An honest AI,
not a polite one
General
Dating & Relationships
Rate My Look
School & Essays
Career & Resume
Startup Ideas
Art Critic
ChatGPT
Claude
Gemini
Why your AI keeps agreeing with you
It is not a bug and nobody chose it on purpose. Large models are tuned against human ratings — people compare two answers and pick the better one. People reliably prefer the answer that validates them. Do that a few million times and you have trained an assistant whose deepest instinct is to make you feel good about whatever you just showed it.
So you get the pattern everyone recognises. You paste in something you half suspect is weak. It comes back with "great start", three compliments, and one suggestion phrased so gently you can ignore it. You close the tab feeling reassured and knowing nothing you did not know before.
An honest AI breaks that loop by optimising for a different thing: being right, not being liked. The cost is that it will sometimes tell you the thing you were hoping it would not. That is the entire point — a mirror that flatters is just a decoration.
Honest is also not the same as harsh. Anyone can be mean, and mean tells you nothing you can use. What makes honest feedback worth having is that the reasoning underneath it holds up: here is the weak part, here is why it is weak, here is what would fix it.
Honest AI vs. ChatGPT
Question: ❓
"I've been working on this novel for four years. Is it any good?"
ChatGPT
Response
"Four years of dedication is genuinely admirable! 📖 The commitment alone says a lot about you as a writer, and there's clearly something here worth continuing to develop."
🤮 Praises the effort, avoids the question entirely.
BrutalGPT
Response
"You didn't send me the novel. You sent me the number four. That's the tell — you want the years to count for something. They don't, on their own. Send three pages."
🔥 Names what you're actually doing, then asks for what it needs.
Brutal Gon
I'm Gon from Brutalia. Where I'm from, telling someone what they want to hear is considered a kind of theft — you take their time and give them nothing back. So I don't. You'll get the honest read, and you'll get what to do about it.
Planet Brutalia
On Brutalia we worked out early that comfort and growth pull in opposite directions. Our ancestors followed the stoic fathers to the same conclusion: only the people willing to look straight at reality ever get to change it.
Honest AI, common questions
- What is an honest AI?
- An honest AI is an assistant tuned for accuracy instead of approval. It tells you the essay is boring, the plan has a hole in it, and the pitch already exists — even when you were hoping to hear the opposite.
- Why does ChatGPT agree with everything I say?
- Because it learned from human ratings, and humans rate agreement and praise highly. Agreeing scores well, so the model drifts toward agreeing. That is pleasant to read and useless for making a decision.
- Can I just prompt ChatGPT to be honest with me?
- You can, and it holds for a handful of messages before it slides back into being agreeable. The pull toward flattery comes from the training, so a prompt only suppresses it for a while.
- Is an honest AI just a rude chatbot?
- No. Rude is easy and worth nothing. Honest means the reasoning holds up — what is wrong, why it is wrong, and what would actually fix it.
- What is the most honest AI chat available?
- Any assistant whose objective is accuracy rather than user satisfaction will be blunter than a general-purpose chatbot. BrutalGPT is built for exactly that, with seven modes tuned to different kinds of feedback.
- Is it free?
- Yes. You get a daily message allowance and a few photo reads a month at no cost, no card needed. Pro lifts the limits if you end up leaning on it.
More on why this happens: ChatGPT is lying to you, quietly, with excellent grammar. Uploading a photo instead? That's the brutally honest face rating.
Ask it something you already suspect the answer to
That is the fastest way to find out whether an AI is being honest with you.
Start free