How I Use AI as a Tool to Write Articles (Not as a Replacement)

Let me be upfront about something.

Every article I publish goes through AI at some point in the process. But the articles that rank, get cited, and actually build trust with readers? Those are the ones where AI did maybe 20% of the work and I did the rest.

I've been writing content for affiliate sites since 2013. I started using AI writing tools seriously around 2021 — Jasper AI first, then Frase, then ChatGPT, and more recently Claude. I've tested most of the major tools. And after all of that experimentation, my conclusion is the opposite of what most "AI writing" guides will tell you.

AI alone produces content nobody wants to read.

Not because it's poorly written. Because it has no point of view, no real experience behind it, and nothing a reader couldn't get from asking the AI themselves. In 2026, that's a death sentence for a piece of content.

Here's how I actually use AI to write articles — and why the human part is the only part that matters.

Why Pure AI Content Fails in 2026

According to Ahrefs, 67% of sources cited in Google's AI Overviews don't even rank in the top 10 organic results for the query. And 67% of ChatGPT's most cited pages come from original research or first-hand data.

Read that again. The content getting cited by AI systems isn't the content that's perfectly optimized — it's the content with real, specific, verifiable information from people who actually did the thing.

An AI can write a technically correct article about affiliate marketing. But it can't tell you that my best affiliate income month was $24,862, or that I lost significant traffic when Google's algorithm changed in early 2024, or that I've personally bought and tested over 40 courses. That first-hand data is what gets cited. That's what builds trust.

Pure AI content has none of that. Which is why it doesn't rank and doesn't get cited.

My Actual Process

I use AI at three specific points in the writing process. Everything else is me.

Step 1 — Research and Topic Validation

Before I write a single word, I use AI to understand what questions people are actually asking about a topic.

I'll ask ChatGPT or Claude something like: "What are the 10 most common questions beginners have about affiliate marketing?" Not to copy the answers — to understand the landscape. I then cross-reference this with Google's "People Also Ask" boxes and check what's ranking in the top 10.

This research phase used to take me two hours. AI cuts it to about 20 minutes. That's the legitimate time-saving value.

I also check Google Trends for the topic to confirm it's growing or stable — not declining. There's no point writing a comprehensive guide on something people are losing interest in.

Step 2 — Structure and Outline

Once I know what I'm writing about, I'll ask AI to suggest a logical structure. Not headings — structure. Something like: "What's the logical order for explaining how to start an affiliate blog to a complete beginner?"

I take the output, rearrange it based on my own experience of what actually matters, cut what's irrelevant, and add sections the AI missed entirely — usually the nuanced stuff that only comes from real experience.

The AI gives me a starting point. I make it reflect what I actually know.

Step 3 — First Draft of Difficult Sections

There are sections in every article that are genuinely tedious to write — definitions, background context, "what is X" explanations. These are also the sections readers skim past. For those, I'll let AI generate a draft and then edit it into my voice.

For everything else — my opinion, my experience, what worked, what didn't, specific numbers and results — I write that myself from scratch. No exceptions.

What I Write Myself — Every Time

This is the part most AI writing guides skip.

My personal experience and results. When I write about affiliate marketing, I include real numbers — my $24,862 best month, the sites I've built, the courses I've actually bought and tested. This can't be faked and it's exactly what Google and AI systems are now trained to look for and cite.

My opinions and recommendations. When I say eCom Elites is better value than Amazing Selling Machine, that's a judgment call based on having paid for and gone through both. AI can't make that call. A reader asking "which course should I buy" needs a human answer, not a hedged AI summary.

Anything that requires nuance. "Is dropshipping dead?" needs a real answer from someone who's actually done it — not a balanced AI summary that says "it depends." I write those sections from scratch every time.

The intro and conclusion. These are where a reader decides whether they trust you. An AI-written intro reads like an AI-written intro. Readers can feel it, even if they can't articulate it. I always write these myself.

The Tools I Actually Use

ChatGPT / Claude — for research, brainstorming, and drafting sections I'll heavily edit. I use both depending on the task. Neither one writes publishable content without significant human editing.

Originality AI — I run every article through this before publishing. Not because I'm worried about getting caught, but because it tells me where the AI influence is too high and I need to add more of my own voice. If a section scores too high for AI probability, I rewrite it. This tool has made my content measurably better.

👉 If you'd rather have most of these tools in one place instead of juggling separate subscriptions, AIVille is worth looking at. It bundles AI writing, image generation, content tools, and more under one roof.

I reviewed it in detail — it's a solid option if you're just getting started with AI tools and don't want to piece together a stack from scratch.

Grammarly — free version, every article. It catches mistakes I miss after staring at the same text for too long.

Jasper AI — I used this heavily in 2021-2022 and it taught me a lot about how to work with AI output. Less central to my workflow now but still useful for specific short-form tasks like meta descriptions and email subject lines.

The Structural Format That Gets Cited

From Matt Diggity's research on AI visibility: 72.4% of pages cited by ChatGPT contained a short, direct answer immediately after a question-based heading. The format is simple — heading states the question, first sentence answers it directly, following paragraphs add context.

I now write every article with this structure in mind. Not just for AI citation potential — it's also better writing. Readers who land on a specific question want the answer fast. Give it to them in the first sentence, then explain it.

This is different from the way most blog posts are written — the "build up to the answer" approach that makes readers scroll through paragraphs before they get what they came for. Kill that habit.

Content Freshness Matters More Than You Think

According to Ahrefs, AI assistants prefer content that's 25.7% fresher than what shows up in organic search. On Perplexity, freshness accounts for roughly 40% of ranking factors.

This means updating articles regularly with real changes — not just changing the date. When I update an article, I'm adding new data, updating tool recommendations, removing outdated information, and adding a section on something that's changed since I first wrote it.

I add a "Last Updated" date at the top of every article for exactly this reason. Google and AI systems both use this signal.

The Test Before You Publish

Before I publish anything, I ask myself one question: could an AI have written this?

If the honest answer is yes — if there's nothing in the article that requires me to have actually done the thing, made the money, bought the course, or run the test — I don't publish it. I go back and add the parts that only I can write.

That's the standard. It's the only one that matters in 2026.

For a deeper look at how AI fits into making money online more broadly, my guide on how to make money with AI covers 14 different approaches beyond just content writing.

Frequently Asked Questions About Using AI to Write Articles

Should I use AI to write my blog posts?

Use AI as a research and drafting tool, not as the writer. AI handles research, outlines, and tedious background sections well. But your opinions, personal experience, real numbers, and recommendations need to come from you. Content that could have been written entirely by AI doesn't rank and doesn't get cited by AI systems.

How much of my article should AI write?

In my process, AI does about 20% of the work — research, structure suggestions, and first drafts of background sections I'll heavily edit. The other 80% is me: the intro, conclusion, opinions, real data from my own experience, and anything requiring genuine judgment. That ratio is what produces content worth reading.

Will Google penalise AI-written content?

Google doesn't penalise AI content directly — it penalises low-quality content that lacks genuine expertise and first-hand experience. An article full of AI-generated text with no real opinions, no specific data, and no human perspective will struggle to rank regardless of how well it's written. The fix is adding the parts only you can write.

What AI tools do you actually use for writing?

ChatGPT and Claude for research and drafting sections I'll heavily edit. Originality AI to check where AI influence is too high before publishing. Grammarly free version for catching errors. Jasper AI for short-form tasks like meta descriptions. I run every article through Originality AI before hitting publish — it tells me where I need to add more of my own voice.

What is the best format for content that gets cited by AI?

According to research, 72.4% of pages cited by ChatGPT contain a short direct answer immediately after a question-based heading. The format is: heading states the question, first sentence answers it directly, following paragraphs add context. Write every section this way and you significantly improve your chances of being cited in AI-generated answers.

How do I know if my content is too AI-heavy?

Ask yourself one question before publishing: could an AI have written this? If there's nothing in the article that requires you to have actually done the thing, made the money, bought the course, or run the test — go back and add those parts. That's the standard that matters in 2026.

Related Articles