What creators started building on KDPBuilderA first-party platform snapshot, May–July 2026

1. Puzzle and coloring books were the most common project types — 42.4% combined (205 of 484), effectively tied at 21% each. Illustrated or activity titles (adding children's books) reached 59.1% (286 of 484).
2. Prose books were a minority of starts. Fiction was 8.1% (39), nonfiction 6.6% (32); fiction, nonfiction, business and memoir together were 16.3% (79 of 484).
3. Print was the default output. 95.2% of projects (461 of 484) included a print edition; 4.8% (23) were ebook-only.
4. Large-format trim dominated where a size was set. Of 293 projects with a trim size, 52.9% chose 8.5×11 and 20.1% chose 6×9. 191 projects (39.5% of 484) had no size set.
5. Starts accelerated within the window. Projects per calendar day rose from 4.8 in June to 13.8 in 1–23 July (≈2.9×). This measures activity on one growing platform, not adoption of AI publishing.
6. Not in this data: whether the books were finished, published on any store, or sold. 38 of the 484 projects have a recorded successful file download, but download logging only began on 21 June 2026, so that figure is not a completion rate.

All proprietary figures were reproduced on 9 September 2026 with read-only queries against the production database, using a fixed cutoff so the tables can be re-run and audited. The first-edition (24 July 2026) numbers could not be reproduced with a single consistent rule set; the differences are listed in the change log below.

Terms, kept separate

Related activity outside the 484, for completeness: 20 quick-generate sample books, 138 bring-your-own-manuscript formatter projects (62 exported), and 215 standalone cover-design jobs (211 completed) were created in the same window. These are different products and are not mixed into the project tables.

• Puzzle + coloring: 205 of 484 = 42.4%.
• Illustrated or activity (puzzle, coloring, children's): 286 of 484 = 59.1%.
• Low-content-style (puzzle, coloring, low-content, workbook, planner): 307 of 484 = 63.4%.
• Prose (fiction, nonfiction, business, memoir): 79 of 484 = 16.3%.

Interpretation is limited to this platform: KDPBuilder markets itself heavily to puzzle, coloring and children's-book creators, so the mix reflects who chose the tool as much as what people are publishing with AI in general.

461 of 484 (95.2%) included a print edition. Given Table 1 this is unsurprising: coloring, puzzle and activity books are print products.

8.5×11 and 6×9 together account for 214 of 293 (73.0%). Because 39.5% of projects had not set a size, shares here should not be read as shares of all projects.

We previously claimed a “roughly 9:1” ratio of commercial to literary fiction. With 27 usable records and most genres below five projects, that claim is withdrawn.

June to July, projects per day rose about 2.9× (4.8 → 13.8). We no longer report a May-to-July multiple: nine days of May is not a month. This series measures starts on one platform during a period when the platform itself was growing and marketing; it says nothing about AI self-publishing adoption in general.

• Self-selected sample. Users chose a tool that specialises in visual and low-content books. Shares describe KDPBuilder users, not authors, AI users, or Amazon KDP publishers.
• Starts, not outcomes. A project is counted when it is created. We do not know how many became finished files, store listings or sales.
• Incomplete download log. Successful download tracking began 21 Jun 2026; the 38-project figure is a floor for the later part of the window only.
• Missing fields. 39.5% of projects had no trim size and 12 of 39 fiction projects had no genre; those tables use smaller denominators.
• Short window. 62 days on a new product. Month-to-month change partly reflects marketing and product launches.
• No independent review. Queries and tables were prepared and checked by the KDP Builder team only.

The first edition was published under the title “The State of AI Self-Publishing 2026”. On re-examination its counts mixed denominators and its conclusions were broader than the sample. The corrections below were applied on 9 Sep 2026; the URL is unchanged so existing links still resolve.

One CSV containing every count and share on this page (Tables 1–5 plus derived groupings), with the denominator and inclusion note on each row. It contains no personal data and no per-creator rows.

Licence: CC BY 4.0 — reuse with attribution to KDP Builder and a link to this page. The read-only query definitions (window, exclusions, small-cell rule) are summarised under Methods; the query text is available on request.

• Market-size and growth estimates for self-publishing. (automateed.com; Grand View Research — estimates differ by source.)
• U.S. self-published title-volume estimates. (selfpublishing.com — methodology not reviewed by us.)
• Author surveys on generative-AI use. (authors.ai, 2025 — self-reported survey; sampling not reviewed by us.)

KDP Builder is an AI-assisted publishing studio for independent authors. It helps creators write manuscripts, generate covers and print-ready interiors, and export an upload package for Amazon KDP — covers, interior PDF, description, keywords, categories and a pre-upload checklist. KDP Builder is not affiliated with or endorsed by Amazon.

1. selfpublishing.com — Publishing by the Numbers (U.S. title volume estimates)
2. automateed.com — Self-Publishing Statistics (market size, KDP market share estimates)
3. Grand View Research — U.S. Self-Publishing Market Report (market-size estimate)
4. authors.ai — How Authors Are Really Using AI (2025 author survey)

Listed for transparency only. These sources were not audited by KDP Builder and none of their figures are used in the tables above.