Quick Summary
ChatGPT feels worse to use now because of five key reasons: it struggles to follow negative instructions (“don’t do X”), it over-explains with unnecessary context, it holds onto context from previous sessions unexpectedly, it gives contradicting answers, and the free tier has become significantly limited compared to when it launched. These aren’t just your imagination, thousands of users are reporting the same issues on Reddit and forums. If you’re frustrated with ChatGPT not working the way it used to, you’re not alone, and there are better alternatives available.
ChatGPT not working like it used to? Here are the real reasons it’s gotten worse and the alternative that actually understands you.
There are now thousands of posts on Reddit with titles like
“Does anyone else feel ChatGPT output has worsened?”
and “It’s become unbearable.”
They’re from paying Plus subscribers, people who are literally paying $20/month and still feeling frustrated.
So what’s actually going on? And more importantly, is there a better option?
Let me break it down.
The 7 Reasons ChatGPT Feels So Bad Right Now
1. It Can’t Handle Negative Instructions
This is the biggest complaint I see. You tell ChatGPT: “Don’t mention XYZ in this speech.”
What does it do? It literally mentions XYZ, usually right at the end, with a sentence like: “So, you shouldn’t think this is about XYZ. It’s about [topic].”
One Reddit user put it perfectly: “When you create your prompts, talk about what you want to include, NOT what you want to exclude. The moment you type things into the system it becomes part of that thread and the more things you tell it NOT to do, the more befuddled it will be.”
The model literally weights “don’t mention XYZ” as a signal to include XYZ in its attention mechanism. It’s counterintuitive, but that’s how these models work. And ChatGPT doesn’t automatically compensate for this the way some other models do.
2. It Over-Explains Everything
Remember when ChatGPT used to give you short, direct answers?
Yeah, me neither.
Now, even for simple queries, you’ll get paragraphs of unnecessary context. One user on Reddit complained: “It answers a long pile of unwanted shit on each iteration.”
You ask a simple question. You get a lecture. You ask it to shorten the response. It shortens it slightly, but still includes three paragraphs of preamble.
This is particularly painful when you’re working on something time-sensitive, like drafting client emails or generating ad copy. You don’t have time to wade through AI waffle. You just need the answer.
3. It Holds Onto Context From Previous Sessions
This one is subtle but infuriating.
You start a new conversation about Project A. But ChatGPT subtly pulls in context from conversations you had last week about Projects B, C, and D. One user on Reddit said: “It holds on to more contexts, and randomly spits out information from like 10 sessions ago.”
So now you’re trying to write something for Client X, and ChatGPT keeps referencing strategy from Client Y.
This is especially painful for professionals managing multiple client accounts. You can’t afford cross-contamination of context. One client’s branding, tone, and strategy should stay separate from another’s.
4. It Gives Contradicting Answers
Here’s a fun one: ask ChatGPT the same question twice in the same conversation.
Chances are, you’ll get different answers. Not because the question was ambiguous, because the model is inconsistent.
One user on Reddit described it: “It literally gave me contradicting answers for something I was doing, one minute apart. It has its off days for sure.”
This is a nightmare for anyone doing analytical work, research, or content creation where accuracy matters. You can’t trust the output the first time. You have to double-check everything. At that point, what’s the point of using AI at all?
5. It Ends Responses With Questions (Like It’s Prompting You)
This is a newer complaint that’s been popping up.
Several Reddit users noticed that ChatGPT now ends responses with questions, as if it’s prompting the user back, rather than completing the task.
One user said: “Mine started ending every string with a question as if it is prompting me.”
So instead of giving you a completed deliverable, ChatGPT turns every response into a conversation starter. “Here’s your email draft. Would you like me to add anything else? Should I make it longer? Do you want me to adjust the tone?”
It’s almost like ChatGPT is trying to keep you in the conversation to burn through your message limits faster.
6. The Free Tier Has Become Significantly Limited
Let’s be honest: the free version of ChatGPT was impressive when it launched. But over time, OpenAI has progressively limited what free users can access.
The models available to free users are older. The rate limits are tighter. And the experience is noticeably worse than what Plus subscribers get.
If you’re on the free tier, you’re using a gimped version of a tool that’s already showing signs of decline. That’s a double whammy.
7. It’s Bloated With System Prompts
This is a technical issue but it affects user experience.
Multiple Reddit users have noticed what appears to be heavy “system prompt” interference, layers of instructions baked into ChatGPT that affect how it responds. One user said: “It feels like they’re adding more lines to the pre-prompt moment. I can see it like, checking all its pockets for context and looking nervous while it can’t locate something but pretending it’s not nervous and totally plays it cool.”
In plain English: ChatGPT is carrying around so much internal baggage that it slows down and gets confused about what it’s supposed to do for you.
Why AI Failing Hurts Your Work
Here’s what happens:
→You spend more time editing
than you would writing from scratch. The time savings you expected? Gone.
→You can’t trust the output.
Contradicting answers mean you have to verify everything manually.
→Client work suffers.
Cross-contamination of context means you might accidentally mix up client brands, strategies, or tone. That’s reputation damage.
→You’re constantly re-prompting.
What should take one prompt takes five. The “quick task” you wanted done in 2 minutes takes 20.
One Reddit user summed it up perfectly: “I don’t care if they can solve PhD level problems better now. Just understand me better.”
That’s the heart of it. The flashiness doesn’t matter. Reliability matters. Understanding matters. Getting the job done without friction matters.
And right now, ChatGPT isn’t delivering on any of those.
OpenCraft AI: The Alternative To ChatGPT That Actually Works
Let me be clear: I’m not saying this to trash ChatGPT. We even have the latest GPT model in OpenCraft AI.
Used it extensively. It changed how I work. But I’m also not going to pretend it’s not frustrating right now, because it is, and you know it is.
Here’s what OpenCraft AI does differently compared to ChatGPT
1. It follows instructions, ALL of them.
We built our instruction-following engine differently. Negative instructions, complex multi-part prompts, specific tone requirements, they all work. You don’t have to rewrite your prompts in a specific way for the AI to understand.
2. It doesn’t over-explain.
OpenCraft AI gives you what you ask for. Short response requested? You get a short response. Bullet points? Bullet points. No unnecessary preamble, no follow-up questions unless you ask for them.
3. Clean session management.
Each conversation in OpenCraft AI is truly isolated. There’s no context bleeding from previous sessions. What you do for Client A stays for Client A. No cross-contamination. No surprises.
4. Consistent outputs.
We optimized for reliability. Ask the same question twice, you get the same answer (unless you explicitly ask for variation). This matters for anyone doing analytical work or content creation where accuracy is non-negotiable.
5. Multi-model access without the bloat.
OpenCraft AI gives you access to GPT, Claude, Gemini, and Grok, all in one place. You can switch between models mid-conversation depending on what you’re working on. No heavy system prompts. No hidden baggage. Just clean, focused AI that does what you tell it.
6. Persistent memory that works.
Store your brand guidelines, preferred tone, and client specifics once. OpenCraft AI remembers them across sessions, so you don’t have to re-explain yourself every time.
Try OpenCraft AI For Free
If you’re frustrated with ChatGPT not working the way it used to, you don’t have to suffer through it.
OpenCraft AI gives you the reliability and understanding you’re looking for, all in a cleaner, faster interface.
Start using it in under 2 minutes.
How Other Tools Compare To OpenCraft AI
ChatGPT (Free Tier)
What it does well: The interface is familiar. The model is capable when it wants to be. There’s a large community with prompt templates.
Where it falls short: Instruction following is inconsistent. Over-explanation is baked in. Context bleeding between sessions is a known issue. Free tier is heavily limited.
Verdict: Fine for experimenting, but not reliable for professional work.
Claude (Anthropic)
What it does well: Better at nuanced, thoughtful responses. Strong on long-form analysis. Generally more consistent than ChatGPT.
Where it falls short: Slower response times. Can be overly cautious with certain topics. Doesn’t have the same level of community support or plugin ecosystem.
Verdict: A solid alternative, but you still need to pay for multiple tools if you want image generation + writing + analysis.
Gemini (Google)
What it does well: Fast. Good integration with Google ecosystem. Strong on research and factual queries.
Where it falls short: Can be overly literal. Less flexible for creative tasks. Still catching up on instruction following.
Verdict: Good for specific use cases, but not a complete replacement for ChatGPT.
OpenCraft AI
What it does well: Multi-model access (GPT, Claude, Gemini, Grok in one place). Clean instruction following. No context bleeding. Persistent memory. Image generation + writing + analysis in one tool. $25/month includes everything.
Where it falls short: OpenCraft AI is newer player, and only has a small online community
Verdict: Best overall value for professionals who need reliability, consistency, and multiple AI capabilities without the bloat.
FAQ: Addressing Concerns About AI & OpenCraft AI
Is ChatGPT actually worse than before, or is this perception?
The perception is real, even if the causes are debated. Thousands of users on Reddit and forums report the same frustrations: harder to get good outputs, more re-prompting required, and inconsistent answers. Whether this is due to model changes, system prompt bloat, or something else, the user experience has degraded for many.
Is the free version of ChatGPT worth using?
Honest answer: not really for professional work. The free tier uses older models with stricter rate limits. You’re essentially using a demo version of a tool that’s showing signs of decline. If you’re serious about using AI for work, the free tier won’t save you enough time to justify the quality loss.
Why does ChatGPT ignore my instructions?
ChatGPT’s model architecture weights all tokens equally, including “don’t” and “not.” So when you say “don’t mention X,” the model actually increases attention on X, making it more likely to appear. This is a fundamental limitation of how these models work, not a user error.
What’s the best alternative to ChatGPT right now?
For most professionals, OpenCraft AI offers the best balance of reliability, instruction following, and value. You get access to GPT, Claude, Gemini, and Grok in one place, with clean session management and persistent memory, for $25/month versus $80+/month for individual subscriptions.
Can I use multiple AI models without switching platforms?
Yes. OpenCraft AI is designed exactly for this. You can switch between GPT, Claude, Gemini, and Grok mid-conversation depending on what you’re working on. No need to manage multiple accounts or browser tabs.
Does OpenCraft AI have the same context bleeding issue as ChatGPT?
No. We built OpenCraft AI with isolated session architecture. Each conversation is truly separate. There’s no cross-contamination from previous sessions, which is critical for professionals managing multiple clients or projects.
Is $25/month worth it versus ChatGPT’s $20/month?
Yes, because OpenCraft AI includes everything: GPT, Claude, Gemini, Grok, image generation, persistent memory, and multi-model switching. ChatGPT’s $20/month only gets you ChatGPT. If you need image generation on top, you’re paying another $10-30/month. OpenCraft AI gives you more models, more capabilities, and cleaner outputs for roughly the same price.
You Deserve Better AI
The frustrations you’re feeling with ChatGPT aren’t in your head. They’re real, they’re widespread, and they’re valid.
You don’t need a tool that can solve PhD problems. You need a tool that understands you, first time, every time. One that gives you what you ask for without the waffle. One that keeps your client work clean and separated. One that doesn’t contradict itself.
That’s what we built OpenCraft AI to do.
If you’re ready to stop wasting time re-prompting and start actually getting work done, try it free.
