Every writer I know is using AI in some form. Most won't admit it. A few are loudly proud of it. Almost nobody is talking honestly about what it's actually good for and where it falls apart.

I've been experimenting with AI writing tools for over two years now. I've used them to draft articles, brainstorm headlines, outline structures, research topics, and edit prose. Some of these use cases are genuinely transformative. Others produce the kind of hollow, committee-written content that readers scroll past in half a second.

Here's what I've learned about how to use ai for writing without losing what makes your voice yours.

What AI is actually good at

Let me be specific. Not vague "AI can help with writing" statements. Specific use cases where AI tools save me real time every week:

Outlining. I give the AI a topic and my angle. It returns a structure — sections, subpoints, logical flow. I disagree with 40% of the suggestions and rearrange the rest. The value isn't the outline itself — it's having something to react to instead of starting from a blank page. Reacting is faster than creating from nothing.

Research summaries. I need to understand a topic quickly — say, the latest Google algorithm changes or how a specific platform's monetization works. AI can synthesize information from multiple sources into a readable briefing in seconds. I still verify everything, but the initial synthesis saves 30 to 60 minutes per article.

Headline variations. I write a headline. I ask the AI for 15 variations. Most are worse than mine. Two or three are genuinely better. That's a net positive in under a minute.

Editing for clarity. I paste a paragraph that feels clunky. I ask the AI to simplify it. The result is usually too bland, but it shows me which sentences were unnecessarily complex. I rewrite them myself, informed by the AI's simplification. The final version is mine, but cleaner.

SEO suggestions. AI is surprisingly good at identifying related keywords, suggesting H2 structures that match search intent, and spotting gaps in an article's coverage relative to what's ranking on Google.

What AI is terrible at

Equally specific:

Voice. AI doesn't have one. It has a default mode — slightly formal, relentlessly balanced, allergic to strong opinions. Every AI-generated paragraph sounds like a press release reviewed by a committee. If you use AI to write your prose, you'll sound like every other writer using AI to write their prose. That's the opposite of building an audience.

Personal experience. AI can't write "I tried this for six months and here's what happened." It can only write "Many writers have found that..." The difference between those two sentences is the difference between content people share and content people scroll past.

Specificity. Real writing uses specific numbers, dates, names, anecdotes. AI writing uses generalities. "Many successful writers" instead of "Tim Ferriss." "A significant increase" instead of "a 340% jump in three months." Readers trust specificity. AI defaults to vagueness.

Taste. Knowing what to leave out is as important as knowing what to include. AI includes everything. It hedges every claim. It adds qualifiers to every statement. Good writing is confident, selective, and opinionated. AI is none of those things by default.

Humor and surprise. AI can attempt humor. It's never funny. It can attempt surprise. It's always predictable. These are qualities that make writing memorable, and they're exactly what AI can't replicate.

The right mental model

Here's how to use ai for writing without it ruining your work: think of AI as a research assistant and a first-draft accelerator, never as a co-writer.

A research assistant gathers information, organizes it, and presents options. You make the decisions. You choose the angle, the structure, the examples, the voice. The assistant does the legwork. You do the thinking.

The moment you let AI make creative decisions — which stories to tell, which metaphors to use, what your opinion is — your writing becomes indistinguishable from everyone else's AI-assisted writing. And that's a lot of writing now. Standing out requires the opposite: more human judgment, not less.

My actual AI writing workflow

Here's exactly how I use AI when writing an article:

Before writing: I use AI for keyword research, topic validation, and competitive analysis. I ask it to summarize what's currently ranking for my target keyword and identify what those articles miss. This takes 10 minutes and replaces an hour of manual research.

Outlining: I describe the article I want to write — the topic, my angle, and three to four points I know I want to make. I ask AI for a suggested outline. I take 30% of what it suggests, throw the rest away, and add my own sections. The outline is mine, informed by the AI's suggestions.

First draft: I write this myself. Entirely. This is where voice lives, and voice is everything. If you outsource the first draft to AI, you're outsourcing the only part that matters. I might dictate it using voice-to-text tools, but the words are mine.

Editing: After writing, I'll sometimes run sections through AI to check for clarity issues, logical gaps, or sentences that are unnecessarily long. I use the feedback. I don't use the AI's rewrites — they always strip out the personality.

SEO check: I ask AI whether my article covers the main subtopics someone searching my keyword would expect. If I'm missing something important, I add a section. If I've covered everything, I move on.

Total AI time per article: maybe 30 minutes. Total writing time per article: two to three hours. The AI accelerates the process by about 30%. It doesn't replace the process.

The tools I use

I keep it simple. Claude for research, brainstorming, and editing feedback. ChatGPT occasionally for different perspectives on the same prompt. That's it. I don't use Jasper, Copy.ai, or any of the dedicated "AI writing" tools — they're optimized for generating content, not for assisting writers. There's a meaningful difference.

For writers who want a more integrated workflow, I've been building Superwriter — a writing tool designed specifically for writers who use AI as an assistant, not a replacement. Clean interface, no bloat, built for the workflow I described above.

How to keep your voice when using AI

This is the real question. Here's what works:

Never publish AI-generated text without rewriting it. Not editing — rewriting. If the AI writes a paragraph and you tweak two words, it's still AI writing. If the AI writes a paragraph and you rewrite it in your voice using the same underlying point, it's your writing informed by AI. The difference is everything.

Write your opening and closing yourself. The first and last paragraphs are where your voice is most concentrated. Readers decide in the first three sentences whether your writing feels real or generated. Never outsource the intro.

Add what AI can't. Personal stories. Specific numbers from your own experience. Opinions that might be wrong. Jokes that might not land. These are the signals that tell a reader a human wrote this — and they're exactly what AI strips out when it "improves" your text.

Read it out loud. AI writing sounds fine silently. Read aloud, it sounds mechanical. If a sentence doesn't sound like something you'd say to a friend over coffee, rewrite it. This is the fastest test for whether your voice is intact.

Develop a style guide AI can't replicate. Short sentences. First person. Specific numbers instead of "many" or "often." Starting paragraphs with "Here's the thing" or whatever your verbal tics are. These quirks are your voice. Keep them.

The ethics question

Should you disclose AI use? My answer: it depends on what you used it for.

If AI helped you research, outline, and edit — that's no different from using a human editor or research assistant. No disclosure needed. Those are tools in your process, and the final output is yours.

If AI wrote significant portions of the final text — yes, you should disclose. Not because of rules (though some platforms are adding them), but because your readers trust you to be the writer. Breaking that trust has consequences.

The practical test: if you removed all AI-generated text from your article and it no longer existed, the AI wrote it. If you removed all AI-generated text and the article was still 80% intact, you wrote it with AI assistance. Those are different things, and readers know the difference even when they can't articulate it.

Where this is heading

AI tools will keep getting better. The writing they produce will keep getting harder to distinguish from human writing — at least on a surface level. That means the value of genuine human voice, specific experience, and original thinking will increase, not decrease.

The writers who thrive will be the ones who use AI to handle the commodity parts of writing — research, structure, optimization — and invest the saved time into the parts AI can't do: original thinking, personal storytelling, and the kind of strong opinions that make readers either love you or disagree with you. Either reaction is better than the indifference that generic AI content produces.

Learning how to use ai for writing is a skill worth developing. But it's a support skill, not the main skill. The main skill is still writing — having something to say and saying it in a way that only you can.

For more on tools that support the writing process without taking it over, see my complete writing toolkit. And if you're just starting out, my practical guide to writing online covers the fundamentals before you add AI into the mix.

A writer is nothing without a reader. If you found this helpful, consider becoming my dear email friend. Nothing would make me happier.