Does it catch LLM-generated spam?

Yes — the heuristic specifically targets patterns from LLM-generated form spam: repetitive phrasing, dense URL lists, and prompt-injection strings.

Spam score

Glossary

A 0–100 heuristic + AI score applied to every submission, where higher = more likely spam.

Definition

FormLoom's spam scorer assigns every submission a 0–100 spam score by combining a deterministic heuristic (URL density, keyword blocklist, entropy, character repetition, injection patterns) with an optional LLM classifier. The heuristic runs with zero API keys; the LLM classifier activates when an Anthropic API key is present. Submissions above a threshold are flagged as spam in the dashboard but stored (not deleted) so you can review false positives.

Why it matters for form backends

Spam scoring lets you apply a CAPTCHA-free spam wall that catches both bot-generated and LLM-generated junk.

FAQ

Submissions scoring above 70 are auto-flagged. You can adjust the threshold in the form's settings.

Related terms

Honeypot CAPTCHA Form endpoint

Ready to put this into practice?

Add a form to your site in minutes — free access key, copy-paste snippet.

Open snippet generator All terms