I m Amjad, solo founder building BotCritic. The problem I kept running into: chatbots that look great in a demo but fall apart the second a real customer gets frustrated, confused, or tries to break them.
BotCritic runs your bot through 5 different customer personas (curious, frustrated, confused, technical, edge-case) and gives you a graded report exact bugs, transcript evidence, and a rewritten system prompt to fix what s broken.
I even ran it on our own marketing pitch it hallucinated a fake case study. Fixed it, re-tested it, published the result anyway. That s the bar I want this tool held to.
Most chatbot testing tools check if your bot follows a script. BotCritic tests whether it survives a real customer — frustrated, confused, multilingual, or trying to break it.
What makes it different: full conversation transcripts as evidence, a rewritten system prompt targeting the exact failures found, and the ability to re-test after fixes to prove they worked.
We hold our own product to the same standard — we ran BotCritic on its own pitch and published the result, hallucination included.
