Risk Scoring

reCAPTCHA scores each form submission on how likely it is to be a bot, without asking the visitor to solve a puzzle. A Beam agent reads that score as each submission comes in and, under an approved threshold, either lets it through to the next step, such as a signup or contact form handler, or holds it for a closer look. The agent does not change the score itself. Submissions that sit right at the threshold, or come from a form the account has not set a threshold for yet, are routed to a person to clear or block by hand.

Checkbox Verification

Some forms use reCAPTCHA's checkbox or image challenge instead of a silent score, asking the visitor to confirm they are human before submitting. A Beam agent reads the pass or fail result attached to each submission and, under an approved rule, forwards passed entries to the linked system, such as a spreadsheet or a support queue, while blocked ones are dropped. It does not present the challenge itself, only reads the outcome. Submissions with no verification result recorded, which can happen if the challenge was skipped, are routed to a person to check manually.

Abuse Pattern Detection

reCAPTCHA flags patterns across submissions, such as the same visitor hitting a form repeatedly in a short window, that a single score might miss. A Beam agent reads these flags alongside the risk score and, under an approved rule, groups repeat offenders or holds a form temporarily if abuse spikes past an agreed limit. It reports the pattern rather than deciding policy. A sustained spike, or any pattern that looks like it could affect a real customer instead of a bot, is routed to a person to investigate before the form is limited further.