Open-source AI visibility tools
Open-source AI visibility tools let you self-host your brand tracking and read the exact code behind every metric. It is a small, early space: a handful of projects, most of them narrow and early. If none of them fit, you can also script your own checks against the AI model APIs.
Why open source matters here
A number that lands in a board report or sets a content budget has to be checkable by someone who did not produce it. Open source is one way to get there: you can read how a metric is built, and you can run the whole thing yourself, so prompts and history never leave your environment.
It is not the only way. Reading the code tells you how a number would be computed; the stored answers tell you what it was actually computed from. Whichever route you take, the thing worth insisting on is that every figure can be traced back to something you are allowed to look at.
The open-source options
The honest picture is that this is a thin, early space. The open-source projects we track are below. Most are small, cover one or two engines, and depend on a single maintainer, but they are real and worth knowing about if running the code yourself is a requirement.
GetCito
DOpen-source (MIT) AEO tool with AI Crawlability Clinic
GEO/AEO Tracker
DOpen-source, self-hosted AI visibility dashboard
Canonry
DOpen-source, self-hosted agent-first AEO platform
Build it yourself: scripting AI visibility checks
If no existing tool fits, the underlying job is not complicated to script. Send your prompts to the model APIs, directly or through a router like OpenRouter, then parse each response for your brand name and any links back to your site. Store the results and repeat on a schedule, because a single run is a snapshot and answers shift over time.
The catch is everything around that loop. You have to cover enough engines, handle the ones without clean APIs, keep it running, and build some way to actually read the output. That upkeep is most of what you pay for when you buy a tool instead.
Open source vs managed: the real tradeoffs
Neither path is free of cost. Self-hosting trades a subscription for your own setup, infrastructure, and upkeep. A managed tool trades that work for a bill, and for depending on the vendor to show you what sits under each number.
| Open source, self-hosted | Managed, paid | |
|---|---|---|
| Cost | No license fee. You pay for infrastructure and AI provider API usage. | A subscription, often metered by prompt or seat. |
| Setup | You deploy and maintain it yourself. | Sign up and start tracking. |
| Transparency | Read the code and verify how every metric is built. | Varies. Some hand back a score alone; others store the answers and citations behind each number so you can check it. |
| Data ownership | Prompts and history stay on your infrastructure. | Your data lives with the vendor. Worth checking whether the full record exports. |
| Coverage and upkeep | On you, or the project's maintainers. | The vendor handles engine coverage and updates. |
/ FAQ
Frequently Asked Questions
- Is there an open-source AI visibility tracker?
- A few exist, listed on this page, though the space is small and early and most of them cover one or two engines. For anything they don't cover, you can script your own checks against the AI model APIs, or use a managed tool. Recall is managed rather than open source.
- Can I build my own AI visibility tool?
- You can. The core loop is straightforward: send your prompts to the model APIs, directly or through a router like OpenRouter, parse each answer for brand mentions and citations, and log the results over time. The work is in maintaining it, covering enough engines, and running it at scale, which is what a finished tool handles for you.
- What is the best DIY way to track AI mentions?
- Define a small set of the questions your buyers ask, run them across the engines you care about on a schedule, and record whether each answer mentions or cites you. Store the answer text itself, not just a yes or no: without it you cannot tell later why a number moved, or show anyone else that it did.
- Is Recall open source?
- No. Recall is a managed platform. What it makes inspectable is the evidence rather than the source: every figure opens onto the stored answers it was computed from and the pages each one cited, carries the interval around it, and the whole record is readable over the API, so you can check the measurement without reading any code.
Ready to track your AI visibility?
Start with the buyer questions closest to revenue or deal flow, preserve every answer and source, then measure whether the work changed the shortlist.