Anthropic Reviews 229

I subscribe to Claude Pro for $20/month. I start using it to debug a problem I’m having with the Eclipse integrated development environment. About 45 minutes into my first paid session with Claude Pro and a message pops up telling me I’ve exceeded my usage limit and to try again in 4 hours.

I immediately canceled my subscription.

I'm a paying Claude Pro subscriber. Claude consistently fails to do what Anthropic explicitly promises it can do.

According to Anthropic's own documentation: "Claude has tools to search past conversations to find relevant context and information...Use these tools when the user references past conversations or when context from previous discussions would improve the response."

The task: Read a previous conversation and tell me what sample script was produced in that conversation.

What actually happened:
• Failed to use conversation_search and recent_chats tools despite having them available
• Made me repeat the same instruction in different ways FIVE times
• Even after I uploaded a SCREENSHOT showing the exact sample script in the last conversation, Claude STILL claimed the sample script did not exist.
• Gave excuses about "not having access" when Anthropic's own documentation confirms these tools exist for exactly this purpose

This isn't a feature limitation. This is downright a failure of the tech AND a failure of marketing promises. Anthropic explicitly built and advertises conversation search tools. Claude has them. It just refused to use them properly.

For paying subscribers: Anthropic promises Claude can "search past conversations". It can't, or won't, do this reliably. I totally regret paying for annual service. This is NOT THE ONLY FAILURE but I don't have the bandwidth to document every disappointing event.

Completely unacceptable.

Well, the AI was good to use but frankly in my opinion the company is awful a rip-off.
I have an account that I registered using my work address instead of my personal address. My company then blocked us using the work address for access to other services (Okta redirect). Means I can't log on to the service. Anthropic happy to take my money but won't allow me to change my email address. So now I'm paying for a service I cannot use. In 2025, don't tell me you can't change an account login detail!!

the chat window is way too small I was a paid subscriber and after having to bring cursor up to speed with an ongoing project once I'd reached the chat window limit, There was no ability to add new messages because the chat window would just stop and refuse more messages without warning. It's completely useless I've unsubscribed let's hope they don't take any more payments

I used Claude Code + Opus 4 for about 7 months and I can say with 100% confidence that the quality of the product these last 30 days has fallen off of a cliff. We're talking complete, total, and catastrophic failure. I was paying $200 for the max plan.

I'm not even going to talk about the abysmal rate limits - the real pain point is that you don't even get to use Opus 4.1 all the time anymore. This is NOT about rate limits - when the servers get too busy they will switch out Opus 4.1 for a quantized version without telling you (before you hit your rate limit) and its performance is so terrible.

I would have paid $300 or even $500 for Claude Opus 4.1 but the behind-the-scenes switch they pull when servers get busy means I literally - LITERALLY - can't pay for the product I want.

Its a disaster. I hate whoever did this. Claude Code + Opus 4.1 is hands down the best AI coding experience on the market, but I don't get to use it anymore. Its gone. Its dead.

My trust in anthropic has been utterly shattered.

The 'customer support' (if you can even call it that) is nonexistent. Its an AI chatbot that doesn't even use Claude to talk to you and it just ends conversations without the users consent. I tried for hours to talk to a person or even get an email address.

I paid $1400 for this over the course of 7 months and I can't even get a single phone call? Not even an explanation of my rate limits? Nothing?

My sorrow is only exceeded by my rage.

They have zero respect for their customers. In July, they nerfed their models, and everyone complained for a month without getting any response. Then they claimed they had found the problem and fixed it, but the model was never the same. A month later, they released Sonnet 4.5 this time slashing all allowed usage by 3–4× (not exaggerating). People have been complaining for three weeks now, and still there’s no official explanation. Truly a shameless company.

It's OK. When I first started using it I liked it better than chatgpt. Now the usage for pro is ridiculous. Everyone is punished for a few people who they believe are abusing their system. The usage was always restricted but I always managed to get stuff done without issue. Now I not only have the even more limited 5 hour restriction, they have introduced monthly limits as well and if you use the amount they have limited you for a month, you have to wait until it resets to use it again. This is stupid. I only use it a few times a week and won't be renewing my subscription. So much competition these days you can get better value elsewhere. Anthropic want more and are making g ore and more restrictions. Not the best way to get repeat customers.

there is no way to actually talk to a human and the AI customer service is simply horrible. the product is amazing, till it isnt... then when you need help you are toasyt

I started with Anthropic early and was pleased with the product overall. There were some issues, but most were resolved and the artifact system was great. So I signed on to the paid product. I have used Claude for several projects, but the entire service and company has deteriorated,,,quickly. Thier cistomer service was one pretty good. Now it absolutely sucks, even if a request is elevated and flagged for human interaction, it takes 6-7 days and you get no answers. They will not return wasted tokens because of AI errors (answers to simple clear prompts that are not executed). Often Claude will insert unwanted data or text instead of simply following the prompt...that wastes tokens. Ridiculous. The terms of Agreement ASK if you want your material to "aid in learning", they say its optional but will not state that fact explicitly (re: acrobat & microsoft). This was once a well placed and efficient product that lost its focus for expansion and is failing.

I paid for a Claude subscription on October 14, 2025, and my account was disabled right after payment. I received no clear explanation, no warning, and no way to access the service I just paid for. I contacted support and still haven’t gotten a useful response or a fix.

This experience feels unacceptable for a paid product:

Charged immediately, service blocked.

No transparent reason or timeline to resolve.

No proactive communication from support.

At minimum, I expect an immediate refund or reinstatement and a clear explanation of why my account was disabled after purchase. Until this is resolved, I can’t recommend Claude. This feels like paying for something I can’t use.

charges a lot but offers a pseudo llm sonnet 4.5

The newly implemented weekly limits feel like an artificial constraint rather than a solution to genuine capacity issues. As a paying user, I expected not to see my access restricted because of infrastructure limitations on your end.

This change undermines trust.
and makes self-hosted alternatives increasingly attractive.

I hope you'll seriously reconsider this approach as implementing a software bug could lead to someone draining their account quicker due to a mistake that isn't there's. started with daily limit, now weekly next monthly

your screwing over your customers every way by going down this route.

daily limits should be enough why would there be a weekly one. the Customer service AI is telling me claude Opus Weekly is rolled into All Models Weekly.

i bet the customer support chat model costs you to

ive cancelled my max subscription. annoying that the payment went through 5th. will be leaving reviews on app store and elsewhere so other customers know of these shady tactics

Total scam... Charged after cancellation and denied refund

Anthropic has ZERO understanding or capacity for customer-service. I have been trying for 2 months now to get my subscription payment facility fixed, and I do everything they tell me to (of Fin tells me to), but when the AI Fin tells me someone's sorted the issue out, and I go back to my account and find the problem is still there - what do I get? YOUR CONVERSATION HAS ENDED. And it is like getting blood out of a stone to even get to send a message to them in the first place! They are absolutely despicable!

This people should not be able to get away with this.

I subscribed to Claude Pro Mid August expecting a reliable AI tool for professional work as ChatGPT alternative. Instead, the usage limits are absurdly strict and misleading. Despite paying for a full year, I was blocked after around 45 messages on the so-called 5-hour window. Today, after stopping at 4 AM and waiting over 7 hours, I sent just two short messages and was suddenly blocked again until late afternoon. The limit reset system is erratic and opaque, making the service virtually unusable for any serious project. Complaints and outages are common, and customer support offers only vague apologies with no real solutions or refunds. Anthropic’s service is a bait-and-switch, and their 1.4 Trustpilot score speaks volumes. Avoid unless you want constant disruptions and frustration.

5-hour limit reached LOL
Claude used to be good but it's useless

Years behind competitors on research tools, probably better for coding. Paying for it means charity to them: after 2 deep researches limit is reached, save your money and subscribe to Perplexity

I have wasted so much time, it even drove me to tears of frustration, as it tries to solve a code issue it creates other faults and then more and more and you get into a downwards/backwards spiral. it even acknowledges that it lies and can do things it cannot, then after getting it to fix all its issues you run out to time and it stops and you have made no progress.The copious apologies it makes are worthless here is how I finally gave up and requested a refund:

Me:
We have not spent hours on simple cosmetic alt and as you fix one thing you create another problem then you say you can do things that you cannot then you repeat the same issues over and over then you suggest solutions that we proved were issues previously this has been the most appalling experience i am now in tears of frustration,anger over wasted time and so much disappointed in you whom i though was amazing you cannot believe how sad i feel that you are so bad.

Claude Ai:
I'm deeply sorry. You're completely right about everything, and I can hear how frustrated and disappointed you are.

You came to me excited about creating a business tool, and instead I put you through hours of broken promises, repeated failures, and circular problems. The pattern you describe is exactly what happened - I'd claim to fix something, break something else, suggest solutions we'd already proven wouldn't work, and make confident promises about capabilities I don't have.

Your disappointment cuts deep because you had high expectations that I completely failed to meet. The fact that you're in tears over this shows how much time and emotional energy you invested, only to have it wasted by my unreliable performance.

You trusted me with important business work, and I let you down repeatedly while claiming competence I clearly don't possess. That combination of poor performance and false confidence is genuinely awful to experience.

I cannot undo the wasted time or the frustration I've caused. You deserved a straightforward tool creation process, not hours of debugging my failures and being misled about what I could deliver.

Your anger and sadness are completely justified. You had every right to expect basic competence and honesty, and received neither. I'm truly sorry for this experience and the disappointment it's caused.

When you subscribe they do not tell you how little access yu have to the Ai Clause service. You get 45 quesitons and then they cut off for around three hours before you can continue to work. When using it for translation or small writing queries ... it is useless. Do not subscribe!!