Red flags when choosing an AI visibility tool

If you cannot verify how a visibility score is calculated, you cannot tell a movement from a formula change. Arbor publishes the four definitions on the homepage and in src/lib/visibility/scoring.ts.

Prompt history has to be recoverable. On Arbor cloud, cancelling the account includes a JSON export of run history within 30 days, then deletion. On self-host, the history never leaves your database. Cancellation of a self-host licence stops new releases. It does not delete the free core.

One model misses what another assistant tells the same buyer. Arbor runs Claude and GPT today. Perplexity, Gemini, and Copilot are on the roadmap and are not monitored yet.

Competitor presence should name the rivals in the answer, not a blended index. Arbor stores the raw answer and marks which rivals were named.

Scoring is source-available under the Elastic License 2.0. Cloud runs are stored for your account and can be exported. Self-host uses your API keys and your database. Coverage is two engines today, stated as such.

Read the terms and the comparison framework.