How to stop free message abuse in an AI chatbot builder
A free chatbot plan can end up running a real business's support desk. What each message costs, why the basic checks break and how to limit the damage.
The free account on your screen has a better conversations tab than most of your paying customers. Its bot lives on the website of a bike repair shop, and today it told one person that a cracked carbon frame needs an in-person look and another that the shop closes at 2 on Saturdays. The usage panel above those chats says 97 of 100 free messages, which looks like a healthy free user on the 20th of the month. Then you notice the graph under it. It's flat until the 17th, then it climbs about 30 messages a day.
While you're looking at that, a ticket comes in from a different free account, with a different email address and the same shop's name in the signature: "Bot number 6 is nearly out for the month. Can someone help me get number 7 working on our site? The last one took me ages."
That's a scenario, and every number here is an assumption to replace with your own. Start with one reply, because each one carries the shop's help pages and the chat so far:
- Input: 4,000 tokens × $3 per million = $0.012
- Output: 400 tokens × $15 per million = $0.006
- Searching the help pages, plus hosting: $0.002
- Total: $0.02 per message
At 30 messages a day it needs 900 a month. Your free plan gives 100, so that's nine free accounts (900 ÷ 100) taking turns, and every one refills on the 1st. You pay $18 a month to answer that shop's customers (900 × $0.02). Your Basic plan is $59. The shop has paid you $0, this month and every month before it.
One shop is a rounding error. This is what chatbot builder free message abuse looks like when 140 businesses do it:
- 140 businesses × 9 accounts = 1,260 free accounts
- 140 × 900 messages = 126,000 messages a month
- 126,000 × $0.02 = $2,520 a month
- $2,520 × 12 = $30,240 a year
Here's your month with 220 paying customers averaging 1,100 messages and 3,000 real free accounts averaging 30.
Businesses paying you nothing cost more than every real free user combined, and that money is gone before you've earned anything to cover it. Your model provider charges for each reply whether or not anyone paid you for it. Close all 1,260 accounts tomorrow and you won't get a cent of this month's $2,520 back, because the replies have already been sent.
A paying customer at 1,100 messages costs you $22, so you keep $37 of their $59. It takes 69 of those customers to cover $2,520 ($2,520 ÷ $37 = 68.1), which is nearly a third of the 220 you have. If your company spends $22,000 a month, $30,240 a year is about 1.4 months of runway.
Those free bots also work the hours your paying bots do. A bike shop's customers ask about repairs at lunch on a Monday, and 126,000 of your 458,000 monthly messages (242,000 paying, 90,000 real free) means more than a quarter of the load on your shared model quota pays you nothing. The Monday after a long weekend you hit that quota around noon, and for 40 minutes the bots on paying customers' sites reply "Something went wrong." A dental practice on your Basic plan writes in: a patient with a broken tooth got that message on their website. Three customers cancel that week, and $177 a month (3 × $59) leaves with them.
So now you've got a decision. Cut the free plan and you punish everyone testing a bot for the first time. Remove it and you lose the "Powered by" link on every free bot, the one that sends you new signups. Leave it alone and next month's bill has the same $2,520 in it.
Why free messages get abused in a chatbot builder
Most free plans hand over a sample. A chatbot builder's free plan hands over the finished product. Once the bot has read the shop's pages, it does a part-timer's job on the real website, all month, and that's worth far more to the shop than $59.
You can't see it in the traffic either. Every message is a real visitor typing a real question, and every account is under its limit. An account waiting for its turn looks like someone who signed up and lost interest. Some of the 140 won't even be shops. An agency running bots for 30 local clients at $150 each takes in $4,500 a month (30 × $150), and you're paying for its messages.
One shop's month, illustrative, at 30 messages a day:
The obvious fix is a smaller allowance, and it backfires. At 25 free messages the shop needs 36 accounts instead of nine (900 ÷ 25), while someone testing a bot on their own docs runs out that first afternoon. There are better ways to set free limits.
How to lower what a free chatbot message costs you
These shrink the damage even when an extra account gets through.
Count messages on the server, at the model call. The meter in the widget is only a display.
Give free bots their own budget and queue. A separate quota and a hard spending cap mean a busy Monday on the free plan can't silence a paying customer's bot.
Make a free message cheaper. Use a smaller model, fewer help-page passages, less chat history and shorter replies. If $0.02 falls to $0.008 (illustrative), the same 126,000 messages cost $1,008 instead of $2,520.
Put business features on paid plans. Human handoff, lead capture and transcript export are what a live business needs and a tester doesn't.
Talk to your busiest free accounts. A business using every message is a sales lead, and a smaller plan may win it. Write down what one account means and how you'll say no.
Those steps make each free message cheaper. None of them stops account number seven from opening, so most teams reach for the basics next.
The basic checks, and how each one breaks in a chatbot builder
A unique-email rule compares text, and one mailbox can be written several ways. Gmail ignores dots, and plus-tagged mail reaches the same inbox. That's simply how email works.
| Basic fix | What it catches | What it misses | Who it wrongly blocks | What it costs to keep working |
|---|---|---|---|---|
| Email verification | Typos and dead addresses | A shop with several real inboxes | Owners whose code lands in spam | Sending and deliverability work |
| Unique emails | The same address twice | Every different address | Almost nobody | Very little |
| Disposable-domain blocks | Known throwaway services | New domains and ordinary mailboxes | People using privacy forwarding | A list someone updates forever |
| Gmail dots and plus addressing | One inbox written several ways | A second real mailbox | Almost nobody | Upkeep as mail services change |
| CAPTCHA | Scripted signups | A shop owner, who's a person | People using screen readers or privacy tools | Fees and lost signups |
| Phone requirement | Casual second accounts | Staff phones and spare numbers | Businesses with only a landline | A charge for every text |
| Card requirement | Casual second accounts | Anyone with a second card | People who want to test before paying | Fewer signups |
| One per IP address | Repeat signups on one connection | Home, shop and mobile connections | Agencies and coworking spaces | Exceptions and tickets |
| VPN blocks | Known VPN ranges | Everything off your list | Staff on company VPNs | Another list to maintain |
| Rate limits on the widget | A script hammering one bot | 30 polite messages a day | A real customer's busy hour | Tuning per bot |
| Message caps | One account going over | Each new account's fresh 100 | Anyone at the cap | Deciding when a bot goes silent |
| Smaller allowance | Some cost per account | More accounts make up the difference | Everyone testing a bot | Fewer people reach paying |
| Manual review | What a person can spot | Accounts spread over months | Whoever looks odd to a tired reviewer | Hours, plus reading other companies' chats |
A CAPTCHA never had much of a chance here, since the shop owner is a person, and one on the chat widget would make your customer's customers solve a puzzle to ask about Saturday hours. At three minutes each, reviewing 1,260 accounts is 63 hours (1,260 × 3 ÷ 60), due again next month, and you'd be reading conversations between another business and its own customers.
What you'd really like to know is whether another 100 free messages should open at all, and nothing on that list can tell you. The ticket on the 20th only gave it away because someone was kind enough to say so. Stack every row and the owner with nine inboxes gets through on a quiet afternoon, while someone who just wanted to test a bot closes the tab. Blocking real people never shows up on an invoice.
Evaluate each claim on your free messages
Every row in that table looks at a signup or counts inside one account. Your money moves later, when another 100 free messages are about to open, and that's when you need to know whether this claim should go through.
You don't have to build your own abuse prevention system. AbuseZero is made for that moment, and with it, creating a fresh account isn't enough to erase previous activity. It recognizes repeat devices across new accounts, even when email addresses keep changing. Your email rules become its job, because it checks mailbox aliases and disposable domains. Your IP limit and VPN list do too, because it sees beyond an IP address with VPN, hosting, network reputation, Tor and proxy intelligence. Where each row of that table looks at one thing, multi-layer AI risk analysis examines 140,000+ signals, detects linked claims and flags abuse that single checks miss. No list you maintain has a proprietary algorithm, with AI that checks the claim again. AbuseZero has both.
Your server makes one API call before it opens a free allowance, and again before each monthly refill, and gets a clear allow or deny under rules you control. Eligibility can be once per user or once per calendar month. Observe mode shows you what would be denied before anyone is blocked, and the quickstart walks you through your first benefit. Your app still counts the messages.
Frequently asked questions
What is free message abuse in a chatbot builder?
It's one business or person running several free accounts so their bot keeps answering past the free limit.
Should free chatbots be allowed on live websites?
Usually yes, because testing on a real site is what sells the product. Keep the "Powered by" link, hold back the business features and give free bots their own budget.
Is every business with two accounts abusing the free plan?
No. An agency with paying clients or a company with two brands may have good reasons. Decide what you meant to give each one, then let an AbuseZero evaluation apply that rule at each claim.
Keep your free bots for the people testing them
Start with Builder
Pay as you go. Evaluate your first claim today.
See pricing
Builder, Pro, Scale and Enterprise.
Talk to us about Enterprise
High volumes and custom terms.