Prevent fake accounts consuming credits

If your goal is to prevent fake accounts consuming credits, you are protecting unit economics, not chasing bots for their own sake. Fake accounts pass as users long enough to spend free compute, tokens, or seats, then disappear. Rate limits per account do not help when each fake account is new.

The durable fix is to decide before the grant and learn after. Allow trustworthy signups, challenge the uncertain middle, deny clear farms, and report credit burn or conversion against the original decision id.

Why "fake" is the wrong unit alone

Many abusive accounts are "real": real emails, real CAPTCHA solves, real-looking devices. The tell is economic: they consume and never convert. Scoring only "is this a bot?" misses human-operated farms and agent farms that pass bot checks.

Prefer questions like: does this signup deserve this grant, and what happened to the last hundred like it?

A stack that protects credits

Chitmark is that verify → feedback → challenge control plane. Use the economics calculator for free-tier burn, the compare page for CAPTCHA and Radar, and the trial farming guide for the harvest pattern.

• Verify before minting free credits or issuing an API key
• Persist the event id on the account row
• Complete challenge friction when asked (Turnstile, step-up, card hold)
• Send feedback when you see burn, conversion, chargeback, or ban
• Keep rate limits as a backstop after the grant, not the primary gate