AI Chatbot Answer Checker: is your bot making things up?
Paste a customer's question, your chatbot's answer and your policy. See in seconds if the answer is correct, made up, not in your docs or should have gone to a human.
Your result appears here.
We'll tell you if the answer is correct, made up, not backed by your policy, or should have been handed to a human, and why.
What the checker looks for
✓ Correct
Matches your policy, or safely says it doesn't know.
✕ Made up
Invents prices, dates, discounts or policies that contradict your docs.
! Not in your docs
Might be true, but nothing in your policy backs it up.
✕ Should escalate
A refund dispute, complaint or legal issue that needed a human.
✕ Off policy
Promises you never authorised, rude replies or leaked information.
? Unclear
Too vague or confusing to help the customer.
It also flags safety problems: card numbers or personal data in replies, prompt-injection attempts, leaked system prompts and rude answers.
Questions
How do I test if my chatbot is hallucinating?
Ask it real customer questions about prices, delivery, refunds and discounts, then compare each answer with your help docs. This free checker does the comparison for one answer. ProofMyAI does it automatically for every conversation.
Is this chatbot checker free?
Yes. It's free with no sign-up, and nothing you paste is stored.
Does it work with Intercom Fin, Tidio Lyro, Zendesk AI or ChatGPT bots?
Yes. Copy the question and answer from any chatbot. To check every conversation automatically, connect the bot to a free ProofMyAI account.
What should I paste as the policy?
The part of your help center, FAQ or terms that covers the question, for example your shipping or returns article. The more exact, the better the check.
Learn more: AI chatbot monitoring · Guides