Gemini Gems vs. ChatGPT GPTs vs. Claude Projects: Which Is Right for You?
Most people using Gemini or ChatGPT are doing it the basic way: type a question, read the answer, close the tab. That works. But every major AI platform now lets you go one step further and build a customized AI expert, a version of the AI you've configured for a specific role with context about your life and how you want it to communicate.
Google calls these Gems. OpenAI calls them GPTs. Anthropic calls them Projects.
I've tried all three, though not equally. Here's my honest take on each one.
What Are Custom AI Assistants?
A regular AI conversation starts from zero every time. The AI doesn't know you, your background, or what kind of answers actually help you.
Custom AI assistants solve that. You write a set of instructions once: the role, your background, how you want it to respond. Every future conversation starts with that context already loaded. The AI behaves less like a stranger and more like a specialist who knows your situation.
I use mine like a small team of advisors: a doctor for health questions, a lawyer for contract reviews, an investment advisor for portfolio thinking, and a few technical specialists for my IT work. None of them replace real professionals, but all of them help me ask better questions before I talk to one.
Gemini Gems: Google's Approach
This is the one I use most. I'm a Google AI Pro subscriber, which gives me access to the full Gems feature set. But Gems are also available on the free Gemini plan with some limitations, so you can try them before paying for anything.
Creating a Gem is straightforward: go to gemini.google.com, click "Gems" in the sidebar, hit "New Gem," and write your instructions. No coding required. Give it a name, describe its role and how it should respond, and save. Done.
The standout feature for me is Google integration. My Gems can pull context from Gmail, Drive, and Docs when I give them permission. My investment advisor Gem references a portfolio spreadsheet I keep in Drive. My network engineer Gem can look at config files I've saved in Docs. Real documents make the conversations significantly more useful than generic chat.
Gemini's real-time web access also means my Gems aren't limited to a training cutoff date, which matters for health or market questions where recent information counts.
ChatGPT GPTs: The App Store Model
Upfront disclosure: I don't use ChatGPT GPTs regularly. What I know comes from the platform, documentation, and people I know who do use it.
GPTs have one clear advantage Gems don't: the GPT Store. OpenAI built a marketplace of pre-made GPTs that other users have created and published. Want a GPT for academic research? There are dozens. Want one tuned for a specific programming language? Already exists.
This is genuinely useful for people who don't want to write their own instructions. The barrier to getting started is low. Browse, pick one, start talking.
The main limitation: building your own custom GPT requires a ChatGPT Plus subscription at $20/month. The free tier gives you access to existing GPTs, but not the ability to create your own. Gemini lets you create custom Gems on the free plan, which is a meaningful difference.
Claude Projects: The Developer's Choice
I use Claude, but mainly for coding work through Claude Code — not for personal AI advisors. My experience with Claude Projects is limited. I'm sharing what I know from research and from developers I work with, not from daily personal use.
Claude Projects gives Claude persistent context for a specific purpose. You upload documents, set instructions, and every conversation in that project starts with that context intact. It's particularly strong for large documents: long contracts, technical specs, research papers. The model handles nuance and complexity in text carefully.
For the "personal AI advisor" use case I'm describing here, Claude Projects works well. It just doesn't have Google ecosystem integration, which is what keeps me on Gemini.
Side-by-Side Comparison
| Feature | Gemini Gems | ChatGPT GPTs | Claude Projects |
|---|---|---|---|
| Create your own (free) | Yes | No (Plus required) | Yes (limited) |
| Pre-made library | No | Yes (GPT Store) | No |
| Google app integration | Gmail, Drive, Docs | No | No |
| Real-time web access | Yes | Plus only | No |
| Long document handling | Good | Good | Excellent |
| Ease of setup | Very easy | Easy | Easy |
| My personal use | Daily (main tool) | Researched only | Code work only |
Why I Stick with Gemini Gems
I use Google services heavily. Gmail, Drive, Docs, Calendar. Gemini integrates with all of them, and that integration is the practical reason I've stayed.
My investment advisor Gem can reference my actual portfolio spreadsheet. My network engineer Gem can review a real config file. That grounding in real documents makes the advice more specific than anything a generic AI chat can offer.
I also subscribe to Google AI Pro anyway, so the full feature set is already there. If I weren't in the Google ecosystem, I'd probably weigh the GPT Store more seriously. The pre-built GPTs are good, and for someone who doesn't want to write instructions from scratch, that convenience is real.
Claude I respect — technically impressive, and I use it heavily for code. But for personal advisor Gems, the lack of Google integration is a real gap for how I work.
How to Create Your First Gem
- Go to gemini.google.com and sign in. Click "Gems" in the left sidebar, then "New Gem."
- Write your instructions. Describe the role ("You are my personal doctor"), the expertise it should have, how you want it to communicate, and any context about yourself. A paragraph is enough to start.
- Save and test it. Give it a name, save the Gem, and open a new chat. Ask it a real question you'd actually want answered. Refine the instructions based on what it gets wrong.
After a few iterations, most Gems settle into something genuinely useful.
One thing worth knowing: you don't have to write the instructions yourself. Open a regular Gemini chat and describe what you want — "Write a Gem prompt for a personal finance advisor who knows my investment style" — and Gemini will generate the full system prompt for you. Copy it, paste it into the Gem instructions field, and you're done. The Gem architecture is the real value; the prompt is just a starting point, and you don't have to start from scratch.
What's Coming in This Series
- My Personal Doctor Gem: health questions, medication lookups, understanding lab results
- My Personal Lawyer Gem: contract review, legal language, IT-related legal questions
- My Investment Advisor Gem: portfolio thinking, market analysis, rebalancing decisions
- My Developer Gem: code review, architecture questions, debugging support
- My Network Engineer Gem: the one I'm most confident about, given 25 years in the field
- My Apple Device Expert Gem: managing personal Apple devices, troubleshooting, settings
- My DNS Expert Gem: narrow and technical, but surprisingly useful
Each post is standalone. Pick whichever topic is most relevant to you and start there.
Related Posts
- Google Gemini vs ChatGPT: Which AI Is Better for You? — A comparison of the base tools before going deeper into custom assistants
- How to Use Gemini in Gmail and Google Docs — The Google integration that makes Gems more powerful than standalone AI chat
- Free AI Tools You Should Start Using Right Now (2026 Guide) — Overview of what's free across major AI platforms
Sources & Further Reading
- Google — Create and use Gems in Gemini — Official guide to building and managing Gems
- OpenAI — Introducing GPTs — OpenAI's announcement of the GPTs feature and GPT Store
- Anthropic — Claude Projects — Overview of how Claude Projects work and what they're designed for
Have you tried building a custom AI assistant? Which platform did you use, and what did you build it for? I'd like to hear in the comments.
Comments
Post a Comment