Why GPT-5 Feels Like a Creative Downgrade (& How to Fix It)
Z
Zack Saadioui
8/12/2025
You’ve been eagerly awaiting GPT-5, hoping for a quantum leap in your creative writing process. But now that it’s here, something feels… off. Right? If you're a writer, a creator, or just someone who loved brainstorming with previous models, you might be feeling a bit let down. Honestly, you’re not alone.
Across Reddit forums & social media, the consensus is pretty clear for a lot of creative users: GPT-5 feels like a downgrade. People are saying it’s “colder,” “more robotic,” & has lost the “emotional nuance” that made GPT-4o feel like a genuine creative partner. One user on Reddit said it feels like their ChatGPT "got put on lithium," becoming more stable but "boring AF." It seems like in the push for speed & accuracy, something essential for creative work got lost.
The biggest complaint I’ve seen is about its thinking style. Where GPT-4o could juggle multiple ideas, jumping from concept A to B & then back to A to synthesize something new, GPT-5 seems more rigid & linear. It’s great for logical tasks, sure, but that’s not what you need when you’re in the messy, non-linear flow of creating a story or building a world. It’s like it lost its ability for associative thinking, which is the bedrock of human creativity.
So, what gives? It seems like in an effort to make the model "safer" & more efficient, there’s been a kind of "intellectual flattening." It’s less likely to go off the rails, but it’s also less likely to have that spark of unexpected genius. For writers who relied on that spark, it’s a HUGE problem.
But here’s the thing: it’s not a lost cause. You just have to change how you work with it. The days of simple, conversational prompting for high-quality creative output are, for now, on pause. You have to become more of a pilot & less of a passenger. It takes more effort, but you can absolutely still get amazing, creative results out of GPT-5. You just need to learn the new rules of the game.
Let’s dive into how you can stop being disappointed & start getting the creative results you want.
Think Like a Prompt Engineer, Not a Novelist
This is the biggest mindset shift you need to make. With older models, you could be more of a collaborator. Now, you need to be more of a director. Prompt engineering isn't just for techies; it's the new art of giving instructions to an AI. It’s about being incredibly deliberate & strategic with your words.
The Anatomy of a Killer Creative Prompt
A simple "write a story about a sad robot" isn't going to cut it anymore. You need to build your prompts with layers. Here’s a framework I’ve found works wonders:
Assign a Persona: This is your starting point. Don't just ask it to write; tell it who it is.
Bad Prompt: "Write a fantasy scene."
Good Prompt: "You are a world-weary fantasy novelist in the vein of Joe Abercrombie, known for your gritty realism, morally gray characters, & cynical tone. Your prose is sharp, visceral, & avoids clichés."
Provide Rich Context: Give it the raw materials. The more context it has, the less it has to invent from its new, more limited imagination.
Bad Prompt: "The scene is in a tavern."
Good Prompt: "The scene is set in a dimly lit, smoke-filled tavern called 'The Gilded Rat' in the corrupt port city of Oakhaven. The air smells of stale ale, salt, & desperation. The patrons are a mix of grizzled sailors, nervous merchants, & a few shadowy figures in the corners. The main character, Kaelen, is a disgraced knight nursing a drink, trying to forget the battle he fled."
Be Explicit About Style & Tone: Don't hope it will figure it out. Spell it out. Use literary terms. Mention other authors.
Bad Prompt: "Make it sound cool."
Good Prompt: "Write in a third-person limited perspective, sticking closely to Kaelen's point of view. Use a cynical & introspective tone. The prose should be concise, with a focus on sensory details—the stickiness of the table, the low murmur of conversations, the clink of coins. Avoid purple prose. The dialogue should be clipped & realistic."
Give It a Goal & Constraints: Tell it what you want the scene to accomplish & what it should avoid.
Bad Prompt: "Something interesting should happen."
Good Prompt: "The goal of this scene is to introduce Kaelen's internal conflict & hint at his past trauma. A mysterious stranger should approach him with an offer that he initially refuses. The scene should end on a note of tension & uncertainty. DO NOT reveal the specifics of why he's a disgraced knight. Keep the stranger's motivations ambiguous."
When you combine all these elements, you’re not just asking for a story; you’re providing a detailed blueprint. You’re guiding its every decision, which is now, unfortunately, more necessary than ever.
Become the AI's Trainer for a Session
One of the most powerful techniques I've found is to "train" the AI within a single chat session. Since it has memory within a conversation, you can teach it your specific style before you even start asking for new content.
Here's a step-by-step process that works like a charm:
Step 1: The Introduction. Start a new chat & tell the AI what you're about to do. Be direct.
Prompt: "I am going to teach you my creative writing style. I will provide you with several examples of my writing. Do not do anything with them yet. Just analyze them for tone, voice, pacing, sentence structure, vocabulary, & common themes. Just reply with 'READY' when you understand."
Step 2: Feed It Your Work. Copy & paste 3-5 examples of your writing that you're proud of. They should be representative of the style you want to achieve. After each paste, wait for it to process. A few hundred to a thousand words is a good amount.
You:[Paste example 1]
GPT-5: "READY."
You:[Paste example 2]
GPT-5: "READY."
Step 3: Ask It to Synthesize. Once you've given it enough material, ask it to create a style guide based on your work.
Prompt: "Based on the examples I've provided, create a detailed 'Style Guide' for my writing. Break it down into sections: Tone & Voice, Sentence Structure, Vocabulary, Pacing, Characterization, & Themes. Be as specific as possible."
Step 4: The Magic Happens. The AI will now generate a surprisingly accurate analysis of your own writing style. This is valuable on its own, but it's also the internal "instruction manual" it will now use for the rest of the conversation.
Step 5: Start Writing. Now, you can give it a creative prompt, but with a new instruction.
Prompt: "Okay, great. Now, using the 'Style Guide' you just created, write a scene about a young mage discovering a forbidden spell in a dusty library. Make sure it adheres strictly to the style guide."
The difference in output is night & day. You’ve essentially fine-tuned the model for your specific needs within that session. It’s no longer a generic, cold AI; it's an AI trying its best to be you.
Mastering the Technical Knobs: Temperature
If you're using the AI through an API or a platform that allows for more technical control, you NEED to understand the "temperature" setting.
In simple terms, temperature controls the randomness of the AI's output.
Low Temperature (e.g., 0.2): The AI plays it safe. It will choose the most common, predictable next word. This is great for factual summaries or technical writing. This is likely the default "feel" of GPT-5 that’s causing so much disappointment. It's deterministic & boring.
High Temperature (e.g., 0.8 to 1.0): The AI takes more risks. It's more likely to choose less common words, leading to more creative, surprising, & sometimes even strange outputs. This is where the magic lies for creative writing.
Think of it like this: a low temperature gives you a cover band playing the hits exactly as you've heard them. A high temperature gives you a jazz band improvising on a theme. For creative work, you want the jazz band.
If your interface allows it, crank up the temperature to around 0.75 or 0.8 when you’re brainstorming or writing first drafts. You might get some nonsense, but you’ll also get those happy accidents & unexpected phrases that spark new ideas. You can always dial it back down for more focused editing later. Experimentation is key to finding the sweet spot.
Iteration is Everything
Don't expect the perfect draft on the first try. The new workflow is a conversation, a process of refinement. Think of the first output as a lump of clay. It's your job to shape it.
Here’s a practical iterative workflow:
Generate the First Draft: Use a detailed prompt, as we discussed.
Critique & Re-prompt: Read the output. What’s weak? The dialogue? The descriptions? Be specific in your feedback.
Instead of: "Make it better."
Try: "This is a good start, but the dialogue feels a bit stilted. Rewrite the conversation between the two characters to be more informal & filled with subtext. Have them talk around the issue instead of addressing it directly."
Focus on a Single Element: Ask it to do one pass just for sensory details. Then another for character introspection. Then another for pacing.
Prompt: "Now, let's enhance the atmosphere. Go through the scene & add more sensory details. What does the air smell like? What distant sounds can be heard? What is the texture of the objects the character touches?"
Combine & Polish: Once you have several refined pieces, you can combine the best parts & do the final human polish.
This approach turns the AI from a disappointing author into a powerful, multi-faceted editing tool.
A Note on AI for Business & Engagement
It's interesting to see how this push for "safer" & more predictable AI is actually a benefit in other contexts. For creative writing, it's a frustration. But for a business, that predictability is a feature, not a bug.
For example, when a company wants to automate customer service, they don't want a creative, improvising chatbot. They need a reliable, accurate one. This is where platforms like Arsturn come in. They help businesses build no-code AI chatbots trained specifically on their own data. This means the chatbot provides instant, consistent, & accurate support 24/7. It's designed for precision, not poetry. So, if a customer visits a website, the Arsturn-powered chatbot can instantly answer their questions about shipping, products, or policies without going off on a creative tangent.
It’s the same underlying technology, but optimized for a completely different goal. Where a writer wants a chaotic brainstorming partner, a business needs a dependable digital employee. Pretty cool to see how the same tech can be tailored for such different needs. For business communication & website engagement, this kind of focused AI is a game-changer, helping to boost conversions & provide personalized experiences at scale.
When All Else Fails: Know When to Use Other Tools
Here's the hard truth: sometimes, GPT-5 just might not be the right tool for the job anymore, especially in the early, messy stages of creation. Many users who are frustrated with GPT-5 have started using other AI models for their creative work, as hinted in some forums.
Don't be afraid to explore alternatives. Models from other companies like Anthropic (Claude) or even open-source models might give you a different flavor of creativity that aligns better with your process. Each model has its own "personality" based on its training data & architecture. The key is to build a toolkit.
You might use a different AI for initial brainstorming & character generation, then bring that raw material back to GPT-5 for focused outlining & scene-by-scene generation using the prompt engineering techniques we talked about.
Tying It All Together
Look, it's totally valid to be disappointed with GPT-5's creative abilities right out of the box. It is different, & in some ways, it is worse for the freewheeling creative process we got used to. But it's not a creative death sentence.
The key is to shift your role from a passive user to an active director.
Master the art of the detailed prompt. Give it a persona, context, style guides, & goals.
Train it on your own voice at the start of every session.
Embrace iteration as your new workflow.
Play with technical settings like temperature whenever you can.
By being more deliberate, more specific, & more strategic, you can force the "boring" out of it & coax the creativity back in. It’s more work, yes, but it forces you to be a more conscious & analytical writer in the process, which is a skill that will serve you well, with or without an AI partner.
Hope this was helpful. It's a new landscape for all of us, but with the right approach, we can still make incredible things. Let me know what you think & if you find any other tricks that work for you.