I use AI every day. I’ve written two free guides on the subject, one covering how to use AI for writing and one covering where AI fails. I’ve built an entire library of 46 handbooks that integrate AI into writing craft. I’m not anti-AI. I’m anti-stupidity about AI, and there’s an enormous amount of stupidity circulating about what AI can and cannot do for writers.
The short version: AI is a research assistant with encyclopedic knowledge and no memory. A brainstorming partner who never gets tired. An editor who spots patterns you’ve stopped noticing. A collaborator who generates variations faster than you can evaluate them. Used well, it makes you a better, faster writer. Used poorly, it produces generic slop indistinguishable from the thousands of other AI-assisted projects flooding the market.
The difference between those outcomes isn’t the tool. It’s whether the writer understands what the tool actually does.
What AI Does Well
The value is in AI breaking you out of the loop your own brain is stuck in.Share on X
AI excels at brainstorming. When you’re stuck on a plot problem, a character decision, or the structure of a chapter, AI can generate twenty possible solutions in the time it takes to finish your coffee. Not all of them work. Most won’t, but buried in the pile you’ll find approaches that never occurred to you during solo brainstorming. The value isn’t in AI’s answers. The value is in AI breaking you out of the loop your own brain is stuck in. For more, see logical fallacies every writer should recognize and What I Learned in Pascal in 1980 That Every AI Executive Should Know.
AI excels at research support. Need to understand medieval siege warfare for a fantasy battle? AI explains catapult mechanics without forcing you to become a military history scholar. Writing contemporary fiction set in a profession you don’t know? AI provides enough insider knowledge to sound authentic without requiring a career change. It eliminates the Wikipedia rabbit holes that destroy entire writing afternoons. You know the pattern: start looking up one historical detail, click an interesting link, follow another connection, and suddenly you’re reading about 16th-century bread recipes when you only needed to know what people wore to funerals.
AI excels at revision analysis. Feed it a chapter and ask specific questions. Does the pacing drag in the middle section? Is the dialogue between these two characters distinct enough? Does this scene accomplish what I need it to accomplish? AI catches patterns you’ve become blind to after reading your own work fifteen times. It spots repetitive sentence structures, overused words, and sections where the energy drops.
AI excels at tedious but necessary tasks. Grammar checking, formatting consistency, citation management, summarizing research notes, organizing complex information. These tasks eat hours without contributing to the creative work. AI handles them quickly and accurately enough to free your time for the writing that actually matters.
Where AI Fails
AI cannot write in your voice. It writes in AI voice. The output is technically correct, grammatically clean, and emotionally flat. It sounds like it was generated by a committee of English teachers who learned writing from textbooks, not life experience. The words are accurate, the grammar is perfect, and the result feels as engaging as reading appliance warranties.
I spent months testing AI’s limitations when Claude first became available. My twenty years as Director of Computer Operations at Trader Joe’s taught me that you cannot trust a system until you know how it breaks. The same principle applies here. You cannot use AI well until you understand how it fails.
AI struggles with nuance. Irony often flies past it entirely. Sarcasm gets interpreted as genuine enthusiasm. Fresh metaphors get misread or interpreted literally. Cultural context creates gaps that produce errors no human writer would make. If your writing depends on subtext, implication, or the gap between what characters say and what they mean, AI will miss it.
AI drifts in long sessions. Context that was established clearly at the beginning of a conversation degrades over time. Character details shift. Tone wanders. Established facts get contradicted. The longer you work in a single session, the less reliable the output becomes. That’s a characteristic of how these systems work, and no fix is coming.
AI is confidently wrong. It generates plausible-sounding information that’s factually incorrect and presents it with the same confidence as accurate information. It invents book titles, fabricates quotes, misattributes ideas, and creates fictional experts. Every factual claim AI makes needs verification. Every quote needs checking. Every reference needs confirmation. Writers who skip this step publish embarrassing errors.
AI produces generic output when asked to write prose. The sentences are correct. They’re also bland, predictable, and indistinguishable from every other AI-generated paragraph on the internet. Common patterns include overuse of certain transitional phrases, balanced sentence structures that never vary, hedging language that avoids strong positions, and a certain smoothness that lacks the rough edges of human writing. Readers increasingly recognize these tells even when they cannot articulate what feels off.
What Is the Hybrid Approach to AI Writing?
From the podcast
Martin Ricketts runs the same sequence and has a shorthand for it, described on Leaders and Their Stories:
The way I do it’s AI, human, AI, AI, human. So the human does it first.The order matters more than the ratio. Whoever goes first sets the voice, and if that’s the machine, everything after it is editing instead of writing.
The best AI-assisted writing treats AI as a powerful but flawed collaborator.
You bring creativity, judgment, emotional authenticity, and your unique voice. For more on AI as assistant, not author, hear Richard on Tequila and Tech Talk. AI brings speed, breadth, tireless consistency, and pattern recognition. Neither of you is sufficient alone. You’re slow, limited in knowledge, and easily tired. AI is generic, emotionally hollow, prone to errors, and incapable of strategic judgment. Together, you compensate for each other’s weaknesses.
This only works if you maintain clear boundaries about who does what. AI brainstorms. You select. AI researches. You verify. AI drafts. You rewrite in your own voice. AI analyzes. You decide what to do with the analysis. The moment you start accepting AI output without filtering it through your own judgment, the work stops being yours and starts being generic.
I use AI extensively in my ghostwriting practice and in developing my writing handbooks. I use it for brainstorming structures, researching unfamiliar industries, analyzing draft chapters for pacing problems, and generating variations when I’m stuck. I don’t use it to write prose that goes to clients. The prose is mine. The voice is mine or my client’s. AI supports the process. It doesn’t replace the writer.
How to Start
If you’re new to AI-assisted writing, start small. Pick one brainstorming session where you’re stuck on a plot point or a chapter structure. Use AI for that single session. See what happens. Evaluate what was useful and what wasn’t.
The following week, try a different type of task. If you brainstormed first, try revision analysis next. Feed AI a chapter and ask specific questions about it. Notice what works differently. Brainstorming needs open-ended prompts. Revision needs specific questions about specific text.
Build gradually. As you learn what works for your process, integrate more. Add new types of tasks. Develop your prompt library. Find your rhythm. There’s no rush. Better to use AI well for a few tasks than poorly for many.
When things don’t work, and they won’t always work, that’s information. Adjust your approach. Try different prompts. Provide more context. AI-assisted writing is a skill. Like any skill, it develops through practice, failure, and adjustment. The writers who use AI effectively aren’t those who got it right the first time. They’re those who kept refining their approach until it worked.
Why Does AI Have a Voice Problem?
This deserves its own section because it’s the single most important issue for any writer using AI.
AI-generated text has tells. Read any AI output aloud. Does it sound like you? Or does it sound like a competent stranger wrote it? Your ear catches what your eye misses. Awkward rhythms, unnatural word choices, and passages that don’t match your voice all become obvious when you hear them.
If you find yourself performing the text instead of speaking it naturally, something is wrong. Search your document for AI-typical phrases: “it’s important to note,” “in the realm of,” “when it comes to,” “individuals” when you mean “people,” “use” when you mean “use,” “delve” when you mean anything at all. Replace them with how you’d actually say it.
AI tends toward consistent patterns. Similar sentence lengths. Similar structures. Similar rhythms. Human writing has rough edges, unexpected turns, short punchy sentences followed by longer flowing ones. If your prose feels too smooth, too balanced, too polished in a way that lacks personality, AI voice has crept in. Fix it.
My AI Shortcomings guide includes a complete list of patterns to watch for. My Using AI for Writing guide covers the positive side of collaboration. Both are free. Together they prepare you to work with AI effectively instead of becoming dependent on it. The full AI-Enhanced Writer’s Library has 46 handbooks applying these principles to specific craft areas. For one-on-one guidance, book coaching is available. Start with a conversation.