Publish a Book on Amazon KDP With AI: The 9-Step Plan, With Real Numbers (2026)

A version of this plan is doing the rounds on social media: find an evergreen problem, write a better book with AI, upload it to Amazon KDP, run $5 a day in ads, and you're at $2,000–4,000 a month by week ten. The method is mostly good. The numbers are the best case. Here's the full plan, step by step, with what each step really involves and what it really pays.

The rule: you're not selling a book, you're selling a better answer to a problem people already pay to solve. AI makes the writing faster. It doesn't make a weak book sell.

Step 1: Find a problem that never goes away (Google Trends)

Open Google Trends, set the time range to 2004–present and search problems people have had for decades: anxiety, weight loss, dog training, retirement planning, sleep, budgeting. You want a line that's flat or slowly rising over twenty years, not a spike. A spike is a fad. A flat or rising line is a permanent market.

• Switch to "Past 5 years" to check seasonality. Weight loss and budgeting peak every January. That's useful: a book that's live and reviewed by December catches the wave.
• Add a reference term so you can see scale. Trends shows relative interest, not search volume. A rising line at a tiny volume is still tiny.
• Go one level narrower. "Dog training" is huge and crowded. "Training a rescue dog with anxiety" or "retirement planning for self-employed people" are big enough to sell and small enough to stand out.

Output: three to five specific angles on an evergreen problem.

Step 2: Check that people already pay (Amazon)

Interest isn't demand. Money is. Search each angle on Amazon in the Books store and open the top results.

• Look at the Best Sellers Rank on each product page (under Product details). The lower the number, the more it sells. If the top five books on your angle all have ranks in the low tens of thousands or better in the Books store, people are buying every day.
• Look for self-published books in the top results. If the top ten are all from big publishers with famous authors, it's hard to break in. If a few are indie books with plain covers and a few hundred reviews, there's room.
• Note prices, page counts and formats. That tells you what buyers expect to pay and how long the book should be.

Too little competition is a warning, not a gift. No books usually means no buyers.

Step 3: Read the 3-star reviews

This is the best step in the plan, and it's free. Five-star reviews say "great book." One-star reviews often complain about shipping or a damaged copy. Three-star reviews are from people who bought, read, and wanted more. They tell you exactly what's missing.

Read thirty to fifty of them across the top five books and write down every pattern:

• "Too much theory, not enough steps."
• "Nothing for people who work night shifts."
• "Half of it is the author's life story."
• "Outdated: written before the rule change."

Paste the reviews into a document as you go. That list is your book's brief, and it's the copy for your book description too: "A practical guide, no life story, with a 30-day plan for people who work shifts."

Step 4: Turn the gaps into a better outline (ChatGPT or Claude)

Give your AI assistant the gap list, the table of contents of the two best-selling competitors (the Look inside preview shows it) and your reader. Ask for an outline that fixes every gap, with a clear promise per chapter.

You don't need to be the world's leading expert. You do need to know more than your reader, or bring someone who does. That matters most in health, money and psychology. An AI-written anxiety or weight-loss book with no real experience behind it is how you end up with one-star reviews that sink the listing, or worse, give bad advice. For those topics, write from your own experience, interview a professional, or pick a less sensitive angle.

A workable process:

1. Outline with AI, then rewrite it until every chapter answers a gap from step 3.
2. Draft chapter by chapter. Use AI for structure, examples and first drafts; add your own stories, checklists and opinions. That's the part nobody else can copy.
3. Aim for 25,000–40,000 words for practical nonfiction, unless the competitors are clearly longer or shorter.

Realistic time: four to eight weeks of evenings for a first book you're proud of. Two weeks is possible, but it shows.

Step 5: Edit like it matters (Grammarly, and a human)

Run the draft through Grammarly (or the free editing in Word or Google Docs) for spelling, grammar and clarity. Then read it out loud. That's where clumsy AI phrasing, repetition and filler jump out.

The viral version of this plan says to raise your "human score" in Originality.ai. Skip that game. Amazon doesn't rank books by AI-detector scores; readers rank them with reviews. A book that's generic gets 3 stars whatever a detector says. Spend that time on what readers notice: concrete examples, a clear structure, no padding.

Disclose honestly. When you set up the book, KDP asks whether the content is AI-generated. If AI wrote the text or made the cover image (even if you edited it afterwards), answer yes. If you wrote it and only used AI to brainstorm, edit or check, that counts as AI-assisted and doesn't need disclosing. The answer isn't shown on your book page. Amazon also caps how many new titles an account can create, so mass-uploading isn't an option anyway.

If you can, get two or three beta readers from your target audience to read it before you publish. One honest reader beats any tool.

Step 6: Make the cover (Ideogram, Canva or KDP Cover Creator)

The cover is your ad. On a phone it's a thumbnail the size of a stamp, so it needs one bold image and a title you can read at that size.

• Ideogram is good at images with readable text in them. Type your title, niche and style ("bold minimalist self-help book cover, large title text"), generate several options, and pick the strongest. Note: on the free plan your images are public and the credits are limited; check the current terms before using an image commercially.
• Look at the top ten covers in your niche first. Your cover should look like it belongs in that category, only cleaner.
• A paperback needs a full wrap cover: back, spine and front, at exact dimensions that depend on your page count and paper. Get the template from KDP's cover calculator and finish the layout in Canva, or use the free KDP Cover Creator in the upload flow. Most AI images also need upscaling to reach print quality (300 DPI).

A designer used to cost a few hundred dollars and two weeks. Now a decent cover costs an afternoon. A great one still costs a designer, and in a crowded niche that can be the best money you spend.

Step 7: Upload to KDP and price it right

Create a free account on kdp.amazon.com, complete the tax interview (outside the US you fill in a W-8BEN; Dutch residents can claim the treaty rate of 0% US withholding on royalties) and add your bank details.

Then set up the formats. One manuscript, three versions:

Format	Typical price	Royalty	You earn per sale (example)
Paperback, 180 pages	$12.99–16.99	60% minus printing (50% under $9.99)	$14.99 → $8.99 − $3.16 = $5.83
Kindle ebook	$2.99–9.99	70% minus a small delivery fee	$4.99 → about $3.40
Hardcover	$19.99–29.99	60% minus (higher) printing	Less per sale than you'd think; adds a premium option

Amazon.com rates and printing cost (black ink: $1.00 per book plus $0.012 per page for 110–828 pages), checked October 2026. Other marketplaces use their own currencies and print costs.

So "you keep 60–70%" is half true. You keep 60% of the paperback price after printing is deducted, which lands at around 40% of the cover price. Still very good for a product you never touch.

Amazon does print, ship, handle payments and customer service. Review takes up to 72 hours. Order a proof copy before you go live: typos and layout errors look much worse on paper.

Two things the hype skips:

• Royalties are paid about 60 days after the end of the month in which the sale happened. A sale in November arrives in late January.
• Enrolling the ebook in KDP Select (Kindle exclusive for 90 days) adds Kindle Unlimited page reads, which for many new authors are a large share of income. It means you can't sell the ebook elsewhere.

Step 8: One upload, many countries (but mostly one market)

Choose worldwide rights and your paperback goes on sale in 14 Amazon marketplaces: the US, UK, Canada, Australia, Japan, Germany, France, Spain, Italy, the Netherlands, Poland, Sweden, Belgium and Ireland. The ebook reaches a similar set. You set the prices once and Amazon converts them, or you set each one yourself.

Be realistic about where sales come from. An English-language book sells mainly in the US, followed by the UK, Canada and Australia. Germans and Dutch readers buy some English nonfiction, but rarely enough to plan on. "Selling in 14 countries" mostly means "selling in the US, with a long tail."

Step 9: Run $5 a day in Amazon Ads

In KDP, click Promote and Advertise and start a Sponsored Products campaign. Two campaigns make a good start:

1. Automatic targeting, $3/day: Amazon picks the searches and books. You learn what works.
2. Manual keyword targeting, $2/day: 20–40 phrases your buyers type ("anxiety workbook for adults", "retirement planning self employed"), plus the titles of competing books.

Now the honest maths. In popular nonfiction niches, a click often costs $0.50–1.50. If one in ten to fifteen clicks buys, one sale can cost $5–20 in ads against $5.83 in royalty. Many books lose money on ads for the first weeks. That's normal: you're buying data.

Give it three to four weeks, then cut. Pause keywords that spend without selling, raise bids on those that convert, add new ones from the search term report. Judge by ACOS (ad spend as a percentage of sales). For a $14.99 paperback, you break even at roughly 39% ACOS ($5.83 royalty out of $14.99 sales).

Ads aren't your only lever. Ads plus a good cover plus early reviews make Amazon's own recommendation engine ("Customers also bought") start working. That's the real goal: sales you didn't pay for.

The timeline: what the hype says vs. what to plan on

	The viral plan	A realistic plan
First book live	Week 1–2	Week 6–10 (research, writing, editing, proof copy)
First income	$400–600/month by week 6	A few to a few dozen sales a month in month one; paid ~60 days later
Book two	Week 7–8	Month 3–5, same niche
3–4 books	Week 10: $2,000–4,000/month	Month 9–18: $100–1,000+/month, highly variable

Here's why the viral numbers are hard to hit. $500 a month from paperbacks means about 86 sales a month, or three a day, every day, after ad costs. $3,000 a month means about 515 sales a month. Some people reach that. They usually have a niche they know well, five or more books that earn strong reviews, and a year or two of iterating behind them. Plenty of first books sell fewer than ten copies a month. That's not failure; it's the starting data.

The portfolio effect is real (it's just slower)

The best parts of the original plan hold up:

• Stay in one niche. A second book on a related problem ("anxiety at work" after "anxiety and sleep") gets recommended alongside the first. Each book becomes an ad for the others. Link them in the back matter: "If this helped, the next book covers..."
• Stack formats. Paperback, ebook and hardcover from one manuscript cost an evening to set up. A large-print edition helps in some niches, such as retirement and puzzles.
• Audiobooks, with a catch for European authors. ACX (Audible's platform) only accepts rights holders who live in and have a tax ID and bank account in the US, UK, Canada or Ireland. In a royalty-share deal, the narrator works for half of your royalties instead of a fee, and experienced narrators rarely accept that for an unknown book. Living in the Netherlands or elsewhere in the EU? Use another audiobook distributor, or Amazon's AI narration Virtual Voice, which is invite-only and sells only in the US store for now. Audible is also moving to a new pooled royalty model during 2026, so check the current rates.
• Build a list. Put a free bonus (a checklist or workbook) in the book with a link to your email signup. Amazon owns the customer; your list doesn't. Book three launches to people who already liked book one. Our guide to lead magnets worth an email shows how.

Is it worth doing?

Yes, if you treat it as building a small catalogue of useful books, not a ten-week money machine. It's cheap to start (a few dollars in tools and a proof copy, plus whatever you spend on ads), you can stop anytime, and a good book keeps selling for years with no stock, no shipping and no customer service.

It's not worth it if you plan to upload thin AI-written books as fast as possible. Amazon's limits, AI disclosure, reviews and the ad costs all work against that.

Starting this week? Run steps 1 to 3 in one evening. If you can't find a niche where the 3-star reviews hand you a clear gap, keep looking before you write a word. More ways to sell online, from finding a product worth testing to building the store, are on the Sell Online shelf.