Why your reCAPTCHA v3 token gets a low score

Google documents reCAPTCHA v3 as returning a score from 0.0 to 1.0 and an action, which the website evaluates on its backend. A high score on a demo establishes the result for that demo. It doesn't establish the score or acceptance of a token submitted to another site or action.

If the target exposes assessment details, inspect token validity, the action, and relevant reasons alongside the score. Enterprise assessments contain risk information that the website interprets for its own workflow; there's no universal acceptance threshold. If all you can observe is a rejection, treat it as a rejection—not a measured low score.

solvethree's documented reCAPTCHA request has no parameter that forces a minimum score. Check the integration inputs instead of looking for a magic score setting.

The solve uses an assumed product or action.

Compare the task's site and action with the current target flow. For solvethree's reCAPTCHA v3 Enterprise flow, recaptchaType defaults to enterprise-v3; the gateway also accepts v3.

Try this: Use the hostname and exact action from the protected interaction. If the action is uncertain, follow Find the reCAPTCHA action before solving again.

The page or its reCAPTCHA setup has changed.

Compare the current hostname, action and reCAPTCHA type with the profile configured in your workspace.

Try this: Update the profile for the actual page before running another batch. Keep the task inputs matched to that profile.

Solving and submission use different proxies.

Compare the successful solving proxy with the proxy used for the application request. If you supplied proxyCandidates, don't assume the first candidate was the successful one.

Try this: Submit through the successful proxy. With a candidate list, use the returned winning index or the SDK helper to select it.

Your application keeps its own browser identity after solving.

Inspect the outgoing User-Agent and client hints rather than just the values stored in your application.

Try this: Use the returned userAgent and secChUa with the successful proxy when submitting to the intended target. Don't combine the token with an unrelated browser identity or target URL.