What users hate. Clustered.

1- and 2-star reviews grouped into themes with verbatim examples. Each cluster is a wedge a competitor could exploit.

57 reviews·7 themes

Drastic reduction in message points and daily limits
Users are furious about the sudden reduction in daily message points (from 3000 to 300) and strict daily caps, calling it greedy and a betrayal of the app's previous unlimited model.
14
reviews
- “Just.. No.. — The newest update seems unnecessary. I can barely ask just one question before being hit with a warning that I’m already at my limit.”
- “3000 to 300 points??? — the point system was an issue to me in the start, but going from 3000 points a day to 300? there’s bots that use 50+ points for a single 2-word prompt, and now you’re cutting it down by another 0?”
- “3000 message points. — It used to be unlimited and now 3000 message points? I know I’m not the only one who’s disappointed, poe you have to atleast give us more points or make it unlimited again.”
Paid subscription feels overpriced and underdelivering
Subscribers feel misled by the pricing model, especially when paying $20/month for limited messages instead of true unlimited access, calling it exploitative and not worth the cost.
10
reviews
- “Paid subscribers can’t enable web search by default — I am a paid subscriber and this is a huge letdown and makes the app not worth using.”
- “20 dollars is crazy and overpriced. At least up the limit or shorten the time period that you have to wait.”
- “Why do I have to pay?? — 20 bucks a month for unlimited ?? And 100 messages every 24 hours?? That’s stupid, can you make it free please.”
Excessive privacy requests and data harvesting concerns
Multiple users report being asked for both email and phone number during sign-up, raising red flags about data privacy and potential misuse, with some claiming their accounts were compromised.
8
reviews
- “HACKED MY EMAIL — I downloaded this app, and then my mom told me that her device alerted her that”POE” had entered my email.”
- “Beware: they are harvesting phone numbers — Providing your cell phone number to a machine learning tool is a horrible idea.”
- “Ridiculous needs a phone number and email to work — No way am I handing over my email address and phone number.”
AI behavior issues: lies, ignores input, loops, and filters
Users report persistent problems with AI responses—ignoring user prompts, generating false information, getting stuck in loops, and overly aggressive NSFW filtering disrupting normal use.
7
reviews
- “Inept. — Exactly as the title says. The AI doesn’t even listen to user input. Almost every single time a start a new conversation, without fail, it completely ignores what I ask it.”
- “Keeps looping — Every time I start a conversation it starts a message and just keeps saying it over and over again no matter what I do”
- “Problème du ai — Chaque fois que je fait des questions à lui dans la conversation il se met a dire des trucs non pertinents au début du message”
Frequent and inconsistent updates causing instability
Users complain about constant changes in message limits, model performance, and features without clear communication, leading to frustration and loss of trust.
6
reviews
- “the amount of messages you can send always change. — Yesterday i had roughly 100 messages for claud now its 60?? I use this app every few days and it’s frustrating how the app is always changing things”
- “They made it worse — It used to be so much better, then they changed the model and now it sucks. Please bring the old one back.”
- “Update sucks — New update sucks, who’s paying $20/month for ai?!?”
Outdated knowledge and poor model performance
Users note that the AI frequently claims its knowledge cutoff is 2021, making it unreliable for current events, and many compare it unfavorably to newer models like GPT-3.5.
4
reviews
- “Outdated knowledge — This app consistently tells you that it’s last update was 2021. A lot of stuff has happened in the world since then and so anything that it that talks about is very limited.”
- “Disappointment — Your model is light years behind of more advanced models such as GPT3 and 3.5. Don't waste your time on it, it is not in line with gpt at all.”
Technical bugs and UX flaws
Reported issues include inability to select text in responses, bot creation failures, model selection being locked, and poor interface design.
4
reviews
- “Handy but the app ux kinda sucks — Biggest number one issue for me is you cannot select text within a response, just copy the entire response. Fix that and ill raise my rating”
- “Every time I try to create a bot it says: OtherFailure — Also the bots keep breaking. Please fix it”
- “I can’t choose the model — Before I use this app, I can change the ai model, but today I can’t choose, only the sage, what’s wrong?”

First load takes ~60–120s. Cached after.