I came to Claude out of necessity. A client project last year -- complicated brand strategy work for a healthcare company -- required me to process a lot of dense, contradictory information without getting confident-sounding nonsense back. ChatGPT kept giving me polished, wrong answers. Claude was the first AI that would actually say "I'm not sure about this part" instead of just... making something up with full conviction.
That's still why I use it.
But there's a learning curve if you're switching from ChatGPT, and Claude has a few features that aren't obvious from the outside. So here's what I actually wish someone had told me when I started.
What Claude Actually Is (And Who Builds It)
Claude is an AI assistant built by Anthropic -- not OpenAI, not Google. This distinction matters more than you'd think.
Anthropic was founded in 2021 by former OpenAI researchers, including Dario and Daniela Amodei, who left specifically because they wanted to focus more seriously on AI safety. That origin story shows up in how Claude behaves: it's designed to be honest about uncertainty, decline genuinely harmful requests, and avoid the kind of overconfident hallucinations that plague other models.
The current flagship model is Claude 3.7 Sonnet (as of early 2026), with Claude Opus 4 available on higher tiers. Each model generation has gotten noticeably smarter, but the character -- thoughtful, honest, slightly cautious -- has stayed consistent.
You access Claude at claude.ai. There's a mobile app too, but honestly I still do most of my real work on desktop.
Free vs. Pro vs. Teams -- What Each Tier Gets You
Free tier: You get access to Claude, but with message limits that you'll hit pretty quickly if you're doing substantive work. The limits aren't posted clearly anywhere (classic AI company behavior), but plan on running out mid-afternoon if you're using it heavily. Free also doesn't include extended thinking mode or Projects.
Claude Pro ($20/month): This is what I'd call the real tier. You get 5x more usage than free, access to all Claude models including Opus 4, extended thinking mode, and -- the big one -- Projects. If you're using Claude for work at all, this is table stakes. Same price as ChatGPT Plus.
Claude for Teams ($30/user/month, 5-user minimum): For actual teams. Adds a workspace where multiple people can share Projects, and claims not to train on your data. If you're processing client information, this is the one to think about. The data privacy angle is real and matters for professional use.
There's also an API for developers, but I'll get to that separately.
Your First Conversation -- How to Ask Claude Good Questions
OK so here's the thing about Claude: it responds really well to context. Way better than most people expect.
Bad prompt: "Write me a marketing email."
Better prompt: "I'm writing a marketing email for a B2B SaaS product that helps HR teams track employee wellness. The audience is HR directors at companies with 100-500 employees. Tone should be professional but warm -- not corporate. The email is announcing a new feature that lets managers see anonymized team wellness trends. About 200 words."
The difference in output quality is dramatic. Claude isn't magic -- it needs to know what you know. Treat it like briefing a very smart contractor who's new to your project, not like typing a search query.
A few things that help:
- Give it a role: "You're a UX writer with 10 years of experience in fintech apps..." Claude leans into this convincingly.
- Tell it the format you want: "Give me a bulleted list" or "Write this as a short paragraph, not headers" or "Use plain English, no jargon."
- Ask for drafts, not final: "Give me three different opening lines" gets you options; "Write the email" gets you one thing to argue with.
And when Claude hedges -- when it says "I'm not certain about this" or "you may want to verify this" -- take that seriously. It's not being falsely modest. That's actually useful information about confidence level.
Projects -- Claude's Version of Memory and Workspaces
This is the feature that makes Claude Pro actually worth it for professional use.
Projects let you create separate workspaces for different clients, topics, or tasks -- each with its own persistent memory and custom instructions. Think of it as giving Claude a permanent briefing document that it reads at the start of every conversation.
I have a Project for each major client. In each one, I've set:
- Background on the company, their tone of voice, their competitive landscape
- What I'm typically working on with them
- Writing style notes ("they hate the word 'leverage'" -- yes, actually)
The result is that I don't have to re-explain my client's whole situation every time I start a conversation. Claude already knows.
To create a Project: click "New Project" in the left sidebar, give it a name, and add instructions in the "Project instructions" section. You can also upload reference documents that Claude will reference throughout -- brand guides, style docs, whatever's relevant.
Projects also keep your conversation history organized by client/topic, which sounds minor but is actually a big quality-of-life improvement over one giant chat history.
Uploading Files and Documents to Analyze
Claude can read documents. PDFs, Word docs, CSVs, images, code files -- drag them into the conversation window or click the attachment icon.
This is genuinely useful for:
- Contract review: "Flag any clauses that seem unusual or one-sided from a vendor perspective." Claude won't replace a lawyer, but it'll catch things you'd skim past.
- Research synthesis: Upload 4-5 papers or reports and ask Claude to identify common themes or contradictions.
- Data interpretation: Drop in a CSV and ask for a plain-English summary of what the data shows.
- Editing your own work: Paste in a draft (or upload a doc) and ask for specific feedback -- not "is this good" but "does the argument in section 3 logically follow from section 2?"
The file size limit is generous but not unlimited -- very long documents may need to be chunked. And Claude doesn't retain files between conversations unless you're in a Project and have uploaded them to the Project library.
One thing I notice as a designer: Claude reads visual layouts in images surprisingly well. I've uploaded wireframes and asked for UX feedback. The results are hit or miss, but it's worth trying.
Claude's Extended Thinking Mode (When to Use It)
Extended thinking is Claude's "let me actually reason through this" mode. Instead of generating a response immediately, it works through the problem step by step -- you see the reasoning process before the final answer.
When to use it: anything where you want rigor over speed. Complex analysis, multi-step problems, questions where you suspect there might be subtleties a quick answer would miss.
When NOT to use it: quick tasks, creative brainstorming, anything conversational. It's slower and you don't always need the extra reasoning for straightforward tasks.
You toggle extended thinking by clicking the "Extended thinking" option before sending a message (it's in the model selector area). It's only available on Claude Pro.
My honest take: it's noticeably better for analytical tasks. I used it for a competitive analysis last month and the output quality was meaningfully higher than the same prompt without extended thinking. The tradeoff is time -- sometimes it takes 60-90 seconds to start responding.
When Claude Beats ChatGPT (And When It Doesn't)
I use both. Not switching costs -- actual preference depends on what I'm doing.
Claude wins for:
- Long-form writing where voice and tone matter
- Anything requiring nuance or acknowledging complexity
- Document analysis and synthesis
- Keeping context over a long conversation (Claude handles this better than most)
- When you need it to say "I don't know" instead of making something up
ChatGPT wins for:
- Using third-party plugins and integrations (GPT has a much bigger ecosystem)
- Image generation (Claude doesn't do images)
- Real-time web browsing (both have it now, but ChatGPT's feels more integrated)
- Tasks where speed matters more than depth
See our full Claude vs. ChatGPT comparison for the head-to-head breakdown, and Claude vs. ChatGPT for writing specifically if that's your main use case.
The thing people get wrong: these aren't identical tools where one just happens to be better. They make different tradeoffs. Claude is more cautious, more honest about uncertainty, and better at sustained reasoning. ChatGPT is more plugged-in to external services and more willing to just give you an answer (for better or worse).
The Claude API -- What It Is, Do You Need It?
The API is how developers programmatically access Claude -- building apps, automating workflows, integrating Claude into their own products. If you're not a developer, you don't need it.
If you are a developer: Anthropic's API is competitive with OpenAI's. Claude 3.7 Sonnet is fast and affordable for most use cases; Opus 4 is the premium option for tasks requiring maximum capability. The pricing is per-token (input and output), similar structure to OpenAI.
One legitimate reason a non-developer might think about the API: if you're building something with a no-code tool like Make or Zapier that supports API calls, you can integrate Claude into automated workflows. Worth exploring if you want Claude in your existing systems without paying for Claude for Teams.
Tips for Getting Better Responses
These are the things that actually changed my results, not theoretical advice:
Be specific about format. "Explain this to me" gives you an essay. "Explain this in 3 bullet points for someone with no technical background" gives you something actually useful.
Iterate, don't regenerate. Instead of hitting regenerate when you don't like a response, reply with what specifically didn't work. "The second point is too vague -- can you give a concrete example?" This works better than starting over.
Use it for second opinions. Paste in a plan, a document, or an argument and ask "what am I missing?" or "steelman the opposite view." Claude is unusually good at this without just validating everything you say.
Don't over-prompt. I've found that 2-3 sentences of good context beats a 500-word prompt. Claude doesn't need to be micromanaged -- give it enough to work with and see what it does.
Save your best prompts. If you find a prompt structure that works for something you do regularly, keep it in a doc. Your Claude Projects can have prompt templates in the instructions section.
For more on AI writing tools and where Claude fits in the broader landscape, the best AI writing tools roundup is worth reading if you're still evaluating options. And if you're running into specific error messages or issues with Claude, the Claude AI common errors guide covers the most frequent problems and how to fix them.
The thing I keep coming back to with Claude: it feels like working with something that's actually trying to be useful rather than impressive. It'll tell you when it doesn't know. It'll push back if your question has a false premise. It won't write you a confident-sounding three-paragraph answer about something it doesn't actually understand.
That's not the most exciting pitch. But for professional work where accuracy matters -- it's the thing that keeps me using it.
Top comments (0)