How to compare AI visibility monitoring tools

Run the same buyer-intent prompts on each candidate and compare citation rate, company mentions, and rival mentions. Arbor sends those prompts to Claude and GPT and keeps the raw answer next to the score.

Ask whether you can inspect the algorithm. Arbor's four scores are defined in the repository. A vendor index with no formula is not comparable to a percentage of prompts.

Ask who holds the prompt logs. Arbor cloud stores them for the account and exports them as JSON if you cancel. Self-host stores them in your database. Postgres is the default. SQLite is a local configuration, not the hosted product.

Count the models that are called today, not the ones on a roadmap slide. Arbor calls Claude and GPT. Perplexity, Gemini, and Copilot are listed as later work and are not in the run queue yet.

The source shows the scoring and the licence. The hosted service is not free: Starter is £39/mo after a 14-day trial with no card. Self-host of the core is free under the Elastic License 2.0. Premium self-host features need a licence key. That is the trade to weigh, not a perfect scorecard.

The homepage keeps a side-by-side with a typical vendor.