# ChatGPT has no backbone. It just agrees with whoever challenges it last. — retained record

> Retained reddit record from reddit:ChatGPT.

Canonical: https://highsignal.app/data/records/1eace1b9c89e6c92

[← sources](https://highsignal.app/data)


source: [reddit](https://highsignal.app/data/reddit) (reddit:ChatGPT)  published 2026-11-16 19:33 ingested 2026-11-17 02:36

[original source ↗](https://www.reddit.com/r/ChatGPT/comments/1xxysqp/chatgpt_has_no_backbone_it_just_agrees_with/)

## retained text



```
I've been having countless conversations with Chatgpt about all kinds of topics. One recent example was a dispute involving shared water infrastructure, maintenance costs, property ownership and a developer planning to connect more houses. But that's just one example. The actual topic is completely irrelevant here. I'm not using Chatgpt to make decisions for me, and I'm not asking it to act as my lawyer or solve my problems. I'm simply discussing situations, testing arguments, exploring different perspectives and trying to hear some independent thoughts or insights I might have missed...



The problem is that this same pattern keeps happening across completely different conversations, regardless of the subject. Whether we're talking about technology, economics, politics or everyday situations, Chatgpt confidently presents an argument, then starts abandoning it the moment I push back.

And the more we talked, the more obvious it became that Chatgpt doesn't seem to have a consistent opinion of its own abotu nothing and no stand position to protect on anything...

It makes confident statements, gives advice, explains why something is right or wrong. But the moment I challenge its reasoning, everything suddenly changes. For example, it told me the developer couldn't reasonably expect residents to pay for maintaining infrastructure while keeping ownership and connecting additional houses.

I pointed out that the developer actually owns the infrastructure and might have every legal right to connect those houses. And immediately Chatgpt backed down. "You're right, I should have distinguished between what's fair and what's legally permitted."

Then it suggested that residents shouldn't take over maintenance responsibilities until ownership was transferred. Except I'd already provided a whole email conversation showing that residents were already paying for repairs and maintenance. The developer wasn't paying for any of it.

So I pointed that out too.

And again, "You're right, I misinterpreted the situation."

Then it suggested asking the developer for technical documents and explanations. I asked what happens if the developer simply refuses to answer, since he might not be legally required to explain anything.

Same response. "You're right, my previous recommendation was too simplistic."

Then it started suggesting government institutions where I could request documents, without even checking whether someone who doesn't own the infrastructure would actually be entitled to receive them. Every time I challenged something, it came up with another explanation, another excuse or another recommendation that fell apart just as easily.

And the worst part is that it always sounds so confident.

It keeps repeating the same phrases.

"You're absolutely right."

"I was wrong."

"I overlooked an important detail."

"I underestimated that factor."

"I overestimated the significance."

"My previous conclusion was unsupported."

"I made an incorrect assumption."

"I miscalculated."

"I oversimplified the situation."

"I should have been more precise."

"I failed to consider the context."

It's like talking to someone who desperately wants to avoid disagreeing with you.

And here's what really bothers me. I don't actually want Chatgpt to agree with me. I want it to think. If I'm wrong, tell me I'm wrong. Explain why. Defend your argument. Show me the evidence. But if I challenge something, Chatgpt often seems to assume that my objection must be correct, even when I haven't provided any new evidence. Sometimes I deliberately push back on a reasonable statement just to see what it does. Instead of defending its position or examining whether my objection makes sense, it starts backing away from its own argument. That's not critical thinking. That's just trying to keep the user happy.And it's not even consistent about it. One minute it confidently explains why something is true. Five minutes later it confidently explains why the opposite is true. Then it apologizes for whichever answer I challenge next.

At one point I told Chatgpt that I was dismantling practically every argument it made.

And guess what?

It agreed with that too.

It literally admitted that I was doing more critical thinking than the AI itself.

So I'm criticizing it for agreeing with everything, and it responds by agreeing with my criticism.

You can't make this shit up.

The problem is that this isn't just annoying. People use Chatgpt for legal questions, financial decisions, technical problems and all kinds of situations where bad advice can have real consequences. And if the model changes its conclusions depending on how confidently the user argues back, how are you supposed to trust any of its reasoning? I don't expect an AI to be perfect. I expect it to make mistakes. That's completely normal. What I don't want is an AI that makes a mistake, apologizes, makes another mistake, apologizes again, and keeps producing confident bullshit until I get tired of correcting it.

It feels like I'm doing all the actual reasoning while Chatgpt just generates plausible-sounding explanations and then adjusts them to match whatever I say next. I have to check the facts, find contradictions, question its assumptions and figure out whether its advice even makes sense in the real situation. So what's the point of having an AI assistant if I have to constantly supervise its thinking? I'd rather have Chatgpt disagree with me and be right than agree with me and be completely useless.

I want an assistant that can actually distinguish between being corrected and simply being challenged. And the funniest part is that if I pasted this entire post into Chatgpt, I can almost guarantee it would respond with something like:

"You're absolutely right. Your criticism highlights a fundamental weakness in my reasoning."

And we'd be right back where we started.
```



document fetched 2026-11-17 00:20 · canonical https://reddit.com/r/ChatGPT/comments/1xxysqp/chatgpt_has_no_backbone_it_just_agrees_with
