Choosing an AI assistant
Choosing between ChatGPT, Claude, and Gemini works best when you start with your own situation — what problem you are solving, how you work, and what tools you already use — rather than with a feature comparison. This job aid sets up an AI advisor that asks about your use case, your team, and your budget, then answers your specific questions and verifies each feature claim against the provider's own documentation.
Setting up your advisor
Set up a custom GPT or a Gem with the role of an unbiased advisor helping you choose between ChatGPT, Claude, and Gemini. Tell it to verify any feature claim against the provider's own documentation and to cite the source and date. If you don't want to build a custom assistant, you can give the same instructions in a normal chat window.
Questions to ask your advisor
Give your advisor your context first — what problem you're solving, whether you work solo or on a team, which tools you already use (particularly Google Workspace), your comfort level with AI, and your budget. The advisor can guide you through a structured decision or answer specific feature questions directly. For example:
- I use Google Workspace daily. How can I connect AI to my Gmail?
- I want to automate parts of my workflow. Can you help me compare automation options?
- What are Claude skills and what can I use them for?
- I work with sensitive client information. What are the privacy and security differences?
- I love ChatGPT Projects for keeping context... do Claude and Gemini do anything better or different with their Projects feature?
Verify before you commit
Before you rely on any feature claim — this page's, the advisor's, or a comparison table's — check it against the provider's own documentation and note the date you checked. If the documentation and the claim disagree, trust the documentation.
Official sources:
- Anthropic: anthropic.com/news, support.claude.com
- OpenAI: openai.com/index, help.openai.com
- Google: blog.google/products/gemini, ai.google.dev
Remember, as always, that AI can make mistakes. Verify information, especially before you take out your credit card.
Related pages
- prompting-frontier-models — How to write structured prompts that get reliable results from any AI assistant
- building-agents-that-hold-up — How to evaluate whether a task is worth automating with an AI agent, and how to build one that works