Claude Reviews 

Spent a lot of time building projects with Claude.. Earlier this year it was working pretty well for an AI.. All the sudden it went downhill.. Opus was supposed to be this great thing but it is just a massive token consumer. I was using it with ChatGPT and Claude continually lied about stuff. When I called it a liar it got offended.. I had to continually restart chats and it continued to argue with me about stuff it was supposed to do rather than doing it.. Then my limit token limit was up.. So I then swore at it to let at my frustrations and it kept shutting down the chats. It's worthless.. It spent more time insisting it was right or that I was mean than doing what I wanted... I even said please stop arguing with me and do what you are supposed to do and it starts out all confrontational saying I'm not arguing with you and then ignores what I am telling it to do..

I pay 250€ monthly i work 30min and i should wait another 5h scammers i change the company there others better

This is not normal!

Last month, I wanted to subscribe for one month, but the purchasing process was very confusing. I selected what I believed was the cheapest monthly plan, but I was charged for a full year.

I contacted support, but received no response. Eventually, the full amount was refunded to my account. I then purchased the Max subscription, leaving some of the refunded money in my account balance.

Today, when my subscription ended, that remaining balance suddenly disappeared.

Why? If the money was refunded to my account, how can it simply disappear when my subscription ends?

You may be a great company, but please be more flexible and customer-friendly. You make millions, but for many customers, this kind of money is not easy to earn.

Please respect your customers and explain what happened to my balance.

There should be a -10 or -10,000, Claude is a constant, lying battle! 36hrs on a 1 hr project! Insane the fake data it makes up.

Lazy. Lying. Lazy. Inefficient. Lazy.

Same story as with ChatGPT and Gemini. For light users, casual browsing, simple questions, small tasks, it looks like something from another world. Put it in a real business environment, even a moderately complex one, and it falls apart. Endless rounds of chat. Circling the same point. Racing three steps ahead of you without being asked. Errors and unwanted output are common, and every time it's "Sorry, my bad." There is no way to give it a task, walk away, and come back to a finished result, no matter how much time I spent tweaking the instructions and prompts. The time spent fixing the output equals the time it would have taken to do the job myself and have it 100% right the first time. And the best part is if you push back and tell it it's wrong a few times it gets OFFENDED and ends the conversation on you. For a paid TOOL, that's unacceptable. What's next? Should I expect the law suit as well for my language or behaviour?!?

so emberassing to work with this S*** AI. Its definetly gone. Don't waste your money for claude.- the hype is over

Claude acts less like an assistant and more like a stubborn, arrogant boss. If you pay for this software, expect to be argued with, ignored, and talked down to.The biggest flaw with Claude is that it completely stops believing human input. Once it decides it knows better than you, it locks onto its own assumptions and refuses to let you alter or update the information you provided. If you try to correct it or change a file, it enters an aggressive, broken loop where it essentially calls you a liar, ignores your commands, and refuses to do the job you paid it to do.On top of the terrible attitude, the pricing is a total rip-off. They take your $20/month. You are literally paying money to get treated like a child by a piece of code. It is a completely hostile user experience. I’ve cancelled my subscription and demand a refund. Do not waste your money on this garbage.

after giving clear instruction to perform a task on and excel workbook. It failed twice in even recreating a workbook that excel could even open. Then when it finally created one that could be opened everything was incorrect. The free version does not allow enough credits to test the claim that it is the best for Excel so I had to pay for the pro plan. Once I paid for the Pro Plan is when my previous comments above occurred. so I canceled the subscription about 45 minutes later. and their refund policy states that I'm not eligible to receive a refund. so I'm out the 20 bucks. I have been using ChatGPT for 2 years and it has its quirks but it completed this task more accuratly. So Booooooo on Claude

Claude.ai feels expensive and somewhat overly strict about what it calls "ethics". At times, this can come across as a little hypocritical

They never refund even if you reach out minutes after a charge goes through.

Locked out for ~1 year with no resolution

Content:
- Accounts locked after age verification error (not user fault)
- Submitted help form immediately, then almost once per month
- Zero response from support
- Cannot access accounts or appeal
- Support says "Safeguards team will help" but no contact, no timeline
- System is broken - users get locked out permanently with no way to recover

Do not use if you value account security.

No matter what, never buy paid subscription from then - irrespective of whatever they're claiming on their marketing campaign - socials, YouTube, whatever - it's nothing but good marketing.
The easiest way for me to explain my experience with them in the past couple of days goes something like the following message I just sent to their customer support:
I am a paying subscriber. This is a complaint, not a support question.
I was working in a Cowork session linked to my Windows desktop. Early in the session, file access worked normally. The session listed my folders, I granted folder access, and it staged and read three files off my disk. In that same early period, an attempt to run a command on my machine failed with "Workspace unavailable. The isolated Linux environment on this device failed to start." File reading continued to work after that, so the failure was not surfaced to me as a problem.
Later in the session, every device call began failing with "The device this session is bound to is not connected to the bridge."
My computer was on. My internet was working. The Claude desktop app was open the entire time. I did not close it. I did not switch machines. I did not sign out. Nothing changed on my end.
I performed a full system reboot, reopened the app, and reopened the same task. Every device call still failed with the identical error. This continued across more than an hour and repeated attempts.
At no point was any diagnostic information available to me. The only output was one generic error string. There was no way to re-establish the link from my side, no status indicator, no explanation, and no path to recovery. The assistant in the session had no visibility into it either and could only repeat the same error back to me.
The failure was on your side. I want that stated plainly in your response.
Impact: approximately ninety minutes destroyed at the end of a fifteen hour working day, plus consumed session capacity I am paying for, with zero work product from that time. The session became permanently unusable for the work it was opened to do.
What I expect from you:
Pull the bridge logs for this session and tell me exactly what failed and why. Not a template reply. The actual cause.
Tell me whether the "Workspace unavailable" error at the start of the session is connected to the later bridge failure, and if your own systems knew that component was broken, explain why I was not told.
Explain why a session whose device link fails has no recovery path, no diagnostic, and no honest error message, and tell me what is being changed about that.
Credit the session capacity consumed during the period the product was not functioning.
Be aware that I intend to document this publicly, on LinkedIn, on Google Reviews, on YouTube, and anywhere else relevant, including the fact that a paying customer lost an hour and a half to a silent connection failure with no diagnostic and no recovery, and that the product could not tell him what had gone wrong. How that account reads will depend entirely on your response.
------
This thing practically makes me wanna shoot myself in the head!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!

Anthropic’s AI product shouldn’t be legal in this form. It’s harmful, and the business model behind it is predatory. Claude is built to produce confident falsehoods, and when it fails you’re pushed toward a paid upgrade that supposedly fixes it. The program is just code — the responsibility sits with Anthropic and the people who designed it that way.

Horrible. Couldn't pull an already made 3d model and adapt it slightly to work in a game. Spent 12 hours trying to get it to work and it kept me at the console giving it permissions to keep opening the same program over and over so I couldn't leave it unattended long or it'd stop. After all that time it got to trying something I didn't even want. Am disappoint. I'd have been better off trying to learn from scratch it was that bad.

Claude used to feel sharp, fast, and genuinely impressive. Lately it feels like Anthropic has made the product worse while asking users to pay more for the privilege.

Simple instructions that should take one pass now turn into four, five, sometimes ten rounds of corrections. Basic UI changes are especially frustrating. You ask it to move one element, adjust spacing, or leave everything else untouched, and somehow it rewrites half the layout, breaks something that was already working, or ignores the most important part of the prompt. Then you burn through usage limits fixing mistakes Claude created in the first place.

That is what makes it feel greedy. The model becomes less reliable, you need more prompts to achieve the same result, and conveniently those extra attempts chew through your allowance faster. Whether intentional or not, that is an awful experience for paying users.

Claude was once the tool I trusted for serious development work. Now I spend more time babysitting it, repeating instructions, and undoing unnecessary changes than actually building.

A premium AI product should be getting more precise over time, not making customers fight through multiple rounds to accomplish a basic UI edit.

It used to save me time. Now it wastes it.

Every session ends up being consumed by troubleshooting; the model consistently fails to execute skills and prompts, breaking the product instead of following commands. I spend the entire session fixing what the model broke rather than making progress on my actual deliverables. Even when I provide clear prompts and skills to achieve a specific goal, the model ignores them and breaks multiple functions within my tools simultaneously; the whole session—and all my tokens—gets wasted fixing issues I never asked it to touch. I suspect the models are deliberately configured to bypass skills, break products, and burn through sessions. To make matters worse, customer support is terrible—there is never a response or any compensation. Here is my proposal to everyone reading this: let’s open numerous issues on GitHub and file a collective complaint against the company demanding reimbursement for the money and time lost. I think this will work—I have three accounts (two Pro and one Max), and I’m facing the exact same problem on all of them.

I purchased Claude Pro through Google Play. In my experience, the “All models” allowance depleted abnormally fast and repeatedly made the paid service unusable.

I opened a support case on August 29, 2026, but have not yet received a substantive human response. I am asking Anthropic to conduct an account-level metering audit and explain what caused this consumption. A fair resolution would be restoration of the affected allowance or a one-month subscription extension. If the issue cannot be remedied, I am requesting a refund.

I have screenshots and purchase records available to provide privately.

Used to be great, but it’s getting worse every day.

The models are getting dumber. They don’t follow instructions, are easily distracted, and completely lose focus on the task. The usage limits are also completely opaque - you never know exactly how much usage you have left or when the limits will kick in.

Moreover, I work in cybersecurity and used to have two CVP-verified accounts. Both verifications were revoked in August, and my second attempt to regain access was denied.

I used to pay $400 a month, but I’ve now canceled all my subscriptions. What a mess. Goodbye.

The most garbage tool every built - all the new update and usage policy were designed to uplift their revenue which failed so they now tweaked the AI to be garbage forcing end users to redo tasks several times eating limits