Is AI Writing Worth It? My Honest Take After Writing 20 Books with AI

Published April 5, 2026 · 15 min read

Twenty books. Eighteen months. Roughly $27,000 in total royalties. One existential crisis about whether I'm really a "writer" anymore (resolved, mostly). And a very clear answer to the question everyone keeps asking me.

Is AI writing worth it?

Yeah. But not for the reasons you think, and not without some real trade-offs that nobody in the AI hype machine wants to talk about.

I'm going to be genuinely honest here because the internet has enough people who either think AI writing is the greatest thing ever or think it's destroying literature. The truth is messier, more interesting, and more useful than either extreme. I've got 20 books of data to back it up.

The Numbers: 20 Books in 18 Months

Let me lay out the raw numbers before I get into opinions. Here's what 18 months of AI-assisted book publishing looks like:

• Total books published: 20
• Genres: Thriller (6), Romance (5), Sci-fi (4), Fantasy (3), Self-help (2)
• Average word count: 42,000 words
• Total words published: ~840,000
• Average time per book (idea to published): 3.5 days
• Total production cost: ~$3,200 (tools, editing, covers)
• Total revenue: ~$27,000
• Average Amazon rating: 4.1 stars
• Best-performing book: $4,800 lifetime / $320 per month
• Worst-performing book: $180 lifetime / $12 per month

If you'd told me two years ago that I'd publish 20 books in a year and a half and make $27,000 from them, I'd have laughed. Not because the money is life-changing (it's not), but because I spent the previous five years "working on my novel" and had exactly zero published books and zero dollars to show for it.

That context matters. My baseline wasn't "successful author considering AI." My baseline was "person who wanted to be an author and kept not finishing anything." AI changed that equation completely.

What Worked Better Than I Expected

Speed of First Drafts

This is obvious, but I want to be specific about how dramatic the change is. My pre-AI writing pace was about 1,000 words on a good day, maybe 3,000 on a great day. At that rate, a 42,000-word novel would take me 14-42 days of writing. And that's just the first draft.

With ShakespeareAI, I get a complete 42,000-word first draft in about 12 minutes. Twelve minutes. Then I spend 6-10 hours editing it over the next couple of days. Total active work time per book: about 15 hours. Compare that to the 80-120 hours a traditionally-written novel takes at my pace.

That's not a small improvement. That's an 80% reduction in time per book. It's the difference between publishing 2-3 books a year and publishing 12-15.

Genre Fiction Quality

I'll get into the quality nuances later, but the short version: for genre fiction — thrillers, romance, sci-fi — the AI output after editing is genuinely good. Not "good for AI." Just... good. My thriller readers don't know or care that AI was involved. They care that the plot keeps them turning pages and the ending satisfies. AI handles both of those things well.

I was skeptical at first. I assumed AI fiction would feel hollow or formulaic. Some of it does, in the raw draft. But after editing — injecting my voice, sharpening the dialogue, fixing pacing — the final product reads like a solid indie thriller. I covered the full editing process in our guide to improving AI-generated writing, and those techniques genuinely transform the output.

Consistency of Output

This one caught me off guard. When I was writing manually, my output quality varied wildly. Some days I wrote well. Some days I wrote garbage. My mood, energy level, whether I'd slept well — all of it affected the writing.

AI doesn't have bad days. The draft quality is consistent every time. I still have to edit, but I'm always starting from the same baseline of "competent first draft." That predictability is incredibly valuable when you're trying to build a publishing business.

Idea Validation Speed

Before AI, testing a book idea meant committing months to it. Write the draft, see if it works, realize the concept was flawed, start over. Painful.

Now I can test a concept in a day. Generate the book, read through it, see if the story actually works in practice. If it doesn't, I've lost a day, not a month. Three of my 20 books started as completely different concepts that I pivoted after seeing the first AI draft. That ability to iterate fast is underrated.

What Didn't Work (Or Flat-Out Failed)

Here's where most AI writing articles stop. They tell you the good stuff and skip the failures. I'm not going to do that because the failures taught me more than the wins.

My Self-Help Books Underperformed

I published two self-help books with AI. Both underperformed significantly. Average rating: 3.4 stars. Combined revenue: about $900 over 18 months.

Why? Because self-help readers want genuine expertise and personal experience. AI can structure a self-help book well — the advice is logical, the chapters flow — but it can't share real stories of struggle and growth. The two self-help books felt like they were written by someone who'd read about the topics but never lived them. Because... that's exactly what happened.

I could have fixed this by injecting way more personal content during editing, but that would have required essentially rewriting 40% of the book. At that point, the AI wasn't saving me enough time to justify the approach.

Lesson: AI works best for fiction and topic-based nonfiction. For personal memoir, self-help, or anything that requires lived experience as the core value proposition, AI is a poor fit unless you're willing to do very heavy editing.

Book 4 Was a Disaster

My fourth book — a sci-fi novel — was the worst thing I've published. I was overconfident after three decent books and rushed the editing. I spent maybe 2 hours on it instead of my usual 6-10. Published it, felt great about my "efficiency."

Then the reviews came in. "Repetitive." "Characters felt flat." "The ending made no sense." One reviewer wrote "this reads like it was written by an algorithm" and... fair. It was. Because I hadn't done my job as the editor.

That book still sits at 2.8 stars. I thought about pulling it but left it up as a reminder. The AI didn't fail — I failed. I treated the AI output as a finished product instead of a first draft, and readers noticed immediately.

If you're wondering whether AI can write an entire book without any human editing and have it be good — the answer is no. Not yet. Maybe not ever, depending on your quality bar.

Fantasy Was Harder Than Expected

My three fantasy novels required significantly more editing than my thrillers or romances. World-building consistency is a problem — the AI would establish that magic worked one way in chapter 3 and then subtly change the rules by chapter 12. Names would drift. Geography would contradict itself.

I spent 12-15 hours editing each fantasy book compared to 6-8 for thrillers. They turned out well (4.2 average stars), but the time investment was much higher. Fantasy writers considering AI should budget for more editing time. I went deep on this in our tools comparison for 2026.

The Time Savings Are Real, But...

Everyone talks about how AI saves time. And it does. Enormously. But the time calculation is more nuanced than "AI writes in 10 minutes vs months of manual writing."

Here's the real time breakdown for one of my books:

The draft generation is where AI saves massive time. But notice that editing still takes 6-10 hours. Planning still takes 2-3 hours. You can't skip those and get good results. My Book 4 disaster proved that.

The honest framing is: AI turns a 2-3 month project into a 3-5 day project. That's incredible. But it's not "push a button and get a book." It's "push a button and get a first draft that needs real work."

I documented one of these projects start to finish — you can read about it in our 3-day novel experiment.

Quality: An Honest Assessment

This is the section where I'm going to lose some people who want me to say AI writing is amazing and some people who want me to say it's garbage. It's neither.

What AI Does Well

• Plot structure — AI generates well-paced plots with proper story arcs. Beginning, middle, end, rising tension, climax, resolution. The bones are solid
• Dialogue — Surprisingly natural in most cases. Characters sound distinct. Not award-winning dialogue, but good enough that readers don't notice
• Genre conventions — AI nails genre expectations. Thriller readers get twists. Romance readers get tension and payoff. Sci-fi readers get interesting concepts
• Pacing (macro level) — Chapters end on hooks, the story moves forward, there's no sagging middle (usually)

What AI Does Poorly

• Emotional depth — AI can write "she felt sad." It struggles to write the specific, gut-punch kind of sadness that makes you cry. The difference between "he missed her" and the particular ache of finding her coffee mug in the cabinet six months later
• Subtle thematic work — AI doesn't do subtext well. Everything is surface level. If your story's real meaning is supposed to emerge from pattern and implication rather than direct statement, AI will struggle
• Micro-pacing — Within scenes, AI maintains a steady pace. It doesn't know when to slow down for a single devastating sentence or speed up with staccato fragments during action. That's editing work
• Surprise — AI is rarely genuinely surprising. It generates competent, expected choices. The weird, risky, inspired choices that make a scene memorable — those come from humans

The Quality Tier

After editing, my AI-assisted books land in what I call the "solid genre fiction" tier. They're comparable to mid-list self-published books by competent indie authors. Not debut literary fiction. Not bestseller-level commercial fiction. But genuinely enjoyable reads that deliver on their genre promises.

That quality tier is where the vast majority of successful self-published books live anyway. Most KDP bestsellers aren't literary masterpieces — they're well-told genre stories that readers consume quickly and enjoy. AI hits that mark consistently.

What Readers Actually Think

This is the data I find most interesting. What do actual readers — people who paid money for these books — think?

Across 20 books, I have about 480 total reviews on Amazon. Here's the breakdown:

• 5-star reviews: 38%
• 4-star reviews: 29%
• 3-star reviews: 18%
• 2-star reviews: 10%
• 1-star reviews: 5%

Average: 4.1 stars. For indie genre fiction, that's solid. Not spectacular, but firmly in the "readers liked it" territory.

Out of 480 reviews, exactly 3 mentioned AI or anything suggesting the book felt "automated." Three out of 480. Less than 1%. The overwhelming majority of reviews talk about plot, characters, pacing — the stuff readers actually care about.

The negative reviews? They mention things like "predictable ending" (fair — AI endings can be safe), "wanted more depth from the main character" (fair — that's the emotional depth ceiling I mentioned), and "felt rushed in places" (also fair — that's usually a pacing issue I should have caught in editing).

Nobody said "this was clearly written by a robot." Because after editing, it doesn't read like a robot wrote it. It reads like a human wrote a fast genre novel. Which is essentially what happened.

The Money: What 20 AI Books Actually Earn

Let me be transparent about the financial side because this is what most people really want to know.

Some important context:

The revenue is not evenly distributed. My top 5 books generate about 60% of total revenue. My bottom 5 generate about 8%. Publishing is a hits business even with AI. Volume helps because more books means more chances to find a hit, but not every book will perform.

Revenue grows over time. My first month of publishing made me $47. Month 6 was about $600. Now I'm at $1,800/month. Each new book adds to the catalog, and older books continue to earn. The compounding effect is real but slow.

$1,800/month is not "quit your job" money. Let me be clear about that. I still have a full-time job. AI book publishing is a side income that I spend about 15-20 hours per month on (roughly one new book per month at this point). It's meaningful extra money, not a livelihood. Some AI book publishers make $5,000-$10,000/month, but they've been at it longer and have bigger catalogs.

For more on the financial side, our post on making money writing books with AI goes deeper into strategies and realistic expectations.

What Surprised Me Most

Editing Is Where the Magic Happens

I went into this thinking the AI generation was the important part. I was wrong. The editing is where a book transforms from "generic AI output" to "my book." The voice, the personality, the specific choices that make readers connect — all of that comes from the editing phase.

I've started thinking of myself as a "director" rather than a "writer." The AI generates the raw footage. I shape it into the final cut. That reframe made me better at the process and more satisfied with the results.

My Writing Got Better

This was the biggest surprise. After editing 20 AI-generated books, my own writing improved. Dramatically. I can identify weak prose instantly now. I know what makes dialogue sound natural. I understand pacing at a structural level because I've edited 840,000 words of it.

Editing AI text is, unexpectedly, one of the best writing courses I've ever taken. You see the patterns — good and bad — at a scale that would take decades of manual writing to accumulate.

Genre Matters More Than AI Quality

My best-performing books aren't my best-written books. They're the ones in popular genres with strong keywords and good covers. My Book 7 — a thriller about a missing person in a small town — earns $320/month. My Book 14 — a sci-fi novel that I think is much better written — earns $90/month. The market doesn't care about prose quality as much as genre fit and discoverability.

This was humbling. But it's also freeing. It means you don't need perfect AI output. You need the right concept in the right genre with the right packaging. The writing just needs to be "good enough," and AI clears that bar for genre fiction.

The Ethical Worry Faded

I spent the first few months feeling vaguely guilty about using AI. Was this "real" writing? Was I cheating? Was I taking opportunities from "real" writers?

That worry faded as I realized: I wasn't taking anything from anyone. The books I published with AI are books that would never have existed otherwise. I wasn't going to write 20 novels manually. I was going to write zero novels and keep talking about "my book" at dinner parties for another decade. AI didn't replace my writing — it created writing that wouldn't have existed.

And the editing work is genuinely creative. I make thousands of decisions per book about voice, tone, pacing, and character. The AI doesn't make those decisions. I do. The final product is a collaboration, and I'm at peace with that.

Who Should (and Shouldn't) Use AI to Write

AI Writing Is Worth It If You:

• Have ideas but struggle with execution. If the blank page is your enemy, AI removes it entirely
• Want to build a publishing catalog. Volume matters in self-publishing, and AI makes volume possible
• Write genre fiction. Thriller, romance, sci-fi, fantasy, mystery — AI handles these genres well
• Think of writing as a business. If you want to make money from books, AI's speed and cost advantages are massive
• Enjoy editing more than drafting. If you'd rather shape existing text than create from scratch, this is your workflow

AI Writing Probably Isn't Worth It If You:

• Write literary fiction aiming for awards. AI can't produce the kind of prose that wins Pulitzers. If that's your goal, write manually
• Love the process of putting words together. If drafting is the fun part for you, AI takes that away
• Write memoir or deeply personal nonfiction. AI can't share your life experiences. It can help structure them, but the content has to come from you
• Aren't willing to edit. If you want to push a button and publish, you'll get bad results and bad reviews. The editing is non-negotiable
• Feel ethically opposed to AI writing. That's a valid position. If using AI would make you feel bad about your work, don't use it. The guilt would undermine the process

Final Verdict

After 20 books, 18 months, $27,000 in revenue, and a lot of editing — is AI writing worth it?

For me, unequivocally yes.

Not because the AI writes great books on its own. It doesn't. Not because it's easy money. It's not. Not because it replaces the need for human creativity. It doesn't.

It's worth it because it turned me from someone who talked about writing into someone who actually writes. It's worth it because I've published 20 books that real people have read and enjoyed. It's worth it because I earn $1,800/month from a creative side project that takes 15-20 hours per month to maintain.

And honestly? It's worth it because I'm a better writer now than I was 18 months ago, specifically because of how much editing AI text has taught me about the craft.

The tools have gotten remarkably good. When I compare the output I'm getting now to what I got 18 months ago, the improvement is striking. Tools like ShakespeareAI keep getting better at consistency, prose quality, and genre awareness. The books I'll write next year will be better than the ones I wrote this year, and I'm already happy with this year's.

If you're on the fence, here's what I'd suggest: try it. Start with a free tool. Write one book. Edit it properly. Publish it. See what happens. That's a much better way to decide than reading opinions — including mine — on the internet.

The worst case scenario is you spend a weekend and have a published book to show for it. I can think of worse ways to spend a weekend.

For a step-by-step guide on getting from idea to published, our how to publish an AI-generated book walkthrough covers the full process.

Related Articles

• Can AI Write an Entire Book?
• I Wrote a Full Novel with AI in 3 Days
• How to Make Money Writing Books with AI
• 10 Ways to Improve AI-Generated Writing
• Best AI Writing Tools for Authors (2026)
• How to Publish an AI-Generated Book

FAQ: Is AI Writing Worth It?

Is AI writing good enough to publish?

Yes, with editing. Raw AI output is a solid first draft but not publish-ready. After 4-8 hours of editing per book, the quality is on par with mid-tier self-published books in genre fiction. My AI-assisted books average 4.1 stars on Amazon across 20 titles. Not award-winning, but readers enjoy them.

How much money can you make writing books with AI?

It varies wildly. My 20 books generate about $1,800/month combined. My best book makes $320/month. My worst makes $12. The key factor isn't AI vs human writing — it's genre selection, cover quality, and Amazon keyword optimization. AI just lets you publish enough books to find what works.

Do readers know when a book is written with AI?

Generally no, if you edit properly. Out of hundreds of reviews across my 20 books, only 3 mentioned anything about AI. Two were from people who liked the book and were curious. One was negative. Most readers care about whether the story is good, not how it was produced.

How long does it take to write a book with AI?

AI generation takes 10-15 minutes. Editing takes 4-10 hours depending on how thorough you are. Cover, formatting, and upload add another 2-3 hours. Total: I can go from idea to published in 2-3 days working part-time. Compare that to the 6-12 months most traditionally-written books take.

What are the biggest downsides of AI writing?

Three main ones: 1) The emotional depth ceiling — AI struggles with the kind of nuanced emotional writing that makes literary fiction great. 2) The editing requirement — you still need to invest real time making the output your own. 3) The creative satisfaction question — if you love the craft of putting words together, AI removes that part. For some people, that's a dealbreaker.

Is AI writing going to get better?

It already has, dramatically. The books I write now with AI are noticeably better than the ones I wrote 18 months ago — same tools, better models. Context windows are longer, prose quality is higher, consistency across long works has improved. The trajectory is clear: AI writing is getting better fast, and the tools that were "okay" a year ago are "genuinely good" now.