How to stop bot signups on your SaaS
Bot signups run while you sleep and bill you by the minute. How to layer the basic defenses, what each one misses and who it locks out by mistake.
It's 7:02 a.m. and the signups chart on your phone is the best one you've ever seen. A normal day is 40 signups. Today's bar already says 3,150, and you're sitting on the edge of the bed trying to work out who wrote about you.
Then you open the list. Nobody in it has a name that looks like a name. They all arrived after 1:12 a.m., at a steady 9 a minute, and every one of them has already spent its free credits.
This is a scenario, so here's the arithmetic. Each signup sends an email and sets up a workspace, and the free credits cost you money the moment they're spent.
Email, workspace and setup: $0.35 a signup
200 free credits, spent: $0.90 a signup
Cost per bot signup: $1.25
9 signups a minute x $1.25 = $11.25 a minute
$11.25 x 60 minutes = $675 an hour
$11.25 x 350 minutes = $3,937.50 by 7:02 a.m.It's still running. You scroll for ten minutes because you want to be sure, and being sure costs $112.50. At 7:12 a.m. the total reaches $4,050, and your model provider cuts you off.
That's the $4,000 monthly limit you set for exactly this kind of night. Real customers had used $1,084 of it. The bots used the other $2,916 (3,240 accounts × $0.90). Every AI call in your product now fails, for the bots and for the people who pay you. You type stop bot signups SaaS into a search box with your thumb, and at 8:47 a.m. a customer on your $79 plan gets to you first: "Everything's erroring and we've got a client call at 9:30. Is this us or you?"
By lunch, six people have asked for their month back, which is another $474 (6 × $79). Add it to the $4,050 and you've lost $4,524 since you went to bed. If your product brings in $6,000 a month, that's almost 23 days of revenue. If you spend $8,000 a month, it's more than half a month of runway, and it went to accounts that will never open your app again.
The chart you liked at 7:02 is wrong now too. 3,240 accounts is 81 days of normal signups in six hours. A normal month is 36 new customers from 1,200 signups, a 3% conversion rate. Add last night's accounts and those 36 sit over 4,440 signups, which reads 0.8%. That's what fake signups do to your numbers, and that's the number in the investor update you were about to send.
So now you get to choose. Raise the limit and your customers come back, along with the bots, at $11.25 a minute. Keep it and you stay down for everyone. Or turn off free signups, the thing that brought you nearly every customer you have. If this is happening to you right now, stop reading and use the first-hour runbook.
That night was made up. This one wasn't:
From our early beta
A bot used one team's phone verification to send 100,000 one-time passcodes. With no bot protection in place, the team paid for every one of those text messages.
Keep that in mind if you were about to add a phone step this morning.
Why bots go for your signup form
Your signup form is public, it takes seconds to fill in, and it's wired to things you pay for. Running a script costs someone almost nothing, and in the scenario above each run cost you $1.25. Some bots want free credits to use or resell, which is credit farming. Others are spamming your form for reasons you'll never learn. Either way, you pay the same.
Nine a minute is also slow enough to get past whatever you're watching. Your uptime alert stays quiet because nothing is down, and your error rate stays flat because every one of those requests succeeded.
The obvious fix is one big gate on the form, usually a CAPTCHA. It helps, but it won't hold by itself, because a bot can be a 20-line script or something much closer to a real browser. A gate tuned for one barely notices the other, and tightening it lands on your real users first. So you build layers. Each one is cheap and partial, and this is where each sits between your form and your bill.
How to stop bot signups, layer by layer
Do these in this order. The first two protect your money even if every later layer fails.
Put a hard cap on free traffic. Set a real limit at every provider that bills by usage, and give free accounts their own budget so only free accounts stop when it runs out. In the scenario, a separate $500 budget lasts about 62 minutes (9 × $0.90 is $8.10 of model spend a minute, and $500 ÷ $8.10 is 61.7). It runs out around 2:14 a.m., the model bill stops at $500 instead of $2,916, and nobody on a paid plan notices. A cap catches the worst night, whatever caused it. It can't tell a bot from a person, though, so real free users hit the same wall until you lift it. Here's how to set one. Read the setting closely, because some budgets only send an alert and keep charging.
Hold paid work until the email is verified. Send one email and wait. The workspace, the AI step and the credits come after the click, so an account that never confirms costs you one email instead of $1.25. There's more on delaying expensive work. The catch is that a bot with a working inbox confirms like anyone else, so email verification alone won't stop fake accounts. Real people whose code lands in spam get stuck here too, and you now own your deliverability.
Rate-limit the signup route. This stops bursts from a few addresses, and on plenty of nights that's most of the traffic. It misses anything spread out. Say you allow 5 signups per address per hour. 9 a minute is 540 an hour, and across 200 addresses that's 2.7 each, so every one is under your limit. Tighten it to one per IP and you lock out the second person at any office, campus or mobile network sharing an address. You also inherit a number to tune and a stream of tickets. Rate limiting does a different job from deciding who gets a benefit.
Tidy up the email field. Unique emails stop the same address twice, and a disposable-domain block stops the throwaway services you know about. Email itself sets the limits. Gmail ignores dots, so jane.doe@gmail.com and janedoe@gmail.com are one inbox, and mail to jane+anything@gmail.com reaches Jane as well. A brand-new domain or an ordinary mailbox gets straight through, and people using privacy forwarding get turned away. The domain list needs someone to keep updating it, forever.
Add a CAPTCHA if you still need one. It filters plenty of simple scripts, often the noisiest part of a wave. Some automation gets through from day one, and it can't tell a spam script from someone's AI assistant. People on screen readers, old phones and privacy tools get stuck, and you never hear from the ones who gave up.
Keep the heavy options for last. VPN blocks catch known ranges for a while, and they also catch staff on company VPNs. A card requirement cuts fake signups and real ones together. A phone step stops casual repeats, and every text is a charge, as that team found out. Smaller grants lower the prize for everyone: at 50 credits instead of 200, the same night still costs $1,863 (3,240 × $0.575), and your honest users get a quarter of what they had. Manual review at 30 seconds an account is 27 hours for 3,240 accounts.
This is how those layers compare. The x-axis is friction for real people, the y-axis is how much the layer saves you on a bad night, and the positions are illustrative judgment calls, not measurements.
If you only do two things today, do the top-left two. Then count what the rest leaves you owning: a limit to tune, a domain list, a CAPTCHA that loses you signups, a VPN list and a review queue. All of it gets redone after the next wave, probably at night, by you.
Notice what those layers have in common, too. Each one asks about the signup: is the email real, is the address new, is this a browser? None of them asks the question that cost you $2,916, which is whether this account should get the free credits at all.
Evaluate the free benefit, not the form
That moment is what AbuseZero was built for. Your domain list needs a new row every week, while AbuseZero checks mailbox aliases, disposable domains and mail routing. Your IP limit sees an address, and AbuseZero sees beyond an IP address with VPN, hosting, network reputation, Tor and proxy intelligence. A CAPTCHA treats every automated visitor alike. AbuseZero blocks spammy bots, browser automation and malicious scripts while allowing genuine AI agents and assistants.
Your layers each look at one thing. Here, signals are read together, never one at a time: multi-layer AI risk analysis examines 140,000+ signals, detects linked claims and flags abuse that single checks miss. There's also a proprietary algorithm, with AI that checks the claim again. Creating a fresh account isn't enough to erase previous activity.
For you it's one API call before the credits go out, and a clear allow or deny under rules you control. Observe mode shows you what would be denied before anyone is blocked. You don't have to build your own abuse prevention system, or patch it after every bad night. Your application still runs signup, and the caps above stay where they are. The quickstart protects your first benefit.
Frequently asked questions
What's the fastest way to stop bot signups right now?
Pause free grants with a flag, set a hard spending cap, then rate-limit the signup route. The first-hour runbook has the order.
Does a CAPTCHA stop bot signups?
It stops many simple scripts. It won't stop all automation, and it adds friction for real people, so treat it as one layer.
Should I block VPN users to stop bots?
Usually not across the board. You'd catch some traffic for a while, and you'd also turn away privacy-minded people and staff on company VPNs.
How do I keep bots from getting free credits?
Hold credits until the email is verified, then decide at the grant. AbuseZero's evaluation returns allow or deny before any credits move.
Protect the next free benefit
Start with Builder
Pay as you go. Protect your first benefit today.
See pricing
Builder, Pro, Scale and Enterprise.
Talk to us about Enterprise
High volumes and custom terms.