AI Tools for Making Money Online: ChatGPT, Claude and Gemini Compared, Realistic Uses, Limits and Disclosure Rules

I've been making money online since 2010, which means I've lived through a lot of "this changes everything" moments: social media, smartphones, programmatic ads, crypto. Most of them changed something, just never quite what the hype promised. AI chatbots are the first one in a while that genuinely changed how I work every single day. When GPT-4 came out in 2023, I sat down to properly learn it and started pushing it into my daily work and my businesses, one routine task at a time. And they're also the biggest magnet for nonsense promises I've seen since the early "make $5,000 a week from your couch" ebooks.

So this is my honest take: which AI assistants are worth paying for right now, what they cost, what they realistically do for an affiliate, blogger or creator, where they fall flat, and which rules you have to follow so AI doesn't get you into trouble with the law, Google or YouTube.

What AI can actually do for your online income

AI doesn't make money on its own; it makes the work that makes money faster and cheaper. That's the whole game. If you already have something that earns, like a blog, a channel, a store or a freelance service, AI can help you produce more of it, test more ideas and handle the boring parts. If you have nothing yet, AI won't hand you a business, no matter what the guy in the rented Lamborghini says.

This isn't just my opinion. In a controlled experiment published in Science, Shakked Noy and Whitney Zhang gave more than 450 college-educated professionals real writing tasks from their own jobs. With ChatGPT, the average time spent fell by about 40% and the quality of the work rose by nearly a fifth. That's huge, but read it carefully: the people in the study already knew their jobs. AI made competent people faster. It didn't turn people into professionals.

Here's how I'd sort the typical tasks of someone who makes money online:

ChatGPT vs Claude vs Gemini: plans and prices compared

All three big assistants have a free plan and a main paid plan at about $20 a month; the expensive tiers ($100 and up) only make sense if you hit usage limits every day. Below are the US prices from each company's own pricing page. Prices can differ in other countries and currencies, and the model names change every few months, so treat the model column as a snapshot.

For teams there are business plans too: ChatGPT Business is $20 per user a month billed annually ($25 monthly, two users minimum), Claude Team Standard seats cost the same $20 annually or $25 monthly, and Gemini Enterprise Business edition is $21 per seat. For a one-person operation, the individual plans are enough.

So which one should you pick?

Honestly, at the $20 level they're closer than the fan wars on X suggest. My rough take, based on how they work in practice rather than benchmark charts:

• ChatGPT is the all-rounder with the biggest ecosystem: images, voice, custom GPTs and the cheapest step up from free (Go at $8). If you want one tool and don't want to think about it, start here.
• Claude is my pick for long writing and editing, working with big documents and code. Its text tends to sound less "AI-ish" out of the box, which matters if you write for a living.
• Gemini makes the most sense if you live in Google's world: Gmail, Docs, Drive, YouTube. Google AI Plus at $4.99 is the cheapest paid entry of the three.

My real advice: use the free plans of all three for a week on your actual tasks, the same prompt in each, side by side. Then pay for one. You'll learn more in that week than from any comparison article, this one included.

If you're enrolled at a college, check Google's student offer before paying anyone. In August 2026 Google announced a free year of Google AI Pro for eligible US college students and a free year of Google AI Plus for students in 140+ other countries. You'll need to add a card, so put a reminder in your calendar for month eleven; free years have a way of quietly becoming paid ones.

Realistic ways affiliates and creators use AI

The uses that pay off are boring: AI speeds up the work around a product or audience you already understand. Here's where I see it work, roughly in order of how fast it pays back.

Content research and drafting for review sites

For a blog or niche site, AI is a fast research assistant and a first-draft machine. Ask it to group keywords by intent, build an outline from the questions people ask, summarize a product's manual, or turn your voice notes from testing a product into a structured draft. What it can't do is test the product. The paragraphs that make a review worth reading are "here's what happened when I used it", and that has to come from you.

That's also where Google draws the line. Google's own guidance says generative AI is fine for research and structure, but using it to "generate many pages without adding value for users" may violate its scaled content abuse spam policy, and it tells publishers to fact-check and review all AI content before publishing. In plain words: AI-assisted is fine, AI-mass-produced is a ranking risk. If you're starting a site from scratch, my guide to affiliate marketing with no money covers the basics.

Ad creatives and landing page copy

Paid traffic lives or dies on creatives, and you need lots of them to find a winner. AI is perfect for volume: ten headline variants, five angles for the same offer, a landing page rewritten for a different audience. For images, a dedicated tool beats a general chatbot because it outputs ready ad sizes. AdCreative generates ad images and videos, scores creatives before you spend money on them, turns product photos into studio-style shots and shows competitors' best-performing ads. It has a 7-day free trial; the Starter plan is $39 a month (or $29 a month billed quarterly) with 20 credits and one brand. I've written a separate walkthrough on how to use AdCreative to advertise.

One warning from experience watching people do this: AI happily writes claims you can't back up. "Lose 10 pounds in a week" or "earn $500 a day" is exactly the copy that gets ads rejected and accounts banned. Check every claim before it goes live.

The quietly powerful use here is translation. When a campaign works in one country, AI can turn the ads and the landing page into German, Spanish, Italian or Polish in minutes, so you can test the same bundle in a new market without paying a translation agency up front. The model gets you most of the way; spend a few dollars on a native speaker to check the money page, because one awkward phrase in a headline reads as "scam" to a local. The same trick works for a store: product descriptions and answers to customer questions in the buyer's own language open markets you'd otherwise never touch. That's exactly how I use it in one of my own businesses, which has customers all over the world: a product description translated for a buyer from China, a customer's question answered in fluent French, each in minutes instead of a day of hunting for a translator. For landing pages, see how to build a high-conversion landing page.

YouTube and short-form video scripts

Scripts, hooks, titles, descriptions and chapter markers are where AI saves creators the most time. YouTube explicitly lists using AI to create or improve an outline or script, thumbnails and titles as production help that doesn't need disclosure. What does need disclosure is realistic synthetic content: a real person appearing to say something they didn't, altered footage of real events, or realistic scenes that never happened. More on monetizing a channel in how to make money on YouTube.

Email sequences

A welcome sequence, a product launch series, a re-engagement email: AI drafts them in minutes, and you then cut them down and add your personality. Email is one place where a generic AI voice really costs you, because people unsubscribe from boring senders. Use the draft as scaffolding, not as the final version. Where AI really earns its keep in email is personalization: describe two or three segments of your list (beginners, buyers, people who went quiet) and ask for a version of the same email for each. Writing three variants by hand is the kind of job nobody does; with AI it takes ten minutes, and an email written for a specific segment usually lands better than one generic blast. I use it this way myself: personalized replies and content built around what individual clients care about, and targeted campaigns per segment instead of one message for everyone. My take on why the list matters: email marketing is your personal ATM.

Freelance services

Freelancers use AI to deliver faster: copywriting, translation, product descriptions, social media posts. But the market has noticed. Clients pay less for plain text that a chatbot can write, and more for strategy, research, editing and results. If you sell writing, read my guide to getting paid for copywriting, which covers how AI changed rates. A trap freelancers learn the hard way: clients run your work through "AI detectors", and some of them trust the score blindly. Those tools are shaky enough that OpenAI withdrew its own AI text classifier in 2023, citing its low accuracy. So write in Google Docs or anything with version history, and if a client ever says "the detector says 80% AI", you can show them hours of edits instead of arguing with a percentage. If you're starting on Fiverr, see making money on Fiverr without skills.

Small tools and automations

This is the underrated one. Customer support is the easiest win if you sell anything: feed the AI your shipping, refund and delivery policies and let it draft answers to the questions you get every day, so buyers in other time zones get a reply in minutes instead of overnight. Keep a human check on anything involving money or angry customers. Support was one of the first things I handed to GPT-4 myself, along with setting up email campaigns and moderating comments: AI drafted the routine answers, and I kept the tricky ones. Even without coding skills, you can get an AI to write a script that pulls prices into a spreadsheet, a WordPress snippet, a simple calculator for your site, or an automation between your email tool and your CRM. Some people go further and build products on top of AI: I wrote about making money on MCP servers, small connectors that let AI assistants use other tools.

Once your prompt file works, stop pasting it. ChatGPT Projects, Claude Projects and Gemini Gems all let you store instructions and reference files permanently, so put your brand voice, your writing samples and your "never claim income or health results" rules there once. Then every new chat in that project already sounds like you, and you stop paying the "I forgot to paste the style guide" tax.

Where AI goes wrong

AI's biggest weakness isn't that it's bad at things. It's that it's confidently bad at things, in exactly the same tone it uses when it's right. The Harvard Business School study with Boston Consulting Group, "Navigating the Jagged Technological Frontier", showed both sides with more than 750 knowledge workers. On tasks AI handles well, people using it completed about 12% more tasks and finished about a quarter faster, with clearly better quality. On a complex task chosen to be outside AI's abilities, people using AI were 19% less likely to get the right answer than people without it.

Let that sink in: on the wrong task, AI made smart professionals worse. The researchers called the line between good and bad tasks "jagged" because you can't see it in advance. Here's where I see it bite people who make money online:

• Made-up facts. Commission rates, minimum payouts, fees, laws and statistics. The model fills a gap with something plausible. Every number in a money page has to be checked at the source.
• Outdated knowledge. A model learned from data up to a certain date. Ask it about an affiliate program and you may get terms from two years ago, or a program that has since closed.
• Same-sounding content. If you and a thousand other affiliates ask for "the best VPN review", you get roughly the same article. Readers notice, and so do search engines.
• No real experience. AI has never used the product, met your audience or lost money on a bad campaign. That's precisely the part people pay attention to.
• Private data. Pasting customer lists, contracts or login details into a chatbot is a bad habit. Check the privacy settings of whatever plan you use, and keep sensitive data out. Two switches are worth flipping before client work: "Improve the model for everyone" under Data Controls in ChatGPT, and "Help improve Claude" in Claude's Privacy settings. Both decide whether your chats can be used for training. For one-off sensitive stuff, ChatGPT's Temporary Chat isn't saved to your history, isn't used for training and is deleted within 30 days.

Two traps in your head: "I'm too late" and "AI will do it for me"

Most people get stuck with AI in one of two opposite ways, and both cost money. The first is fear: "everyone uses AI now, the internet is flooded, there's no point starting." The second is the fantasy: "AI will write the content, run the ads and I'll just collect commissions."

The fear first. Yes, there's more content than ever. But think about what the flood actually is: thousands of near-identical pages, written by the same few models, with nobody behind them. That makes a real person with real experience easier to stand out, not harder. Your tests, your screenshots, your opinions, your mistakes are the one thing AI can't generate. The practical answer to "I'm too late" is to start small: pick one niche, use AI for the grunt work, and put your own experience in every piece. A month of that teaches you more than a year of watching others.

Now the fantasy, which is more dangerous because people pay for it. US regulators have gone after a whole series of "AI-powered" income schemes. In its September 2024 sweep, Operation AI Comply, the Federal Trade Commission targeted, among others, Ascend Ecom, which allegedly promised passive income from "cutting edge" AI-powered online stores and took at least $25 million from consumers, and FBA Machine, which allegedly promised guaranteed income from AI-powered tools and cost consumers almost $16 million. Then-Chair Lina Khan put it bluntly: "there is no AI exemption from the laws on the books." In March 2026 the agency announced a settlement that would ban Air AI and its owners from marketing business opportunities, after it alleged they falsely claimed buyers would make substantial earnings.

The pattern is always the same: a big monthly income figure, the word "AI" doing the heavy lifting, and a price tag in the thousands. Here's how I'd protect yourself:

1. Ask what you're actually buying. A tool, a course or a "done-for-you" store? If the answer is vague, walk away.
2. Treat any income promise as a red flag. Real tools sell features. Scams sell outcomes.
3. Test the free version first. Every legitimate AI tool I know offers a free plan or a trial. The $20 chatbot already does most of what $3,000 "AI systems" sell.
4. Cap your spend. Decide your maximum before you look at any sales page, not after the webinar.

Disclosure rules: the law, Google, YouTube and Amazon KDP

Using AI is legal and mostly doesn't need disclosing; lying with it does. The rules you need to know come from advertising law and from the platforms you publish on.

The law. Three principles hold almost everywhere. Disclose affiliate links clearly next to the recommendation, whether a human or an AI wrote the text. Never publish fake reviews or testimonials, AI-generated ones included (in the US, the FTC's 2024 rule on reviews names them explicitly), and don't promise earnings or capabilities a tool can't deliver. Beyond that, follow the platforms' rules and the laws of the countries your ads target.

The platforms. Each one draws its own line between AI help and AI content:

YouTube says creators who consistently skip the disclosure may get a label applied for them or face penalties, including content removal or suspension from the YouTube Partner Program. It also says that disclosing doesn't limit a video's reach or monetization, so there's no reason to hide it.

A prompt workflow that doesn't sound like a robot

The quality of AI output depends far more on what you give it than on which model you use. One-line prompts get generic text. Here's the workflow I'd use for any money page:

1. Give it context. Who is the reader, what do they already know, what do you want them to do. "Write a blog post about VPNs" gets mush; "write for a freelancer who works from cafés and worries about public Wi-Fi" gets something usable.
2. Feed it your facts. Paste your test notes, the prices you checked, your opinions. Tell it to use only those facts and to mark anything it isn't sure about.
3. Ask for structure first. Get the outline, fix it, then ask for the draft section by section.
4. Edit like it's a junior writer's draft. Cut the filler, add your stories, check every number.
5. Ask it to criticize itself. "What would a skeptical reader object to here?" is one of the most useful prompts there is.
6. Make them check each other. My favorite cheap trick: paste the draft ChatGPT wrote into Claude (or the other way around) and ask, "List every number, price and factual claim in this text and mark which ones you can't confirm." Different models make up different things, so the second one is surprisingly good at catching the first one's confident nonsense, and it costs nothing on the free plans.

A few prompts worth keeping, rewritten from the generic ones you see everywhere:

• Research: "List the 15 questions a beginner asks before buying [product], grouped by buying stage. Don't answer them yet."
• Review draft: "Here are my test notes on [product]. Turn them into a review section in first person. Use only facts from my notes and mark gaps with [CHECK]."
• Ads: "Write 10 Facebook ad headlines for [offer] for [audience]. Five on pain, five on desired outcome. No income or health claims."
• Email: "Write a 3-email welcome sequence for subscribers who downloaded [lead magnet]. Short, conversational, one link per email."
• Support: "Draft a friendly reply to a customer asking for a refund on a digital product. Our policy is: [paste policy]."

Is a paid AI subscription worth it?

For anyone who writes, sells or creates online at least a few hours a week, a single $20 plan usually pays for itself; a stack of AI tools often doesn't. Run the numbers on your own situation rather than on hype.

Here's an example, not a result. Say a 2,000-word product review takes you five hours to research and write. If AI cuts that by 40%, in line with the Noy and Zhang experiment, you spend three hours instead and save two. At eight articles a month, that's 16 hours back for $20. Valued at even a modest hourly rate, that's an easy yes.

Now the tool stack. Say you pay $20 for ChatGPT Plus and $39 for AdCreative Starter, $59 a month in total, to promote an offer that pays $30 per sale. You need two sales a month just to break even on tools, before any ad spend. If your landing page converts 2% of visitors, that's 100 visitors a month just to cover subscriptions. That's achievable, but it shows why you add tools when a campaign needs them, not because a YouTube video said they're "essential". Plug your own numbers into the Earnings Calculator, and once ad spend is involved, the Campaign ROI Calculator.

What to do next

Start with the free plans, pick one task you do every week, like drafting emails or outlining posts, and run it through AI for a month. Measure the time saved honestly, pay for one plan when you hit the limits, and keep your own experience at the center of everything you publish. Disclose your affiliate links, never let AI invent reviews or income claims, and stay away from anyone selling "AI income" for thousands of dollars.

AI works best as one income tool among several, which fits how I think about multiple streams of income. And if you want honest context on the business side, read the truth about affiliate marketing.