self-hosting really cost?

The code is free. The running costs are not:

• API fees. OpenAI and Anthropic charge $10 per 1,000 web searches, plus tokens. Google's Gemini API grounding costs $14 per 1,000 requests after 5,000 free a month, but its terms forbid storing, analysing or programmatically collecting grounded results and their links, so it cannot be used to track citations.
• Scraping services, if you collect answers from consumer interfaces rather than APIs.
• Hosting and database.
• Maintenance: keeping up with API and interface changes, which can break collection.

A worked example: 100 prompts on 4 engines, weekly, is about 1,730 answers a month. At one to two cents per answer, that is roughly $17 to $35 a month in search fees, before tokens and hosting. Daily tracking of the same set is about 12,000 answers, or $120 to $240. That is comparable to paid mid-market tools, before counting staff time. Open source saves money mainly at low frequency, or when you already run the infrastructure.

Does the collection method matter?

Yes. Most open-source trackers use the engines' APIs. API answers can differ from what people see in the consumer apps, which may use different models, settings and personalisation. Surfer's 2026 comparison of the two methods found that brand lists overlapped by only about 21% to 32%. Ansvisor supports scraping services for consumer responses, but scraping raises cost and terms-of-service questions. OpenAI's terms, for example, prohibit automatically or programmatically extracting data or output from its services. Decide which you need, API approximations or app-level answers, before choosing a project.

How should you evaluate an open-source project?

1. Licence. MIT and Apache allow commercial use freely. AGPL requires sharing modifications if you offer the software as a network service.
2. Activity. Recent commits, releases and answered issues. Several projects stall within months.
3. Engines and method. Which platforms, and whether through APIs or scraping.
4. Data model. Whether it stores full answers and cited URLs, not just scores.
5. Security. API keys and client data need proper handling, especially for agencies.
6. Exit. Whether you can export everything in a standard format.

What about building your own?

A basic tracker is a few hundred lines of code: a list of prompts, calls to each engine's API with web search on, where the provider's terms allow storing results, which Gemini's grounding terms do not, and a table that stores the answer text, cited URLs and whether your brand appears. The hard parts are elsewhere. Detecting your brand reliably across spellings and similar names, resolving cited URLs to real sources, scheduling runs, and handling API changes all take ongoing work. Starting from an existing open-source project usually saves weeks, even if you later change most of it.

Who should use open source?

• Good fit: technical marketing teams, agencies wanting a white-label or self-hosted setup, and companies with strict data rules.
• Reasonable fit: analysts who want a transparent method and are happy to maintain scripts.
• Poor fit: teams without developer time, or anyone who needs dependable support.

Paid alternatives are compared in Athena vs Farse vs Aiclicks and Peec vs Determ vs SE Visible. How to track at product level is covered in how to track product visibility by SKU in LLM answers.

What do people ask about open-source LLM trackers?

Is there a free, no-code option?

Not really among open-source projects. For no-code free checks, use vendor free tiers, Google Search Console and Bing Webmaster Tools.

Can I use open-source trackers commercially?

Under MIT and Apache licences, yes. Check AGPL obligations before offering a modified version as a service.

Are open-source results comparable to paid tools?

Only if the prompts, engines and collection method match. Compare on the same prompts.

What is the most active project?

Among dedicated trackers, Ansvisor was updated almost daily in September 2026.