I Use All Three AI Assistants Every Week. Here's When I Pick Each One
ChatGPT, Claude, and Gemini each do something best. After using all three for months, here's my actual routine: what I use when, why, and what still sucks about each one.
TL;DR
ChatGPT is my daily driver: best all-rounder, best for coding, best plugin ecosystem. Claude is where I go for serious writing and analyzing long documents. Gemini I mostly use for YouTube research and anything Google-adjacent. If I had to pick one: ChatGPT. If I could pick two: ChatGPT + Claude.
I pay for ChatGPT Plus and Claude Pro. I use Gemini's free tier when I need it. That setup costs me $40/month, which is less than my Spotify and Netflix combined — and I get way more value from it.
But I didn't start here. I spent months bouncing between them, trying to figure out which one "deserved" to be my main tool. The answer turned out to be: none of them, and all of them. Each has a lane.
The Quick Comparison
| What matters | ChatGPT (GPT-4o) | Claude (Opus 4) | Gemini (2.5 Pro) | |---|---|---|---| | Writing quality | Good, can be verbose | Best natural prose | Functional, corporate tone | | Coding | Best overall | Good, especially Python | Fine, not great | | Long documents | 128K tokens | 200K tokens | 1M tokens | | Image generation | DALL-E 3 built in | None | Imagen 3 | | Internet access | Yes (Browse) | Limited | Best (native Google) | | Free tier | GPT-4o mini | Claude Haiku | Gemini 2.5 Flash | | Paid price | $20/month | $20/month | $19.99/month | | Best use case | All-rounder | Deep thinking, writing | Research, Google integration |
ChatGPT: The Swiss Army Knife I Open First
Every morning, I open ChatGPT. Not because it's the best at everything — it isn't — but because it's the best at the most things, and the friction to switch tools is real.
What ChatGPT actually does well:
The plugin and GPT Store ecosystem is what keeps me here. When I need to analyze a spreadsheet, I switch to the Data Analyst GPT. When I'm brainstorming article ideas, the custom GPT I built with my writing preferences saves me 20 minutes of context-setting. No other platform has this breadth of specialized tools.
Coding is where ChatGPT pulls ahead clearest. It handles React, TypeScript, Python, SQL — the languages I use daily — with fewer mistakes than Claude, and the data analysis mode (which runs actual Python) catches calculation errors that pure language models miss.
The voice mode on mobile matters more than I expected. I use it to talk through article ideas while walking. Claude has no equivalent, and Gemini's voice mode feels bolted on.
What drives me crazy:
ChatGPT overexplains. Every answer comes with a preamble. I've started adding "no preamble, just the answer" to my Custom Instructions, which mostly works, but the default verbosity is baked into the model. It's also the worst of the three about pretending it knows things. Claude will say "I'm not sure" — ChatGPT will give you a confident, detailed, completely wrong answer with citations it made up.
Claude: Where I Go When Writing Actually Matters
If ChatGPT is the Swiss Army knife, Claude is a really good fountain pen. It does fewer things, but for certain tasks nothing else comes close.
What Claude does better than anyone:
Writing. I wrote the first draft of this article in Claude, and I write all my long-form content there. Claude's prose has a natural rhythm that ChatGPT can't match. It uses fewer words to say the same thing. It varies sentence length without being told. It doesn't sound like a LinkedIn post.
The 200K context window (roughly 150,000 words, or a full book) means I can upload entire documentation sites, long research papers, or full conversation histories. ChatGPT's 128K is fine, but the extra 72K tokens in Claude means I don't need to think about trimming.
Claude is also noticeably better at understanding what I meant rather than what I literally said. I can be sloppy with a prompt and Claude will figure it out. ChatGPT takes me more literally. Gemini takes prompts extremely literally — if you write a bad prompt, Gemini gives you exactly the bad thing you asked for.
What I wish was different:
No image generation. This is the biggest hole in Claude's feature set. When I'm writing and want a quick header image or diagram, I need to switch to ChatGPT or Midjourney. It's a friction that shouldn't exist at this price point.
Claude can also be overly cautious. I once asked it to help draft a blog post about prompt injection attacks — a legitimate security topic — and it initially refused, treating it like I was asking for exploit instructions. ChatGPT handled the same request without issue. Claude's safety stance is principled but sometimes gets in the way of legitimate work.
Gemini: The Specialist I Use Less Than I Should
I want to use Gemini more. Google gave it a 1M token context window and deep integration with the tools I already use — YouTube, Maps, Gmail, Google Docs. In theory, this should be incredible.
In practice, Gemini is excellent for a narrow set of tasks and mediocre outside them.
What Gemini genuinely excels at:
Research that involves Google's ecosystem. "Find me recent YouTube tutorials about Next.js 16 App Router" — Gemini pulls actual video timestamps and summarizes content. Neither ChatGPT nor Claude can do this. The same goes for finding academic papers, Google Scholar citations, Maps data, and real-time web information.
The 1M token window is absurdly large. I once uploaded an entire open-source project (every source file, concatenated) and asked it to find all the places where we were using a deprecated API. It found 14 instances across 200+ files in about 30 seconds. That's not something ChatGPT or Claude can match with their current context windows.
Gemini is also the cheapest for high-volume tasks. If you're processing a lot of text and don't need the highest quality, Gemini Flash handles it for essentially free.
What stops me from using it more:
The writing is... fine. Technically correct, grammatically perfect, and completely devoid of personality. Everything sounds like an internal memo at a large corporation. "We are excited to announce..." energy. If you just need information delivered clearly, Gemini does the job. If you want your writing to sound like a human wrote it, use Claude.
The regional feature gaps are real and annoying. What works in the US might not work in Europe. What's available in English might not exist in other languages. Google has been promising to fix this for years. It's still a problem.
The Setup That Works for Me
Here's my actual daily pattern:
- Morning, quick tasks → ChatGPT. Emails, code snippets, quick research.
- Afternoon, deep work → Claude. Long-form writing, document analysis, technical decisions that need careful reasoning.
- Research-heavy tasks → Gemini. Anything that benefits from YouTube, Maps, or real-time search.
- Image needs → ChatGPT (DALL-E) or Midjourney.
- Coding sessions → ChatGPT with occasional Claude for code review.
If you can only pay for one, pick based on your main activity:
- Writing and analysis → Claude
- Coding and general use → ChatGPT
- Google ecosystem and research → Gemini
If you can pay for two (about $40/month total): ChatGPT + Claude. This covers 95% of use cases. Add Gemini Free for the Google-specific stuff.
What's Still Missing From All Three
None of them handle images as input as well as I want. Uploading a UI mockup and getting usable code is still hit-or-miss. They all struggle with very specific edge cases that any human domain expert would catch instantly.
They're all improving at different speeds. Six months ago Claude was clearly ahead at writing; the gap has narrowed but still exists. A year ago ChatGPT's coding was dominant; Claude has mostly closed that gap. The competition is real, and as a user, it's great.
The best tool depends on what you're doing right now. Don't be loyal to one. Be loyal to getting the best output.
FAQ
Which one should a beginner start with?
ChatGPT. It has the largest community, the most tutorials, and the gentlest learning curve. Start free, learn the basics, then try Claude if you find the writing frustrating or Gemini if you live in Google's ecosystem.
Can I just use the free tiers?
Honestly, yes. GPT-4o mini and Claude Haiku are surprisingly capable. The paid tiers get you smarter models, longer conversations, and extra features — but the free versions handle 80% of everyday tasks. Upgrade when you feel the ceiling, not before.
Is Claude really that much better at writing?
Yes. If you publish written content regularly, the difference pays for itself within a week. ChatGPT writes that's grammatically correct but vaguely corporate. Claude writes that sounds like a smart colleague explaining something. For casual use, ChatGPT is fine. For anything you're putting your name on, Claude is worth the $20.
Why does Gemini feel so different from the others?
Gemini was built by a search company thinking about information retrieval. ChatGPT and Claude were built by research labs thinking about conversation. That origin difference still defines how each product feels. Gemini answers questions. ChatGPT and Claude have conversations.
Are there other AI tools I should look at?
Perplexity for research. Cursor or GitHub Copilot for coding inside an IDE. Midjourney for AI images. Each does one thing better than the big three. But for general-purpose AI assistance, ChatGPT, Claude, and Gemini are still the main show.