It was around 11 pm on a Tuesday, and I had three client blog posts due by morning. I’d already rewritten the intro to the first one four times. My coffee had gone cold an hour earlier, and I didn’t notice until I took a sip and nearly spat it back out.
A friend who does freelance dev work kept telling me to “just use Claude for the first draft.” I’d brushed it off for months. It felt like cheating, or like it would spit out generic garbage I’d have to rewrite anyway.
That night, out of pure desperation, I finally tried it.
I’m not going to say it changed my life overnight. My first real attempt at using it for paid work was rough, and I made a mistake that still makes me cringe a little (more on that below). But months later, Claude is genuinely part of how I earn a living now. Not as a replacement for me, but as the fastest, most patient junior assistant I’ve ever worked with.
If you’re wondering whether you can actually make money with Claude AI, or if it’s just another overhyped tool people talk about online, here’s the real version. Mistakes included.
Quick reality check before you get excited
Nobody hands you money for opening a Claude tab. There’s no button that says “deposit $500 here.”
What Claude actually does is make you faster at things people already pay for: writing, coding, customer support, research, and admin work. You still need a real skill, a useful service, or a quality product to make it valuable.
Anyone telling you this is passive income while you sleep is selling you something. Every method below still takes real hours, especially at the start.
What actually worked for me
1. Freelance writing (my bread and butter)
This is the one that pays my rent, so let’s start here.
I use Claude.ai to draft blog posts, product descriptions, and email sequences for clients I find on Upwork and Contra. The free plan is fine for testing this out. I moved to Pro (around $20 a month) once I was using it daily and kept bumping into limits.
Here’s roughly how I work:
- I set up a Project in Claude for each client, and drop in a few of their past posts, their style guide if they have one, and notes on their audience.
- I ask Claude for a rough draft based on an outline I wrote myself, not the other way around. If you let it pick the structure, everything starts sounding the same.
- I rewrote the intro and conclusion almost entirely by hand, since those are where “AI voice” shows up the most.
- I fact-check every number, date, and claim before it goes anywhere near a client.
That last step exists because of my most embarrassing early mistake. I once sent a client a draft with a very confident, very specific statistic in it. It sounded right. It wasn’t. Claude had essentially made it up, and I hadn’t double-checked because I was in a hurry. The client caught it before publishing, thank goodness, but it was an uncomfortable email to write.
Rates for this kind of work vary a lot by niche, but something like $40 to $150 per blog post is a pretty normal starting range on Upwork, climbing higher once you’ve got a portfolio and repeat clients. Nobody’s getting rich on volume alone, but it adds up.
2. Building tiny tools with Claude Code, even though I’m barely a coder
This one surprised me. I’m not a developer. I took one Python class in college and forgot most of it.
Claude Code is Anthropic’s agentic coding tool. You can run it from a terminal, the desktop app, or even the mobile app, and describe what you want in plain English. It writes, tests, and fixes the code itself.
I used it to build a simple invoice generator for a friend who runs a small landscaping business. It took an afternoon, not a computer science degree. He now pays me a small monthly fee to maintain it.
I’ve also used it to build little Chrome extensions and WordPress tweaks for other freelancers who don’t want to learn to code themselves. If spreadsheets are more your speed than code, Claude’s Excel and PowerPoint tools do a similar job on the business side, cleaning up a messy budget or turning rough notes into a client-ready deck in an afternoon instead of a week.
3. Chatbots for local businesses (the one that surprised me most)
A few months in, I pitched a local bakery on a simple FAQ chatbot for their website: hours, allergen info, custom order policy, that kind of thing.
This one uses the Claude API rather than the regular Claude.ai chat subscription. You need to sign up separately at platform.claude.com, which is Anthropic’s developer platform. Then add a small amount of billing credit and create your API key. It’s pay-as-you-go per message, and if you use the cheapest model (Haiku) for something this simple, the monthly running cost for a small business is usually just a few dollars.
My first version of that bakery bot was too confident for its own good. A customer asked about a discount that didn’t exist, and the bot politely made one up. I hadn’t restricted what it was allowed to talk about, so it improvised. I rebuilt it with a much narrower set of instructions and a clear “I don’t know, here’s the owner’s number” fallback, and it’s been solid since.
If you’re not comfortable coding this yourself, tools like Zapier now let you connect Claude to thousands of other apps without writing much code at all, which is a decent starting point.
4. Turning one client call into a week of content
This is the laziest one, in the best way.
I take a transcript from a podcast episode, webinar, or client Zoom call, drop it into a Claude Project, and ask for a week’s worth of LinkedIn posts, a short newsletter blurb, and a handful of social snippets, all pulled from that single hour of talking.
I charge this as a small monthly retainer for a couple of clients who’d rather talk than type. It’s one of the easier services to sell, because most business owners hate writing social posts but love talking about their own work.
5. Digital products and templates
I’ve dabbled in this less, but it’s real. People use Claude to help draft and organize ebooks, Notion templates, and worksheets, then sell them on Gumroad or Etsy.
The honest caveat: this market is crowded, and “AI helped me write it” isn’t a selling point to buyers. What sells is a specific, well-organized answer to a specific problem for a specific audience. Claude just makes the drafting faster, not the idea better.
6. Selling Claude skill packs (the newest one, and my favorite recent discovery)
This is an angle I only stumbled into recently. Claude Cowork and Claude Code both support “Skills” and plugins, basically packaged instructions that teach Claude how to handle a specific, repeatable task really well.
People are building and selling niche skill packs for specific industries, things like a contract-review pack for real estate agents or a customer-service pack for e-commerce stores, then selling them as a one-time download through their own site or Gumroad, separate from whatever Claude plan the buyer already pays for.
I built a small one for freelance photographers (client intake, contract reminders, gallery delivery emails) and sell it as a one-time purchase. It’s not a huge earner yet, but it’s the closest thing on this list to actual passive income, since I built it once and it keeps selling on its own.
If you want to actually start this week
- Sign up for a free Claude.ai account and use it for a few days before spending anything.
- Pick one method above that matches a skill you already have. Writers should start with #1, anyone comfortable poking at software should try #2, business-minded folks should look at #3 or #6.
- Set up a profile on the relevant platform (Upwork, Fiverr, Contra, or Gumroad) and do two or three practice pieces before you pitch anyone.
- Price modestly at first, then raise your rates once you have a couple of happy clients and real proof it works.
- Upgrade to a paid plan only once you’re actually hitting the free plan’s limits, not before. Pricing and usage caps shift fairly often, so it’s worth double-checking claude.com/pricing before you budget around it.
Mistakes that cost me time (and a little embarrassment)
Publishing raw output without editing. It reads generic, and clients notice the pattern faster than you’d think. Always rewrite the sections that reflect your own voice and personal style.
Not fact-checking anything. Claude sounds confident even when it’s wrong, so treat every number, quote, and claim as unverified until you’ve checked it yourself.
Competing purely on speed. I underpriced myself early on because I could turn work around fast, which just meant more work for the same money. Speed is a nice bonus, not the whole pitch.
Skipping the fine print. Some client contracts and freelance platforms have specific rules about disclosing AI-assisted work. Read them before you start, not after a client asks why you didn’t mention it.
Deploying anything client-facing without testing the weird edge cases first. My bakery bot taught me that one directly, and it’s an easy step to skip when you’re excited to launch.
Cramming everything into one giant conversation. Separate Projects per client or per task, keep answers consistent instead of muddled.
So, is it actually worth it?
Honestly? Yes, for me, but not because Claude is magic. It’s because it removed the boring, slow parts of work I was already doing and let me take on more of it without burning out completely.
If you’re expecting a side hustle that runs itself, this probably isn’t it. If you’re willing to pair a real skill with a tool that’s genuinely good at tedious work, there’s real money here.
Start small. Pick one thing from this list, and see if it actually fits how you work before you build a whole business plan around it.
