The State of AI Self-Publishing 2026

Most of what gets written about “AI self-publishing” is either hype (“write a bestseller in an afternoon”) or panic (“the Kindle store is drowning in slop”). Both narratives assume the same thing: that people are using AI to mass-produce novels. Our data says that assumption is almost entirely wrong.

KDP Builder is an AI self-publishing studio built specifically for Amazon Kindle Direct Publishing. In our platform's first quarter, 974 creators started 775 books. We looked at what they actually chose to build — the formats, the trim sizes, the genres, the print-vs-digital decisions — and the picture that emerges is very different from the headlines.

This report is our attempt to replace opinion with evidence. Every proprietary figure comes from our own platform. Every external figure is cited so you can check it yourself. Where our sample is thin, we say so.

— The KDP Builder team

1. 01.AI self-publishing is a low-content and illustrated-book phenomenon, not a novel factory. Puzzle and coloring books alone are 49% of all books created on the platform. Add children's, planners, journals and other low-content formats and 70% of AI-built books are illustrated or activity titles. Fiction is just 8.7%.
2. 02.Print is the default, not the exception. 96% of projects target a print edition; only 3.7% are ebook-only. The “Kindle-first” mental model is outdated — AI self-publishers are print-on-demand publishers first.
3. 03.The most common book size is 8.5×11 — an activity-book format — not the 6×9 novel trim. 8.5×11 is 49% of all projects that specify a size; the classic 6×9 novel trim is a distant second at 23%.
4. 04.When people do write fiction with AI, they write genre fiction. Commercial genre fiction outnumbers literary fiction roughly 9:1, led by mystery and romance (tied for first).
5. 05.Adoption is compounding fast. Monthly book creation grew nearly 7× in the first quarter (33 → 229 → 367), even before a full month of July was counted.
6. 06.The industry backdrop makes this matter. The U.S. produced an estimated 3.5 million self-published titles in 2025 (up ~39% YoY), and roughly 45% of authors now use generative AI in some part of their workflow.* Our data shows where that AI energy is actually going.

* External figures are third-party estimates — see “Industry context” and References.

What the dataset is. This report is built on first-party data from the KDP Builder platform, covering May–July 2026 (the platform's first full quarter of operation):

• 629 structured, multi-phase book projects
• 146 additional quick-generate books
• 775 books in total
• created by 974 registered creators
• plus 138 “bring-your-own-manuscript” formatter projects and 125 standalone cover designs, used as supporting signals

All figures are aggregated and anonymized. No titles, personal data, or identifiable creator information is included or published.

What the dataset is not. We are deliberate about the limits:

• Single-platform sample. These are KDP Builder users. Because our tooling is unusually strong for puzzle, coloring and low-content books, our audience may skew toward those formats. We flag this wherever it could bias a finding.
• A young platform, measured over one quarter. Growth figures reflect platform adoption, not the whole industry's growth rate.
• “Created” is not “sold.” We measure what people choose to build and prepare for upload — not final Amazon sales or earnings.
• Optional fields. Genre and trim size are optional in some flows, so percentages for those are reported against the subset where the field was specified (and we state the n).

External figures (market size, industry-wide AI adoption, etc.) come from third-party sources, are cited inline and in the References section, and should be independently verified before republication.

The single most important thing our data shows: the people using AI to publish are overwhelmingly making visual, activity, and low-content books — not novels.

Across the 629 structured projects:

The headline numbers:

• Puzzle + coloring = 40% of structured projects (253 of 629). Including the 146 quick-generate books, puzzle and coloring titles are 49% of all 775 books — nearly half of everything created.
• Activity & low-content formats (puzzle, coloring, low-content, planner, journal) = 54%.
• Add children's picture/activity books and illustrated-or-activity titles reach 70%.
• Text-forward books (nonfiction, fiction, workbook, cookbook, business, memoir) are the minority at 30%.
• Fiction is just 8.7%. The novel-writing robot is a media fixation, not what the data shows.

Why it matters. The public conversation about “AI slop flooding Kindle” imagines endless AI novels. In reality, AI is being used most where it removes production drudgery — generating 100 unique mazes, a themed coloring set, or a professionally laid-out planner interior — not where it tries to replace authorial voice. This aligns with the broader low-content market's 2025 shift away from mass-duplicated junk and toward niche, well-produced activity books.

If you assume AI self-publishers are pumping out 99-cent Kindle ebooks, the data will surprise you.

• 96% of projects include a print edition.
• Only 3.7% are ebook-only.

Why it matters. Print-on-demand — not Kindle ebooks — is the center of gravity for AI self-publishing. It makes sense once you accept Finding 1: you can't sell a coloring book, a puzzle book, or a children's picture book as a reflowable ebook. The dominant formats are inherently physical. Any tool, marketplace, or piece of advice built on an “ebook-first” assumption is aiming at the wrong 4%.

Trim-size choices confirm the low-content/illustrated skew. Among the 422 projects that specified a size:

• The large-format 8.5×11 — the size you'd pick for a puzzle, activity, or workbook — is the single most common trim at 49%.
• The classic 6×9 novel trim is second at 23%.
• Add the children's square (8.5×8.5) and large/illustrated formats dominate the physical shape of AI books, just as they dominate the content types.

Fiction is a minority of activity (8.7%), but it's revealing. Among the 41 fiction projects with a genre specified:

• Commercial genre fiction outnumbers literary fiction roughly 9:1.
• Mystery and romance are tied at the top — the two most commercially reliable indie categories.

(Small sample: n = 41. Directional, not definitive — but it points the same way as the rest of the data: AI self-publishers are commercially minded, targeting proven-demand categories rather than prestige.)

Why it matters. This mirrors what actually sells for indies. AI-assisted authors aren't chasing literary awards; they're writing to market in mystery, romance, and thriller — the categories with the deepest, most habitual readerships.

• Nearly 7× growth from month one to month three — and July was still incomplete when measured.
• This is platform-adoption growth (a new product), so read it as a signal of demand intensity rather than an industry-wide growth rate.

The demand pattern is consistent with the external picture: a self-publishing market growing double digits annually, and AI adoption among authors crossing the ~45% mark. The energy is real; our data shows it flowing into production-heavy, print-first book formats.

Our platform data sits inside a larger, well-documented boom. These figures are from external sources and are provided for context only:

• Market size: the global self-publishing market was ~$1.85B in 2024, projected toward $6.16B by 2033 at a ~16.7% CAGR. (automateed.com; Grand View Research)
• Volume: the U.S. produced an estimated 3.5 million self-published titles in 2025, up ~38.7% YoY; self-published titles now outnumber traditionally published ones by more than 5:1. (selfpublishing.com)
• Platform dominance: Amazon KDP holds an estimated 65–70% of the self-publishing market. (automateed.com)
• AI adoption: a May 2025 survey of 1,200+ authors found ~45% use generative AI, most often for research (81%), marketing copy (73%), outlining (72%) and editing (70%). (authors.ai; manuscriptreport.com)
• Disclosure gap: ~74% of AI-using authors don't disclose it, while ~67% of readers say they'd be less likely to buy a book known to be AI-written. (manuscriptreport.com; wifitalents.com)
• Low-content shift: the KDP low-content market has moved from “volume-at-all-costs” to niche-specific, quality-first production. (merchtitans.com; vappingo.com)

How to read this alongside our data: the industry is growing and AI-curious, readers are wary of undisclosed AI prose, and the low-content market rewards quality and niche. Our platform data shows AI self-publishers responding rationally — concentrating on visual/low-content, print-first formats where AI adds production value and the “is this AI-written?” trust problem barely applies.

1. The “AI novel flood” narrative is misdirected. The real story is an activity- and low-content-book boom. Expect Amazon's quality enforcement and buyer expectations to keep pushing this segment toward betterproduction, not less of it.
2. Print-on-demand is the strategic center. Tools, services and advice should assume a print-firstcreator making a large-format book — not an ebook-only author.
3. Niche + quality beats volume. The mass-duplicate low-content playbook is dead; the winners pair AI production speed with genuine differentiation and professional design.
4. Genre discipline wins in fiction. AI lowers the cost of writing to market — the authors using it are (sensibly) aiming at mystery, romance and thriller readerships.
5. Disclosure and trust will be the fault line for AI prose — which is one more reason the current center of gravity is visual and activity books, where the trust penalty is minimal.

KDP Builder is the complete AI self-publishing studio for Amazon KDP. It helps independent authors write manuscripts, generate genre-aware covers and print-ready interiors, and export a full upload package — covers, interior PDF, description, keywords, categories, author bio, and a KDP checklist — in one place. Start free with 75 credits, no card required.

A. Sample composition (May–Jul 2026)

• B. Book-type distribution — see Finding 1 table (n = 629 structured projects).
• C. Output format — see Finding 2 table (n = 629).
• D. Trim size — see Finding 3 table (n = 422 specifying a size).
• E. Fiction genre — see Finding 4 table (n = 41 specifying a genre).
• F. Monthly creation — see Finding 5 table.

All proprietary data aggregated and anonymized from the KDP Builder platform. Percentages may not sum to 100 due to rounding.

1. selfpublishing.com — Publishing by the Numbers (2025 U.S. title volume, self-vs-traditional ratio)
2. automateed.com — Self-Publishing Statistics (market size, CAGR, KDP market share)
3. Grand View Research — U.S. Self-Publishing Market Report
4. authors.ai — How Authors Are Really Using AI (2025 author survey)
5. manuscriptreport.com — AI Publishing Statistics (adoption, disclosure)
6. wifitalents.com — AI in the Publishing Industry Statistics (reader sentiment)
7. merchtitans.com — Low-Content Books on Amazon KDP (market shift to niche/quality)
8. vappingo.com — Low-Content Books Income (income realism, niche strategy)

External statistics are third-party estimates reported by the sources above and have not been independently audited by KDP Builder. Verify before republication.