Skip to main content
This page contains recommendations for getting the most out of Roundtable Proof-of-Human. This will help you deploy Proof-of-Human effectively, calibrate your risk thresholds, and maintain your detection pipeline over time.

​
Deploy the tracker on as many pages as possible

Roundtable Proof-of-Human is designed to run on every page of your application. The more pages you track, the more behavioral data is collected, and the more accurate your risk scores become. Unlike traditional CAPTCHAs that create a single checkpoint, continuous cross-page tracking makes it much harder for bots to game your system by temporarily “acting human”. Pages with text input fields are especially valuable — typing is one of the best signals for distinguishing humans from bots and AI agents.

​
Gather data before gating

Place the tracker as early as possible in the user journey and check risk scores as late as possible, ideally checking at the moment of a critical action like a survey submission, account creation, or payment. This gives Roundtable the maximum amount of behavioral data to work with before you need to make a decision. The longer a session runs before you call the API, the more confident the risk score will be. As a starting point, we recommend the following thresholds:
  • 70 and above: Auto-block or reject the action. This range has a high likelihood of bot or fraudulent activity.
  • 50–70: Flag for manual review or require additional verification (e.g., email confirmation, a secondary check).
  • Below 50: Allow the action to proceed.

​
Adjust thresholds based on sensitivity

These defaults work well for most use cases, but you should adjust them based on your tolerance for false positives and false negatives. High-sensitivity workflows (such as highly paid surveys, financial transactions, or password resets) should use lower thresholds, while lower-stakes actions can afford higher ones.

​
Start in monitor-only mode and set initial thresholds

Before enforcing any automated blocking rules, we recommend running Roundtable in observation mode for a period. This just means adding the tracker to your site or project and letting data accumulate without acting on it. You can then review sessions in the Dashboard or retrieve them via the API to calibrate your thresholds against real traffic.

​
Adjust thresholds over time

Your thresholds and rules should evolve as you learn more about your traffic patterns. Regularly review blocking rates and false positive reports in the dashboard, and adjust your thresholds accordingly. What works at launch may need tuning as your user base or threat landscape changes.