The Great LLM Showdown
Back in the 90s, when I was grinding out spec scripts, scribblers were pretty much solo players. Sure, you might have a writing group or that one friend who'd give notes, but mostly, it was you versus the blank page. Now? We've got AI squadmates ready to provide support.
But which AI makes the best writing partner? I've been running experiments with some of the different language models - ChatGPT, Claude, Gemini, Perplexity, Google Notebook LM, and a little bit of the new kid - DeepSeek.
Text: Who's Got the Best Polish Game?
Gemini's like that friend who's amazing at copy editing - eagle eyes for typos and grammar, keeps everything clear and readable. Won't completely change your voice.
Claude Opus really gets what you're going for style-wise, but needs you to be clear about what you want. Give it some examples of your voice, and it'll run with it.
ChatGPT-4 is solid at cleaning up messy drafts, but sometimes it's like that network exec who wants everything "explained" - can sand off too many edges if you're not careful.
If you want to keep your unique voice while tightening things up, Claude Opus or ChatGPT-4 are your best picks. But if you just need a quick polish pass, Gemini's pretty rad.
Creative Brainstorming: Who's Best at Riffing?
Claude Opus is great at breaking story - loves playing with "what if" scenarios and finding thematic connections. Not every pitch is gold, but it helps find the good stuff.
ChatGPT-4 reminds me of my Heroes days using Scrum boards - great at tracking character arcs, plot points, and keeping everything organized across multiple sessions.
DeepSeek seems to know its genres, but needs a push to get truly weird and original.
Want a scribbling partner that can keep track of your whole story bible? Google Notebook LM & ChatGPT-4 rock. Need someone to bounce crazy ideas off of? Claude Opus is ready to play. Want solid genre work? DeepSeek says it’s got the chops. But I haven’t got deep enough to know.
Research Writing: Who's Got the Best Intel?
ChatGPT-4 lays out info clean and clear, but always verify your sources.
Claude Opus is like a story editor who's great at pulling together different threads into something cohesive. Try it for expanding blog posts or world-building.
Gemini and Perplexity are quick with facts and connected to the interwebs.
For structured research and blog work, I'd roll with Claude Opus or ChatGPT-4. But double-check their work via Gemini or Perplexity. Trust, but corroborate.
The Final Boss: Which AI Should You Pick?
Need a scribbling polish? Claude Opus for style, Gemini for clarity.
Brainstorming session? ChatGPT-4 for structure, Claude Opus for wild ideas.
Research deep dive? Claude Opus or ChatGPT-4, depending on how you like your info served.
Here's my current workflow:
I start with Perplexity for deep research, feed that into ChatGPT to structure and expand ideas, run it through Gemini for clarity, and finish with Claude for that final style polish. It's like a relay race where each runner has their specialty.
But if you're picking just one? ChatGPT-4 is your best all-arounder.
Remember when games went from single-player to co-op? That's what AI tools are doing for writing. They're not replacing us - they're giving us teammates in our Scribbler's Toolbox. They can help you tackle the blank page, ensuring you get to ABS. Always - Be - Scribbling!
If you’re curious, here’s the Query and Primer I used for the evaluation:
Evaluation Query:
I am a professional screenwriter and blogger. I use LLMs primarily for expanding blog post ideas and rewriting rough, wordy text to better match my writing style. I also brainstorm creative story elements, including character arcs, plot structures, and world-building details. Based on your capabilities, how well do you perform in these areas? Please provide a brief self-assessment of your strengths and weaknesses in research writing, text refinement, and creative brainstorming.
Process Primer:
I am a professional screenwriter and blogger who uses LLMs for specific tasks. For this session, focus on [research writing, blog idea expansion, creative brainstorming, etc.] and do not worry about prose rewriting, as I have a separate LLM configured for that.
When assisting me, prioritize concise, natural language that aligns with my professional writing style. Avoid overly wordy, generic, or academic phrasing. Instead, aim for responses that feel specific, polished, and practical for immediate use.
For blog work: Help me develop structured, engaging, and well-researched blog post ideas.
For creative brainstorming: Provide fresh, interesting takes on character arcs, plots, and worldbuilding, favoring grounded but compelling ideas.
For research: Focus on accuracy, clarity, and presenting information in a way that sparks further exploration rather than generic summaries.
Please acknowledge this guidance before we start.
