Self-publishing has a supply problem and an opportunity problem at the same time. More people can release a book, but a new listing competes for limited reader attention. A title-count chart alone cannot tell an author whether a particular book will reach anyone. An income survey cannot tell a first-time creator what they will earn. This guide gathers public evidence about U.S. title output, indie author income, and AI-assisted publishing, then turns each number into a decision a writer can actually make.
Scope: figures below were reviewed in September 2026. This is a synthesis of named outside sources, not an EarnDraft survey or a count of our users. The large U.S. output totals derive from Bowker data reported by Publishers Weekly; the author-income figures come from the organizations that surveyed authors. Every number has a population and observation period. We do not present another company’s private “books created” counter as a self-publishing industry total.
A quick evidence table
| Question | Best available public figure here | Source and period | Read it as |
|---|---|---|---|
| How many self-published U.S. ISBN titles? | More than 3.5 million in 2025 | Bowker, reported by Publishers Weekly, March 2026 | ISBN-linked title output in one country |
| How much did that output grow? | 38.7% over 2024 | Same source | Change in the registered self-published category |
| How many traditional U.S. ISBN titles? | 642,242 in 2025 | Same source | A registered-title comparison, not a sales comparison |
| What did all surveyed authors earn from books? | $2,000 median in 2022 | Authors Guild survey of 5,699 published authors, reported 2023 | Median among respondents, including part-time writers |
| What did full-time self-published respondents earn from books? | $12,800 median in 2022 | Same Authors Guild survey | A different, more committed respondent group |
| What did committed indie authors report more recently? | $13,500 median gross annual income | Alliance of Independent Authors 2025 survey | Gross survey income under ALLi’s screening rules |
| How many indie respondents earned at most $100 monthly? | 44% | Written Word Media 2025 survey | Share of that survey’s respondents, not every indie author |
The Bowker report as covered by Publishers Weekly, Authors Guild income results, ALLi’s 2025 data drop, and Written Word Media’s 2025 survey should be read alongside one another, not averaged together. They observe different groups and outcomes.
How many books were self-published in the United States?
Bowker compiled more than 3.5 million self-published ISBN-bearing print and ebook titles for 2025, a 38.7% increase from 2024, according to Publishers Weekly’s March 2026 report. The same report says traditionally published titles rose 6.6% to 642,242. A rough ratio is more than five self-published ISBN records for each traditionally published record. This is a striking difference in output, but it does not mean self-published books outsold traditionally published books five to one.
An ISBN is an identifier for a particular publication format or edition, not a universal counter of works or people. One underlying book may have several ISBNs. An author can also distribute an ebook without purchasing a separate ISBN in some channels. Bowker’s registered-title count therefore should not be described as every distinct book released, every author who published, or every book sold. Because books without ISBNs are absent, it is not a complete census of self-publishing activity either. The exact gap is not established by this count.
| Number in a headline | What the underlying record measures | Common mistake |
|---|---|---|
| Self-published ISBN titles | Registered publication records classified as self-published | Calling them unique authors |
| Traditional ISBN titles | Registered publication records classified as traditional | Treating them as copies sold |
| Growth from a prior year | Change in recorded output under the same classification | Assuming identical change in demand |
| Genre title counts | Records assigned subject categories | Treating categories as sales rankings |
| “Books per day” | Annual count divided by calendar days | Suggesting every day saw identical releases |
What should a new author do with the output number? Use it as a reminder that a listing must explain its value quickly. Define who needs the book and what it helps them do. Show representative pages. Make the title and description specific enough to be found by that reader. A claim that “there are millions of books” is not a reason to stop; it is a reason to avoid generic positioning.
Does self-published title growth prove reader demand?
No. Supply and demand must be measured separately. The output count tells us how many registered titles entered a catalog. It does not report unit sales, revenue, reading time, satisfaction, or repeat purchase. A better test for a planned book combines evidence of a reader problem with evidence that the author can solve it better than available alternatives. Search results, forum questions, course feedback, interviews, and existing sales pages are qualitative signals, not automatic revenue predictions.
The Association of American Publishers’ 2025 StatShot release offers a separate view of U.S. publishing revenue, reporting aggregate revenue for books and course materials. It cannot be placed directly under Bowker’s title count to calculate revenue per self-published title: the populations, channels, product types, and measurement methods differ. This is a classic denominator trap. Divide only when numerator and denominator refer to the same market, period, and items.
Try this comparison exercise. Suppose a local bookstore adds 1,000 new titles to its catalog. If annual sales stay flat, average sales per title may fall; if demand grows faster, it may rise. The title count alone leaves both scenarios possible. Now replace the bookstore with a national online retailer and the lesson remains. Publishing is an access mechanism; discoverability and reader trust must be earned separately.
How much do self-published authors earn?
There is no single honest “average self-published author income.” The answer changes with the definition of an author, hours worked, catalog age, genre, geography, and whether income includes speaking, courses, coaching, and other related work. The median is usually more helpful than the mean because a small number of high earners can lift the average substantially. Even a median must carry its population.
The Authors Guild’s 2023 Author Income Survey included 5,699 published authors and asked about 2022 earnings. Across all respondents, the median from books was $2,000 and the median from all author-related work was $5,000. Among full-time self-published respondents, the median from books was $12,800 and the median from all author-related work was $15,000. These are distinct groups. Quoting the higher figure as a typical outcome for a new part-time author would mislead.
The Alliance of Independent Authors’ 2025 data report reports a median gross income of $13,500 among its qualified indie survey respondents. Its PDF says it received more than 1,800 responses, with 1,520 meeting its screening criteria. “Gross” precedes production, advertising, software, travel, and other expenses. Its respondent pool and time period are not the same as the Authors Guild’s, so the two medians should not be combined into a trend line without careful adjustment.
The Written Word Media 2025 Indie Author Survey reports that 44% of its respondents earned $100 or less per month from books. That is a useful counterweight to success stories, but the same caution applies: the survey audience is not a probability sample of every person who self-published. Readers of an author-marketing outlet may differ from quiet hobbyists, high-volume publishers, and people who never subscribe to publishing advice.
| Income term | Include | Exclude or separate | Why it matters |
|---|---|---|---|
| Gross book revenue | Receipts before expenses | Profit after expenses | A large launch can be unprofitable if ads were expensive. |
| Net book income | Revenue after platform fees, returns, and direct costs | Unrelated service revenue | Better for unit economics. |
| Author-related income | Speaking, teaching, services linked to writing | Sales of unrelated businesses | A business book can create value beyond royalties. |
| Annual median | Middle respondent for a specified group | Forecast for any individual | Half of the measured group lies below it. |
| Monthly band | Respondents within a reported range | Continuous income for each month | Launches and seasonal sales are uneven. |
For a specific title, model your own economics with realistic price, royalty, production expense, and promotion budget. Our KDP ebook royalties guide explains the storefront calculation. A consultant’s book may be evaluated partly by qualified client conversations, while a workbook may be evaluated by direct sales and exercise completion.
Why do income surveys disagree?
They may all be accurate about their own respondents. One survey may recruit through professional writing associations; another through a marketing newsletter. One may filter for full-time authors or a minimum publication history. One may ask about gross revenue, another net book income. They may cover different calendar years, currencies, and countries. A sample of 1,500 can estimate its recruited group carefully and still fail to represent all authors.
A practical reading routine is to locate four lines before citing a survey: recruitment, eligibility, income definition, and time period. If any is unclear, state the uncertainty. Do not convert a monthly income band into an annual median or vice versa. Do not use “average” for a median. Do not merge a gross number from one survey with a net number from another in the same chart without conspicuous labels.
A worked example of misleading comparison
Imagine Survey A asks everyone who has ever published at least one title about book earnings. Survey B recruits active members of a business-minded author organization and excludes respondents who do not meet its commitment criteria. A $2,000 median in A and $13,500 median in B do not prove that income rose more than sixfold. They mainly show that B contains a different population. To estimate a trend, you would need comparable definitions and repeated measures over time. The exercise is fictional; the purpose is to expose the logic behind casual “author income is booming” claims.
What does a book actually cost to publish?
The ability to upload a file at no fee does not make a good book costless. Costs may include editing, fact-checking, illustrations, cover work, permissions, ISBNs, proof copies, design, accessibility testing, and promotion. A creator can do some of the work personally, but then time is a real input. The total varies too widely for a single dependable “average cost” here. Build an itemized budget for the format and quality level your reader expects.
| Cost line | Minimum question | When it is easy to underestimate |
|---|---|---|
| Editorial review | Who checks structure, clarity, and facts? | When subject expertise is assumed to replace reader testing |
| Rights and permissions | Are every image, quote, and worksheet element cleared? | When assets came from mixed tools and sources |
| Cover and interior | Can the book be read comfortably at its actual size? | When only a desktop PDF has been checked |
| Printing | What is the per-copy charge at chosen trim and page count? | When color interiors are assumed to cost like black and white |
| Selling costs | What platform fees, refunds, tax handling, and ads apply? | When gross price is mistaken for take-home revenue |
| Maintenance | Who updates links, prices, and policies later? | When the book promises current advice |
For an ebook, our AI book sale rules guide covers platform disclosure and rights. For a printable, how to sell workbooks considers file delivery and marketplace expectations. Neither substitutes for checking the current official platform terms before launch.
Do most self-published books sell fewer than 100 copies?
That claim appears frequently online, often with an exact percentage, but its chain of evidence is weak. We could not identify a public primary dataset that observes lifetime unit sales across all self-published titles and supports a universal percentage. Amazon does not publish title-level lifetime sales for the entire indie catalog, and direct sales are fragmented across platforms. Treat any exact figure without a traceable denominator and method as an anecdote dressed as a statistic.
The same problem affects “the average indie book sells 250 copies.” Average over what time frame? Which retailers? Are free downloads included? Are books with no ISBN counted? Is the number a mean or median? An author who wants to plan should use scenarios instead of an unverifiable universal average. For example, estimate the break-even number of copies under conservative, moderate, and optimistic reach assumptions, then revisit after a small test launch.
| Planning scenario | Copies | Net contribution per copy | Contribution before fixed costs |
|---|---|---|---|
| Conservative illustration | 25 | $4 | $100 |
| Moderate illustration | 100 | $4 | $400 |
| Stronger illustration | 500 | $4 | $2,000 |
These are arithmetic examples, not market benchmarks. The useful formula is copies × net contribution per copy − fixed costs = project contribution. Replace the inputs with actual format, store, printing, and promotion costs. If the book’s main purpose is to attract clients, calculate qualified inquiries separately rather than pretending the entire value is visible in royalties.
How much of self-publishing uses AI?
The NBER working paper on AI and book releases estimates that AI-containing books accounted for more than half of 2025 releases in its studied Amazon e-book data. That is a research estimate about a particular channel and a particular definition of AI-containing text. It is not a count of every self-published book, and it does not say every paragraph in those titles was machine-written. The paper also considers usage, which is the more relevant counterpart to an expanding supply of titles.
For individual publishers, the decision is operational: if AI helps with ideation or drafting, how will a human verify facts, add original experience, edit voice, protect rights, and disclose generated material where required? The U.S. Copyright Office’s AI initiative addresses human authorship. Amazon KDP’s content guidelines address platform disclosure. Different questions, different authorities. Our AI book writing statistics guide explains why tool-platform project counts must be kept separate from publication counts.
AI output is not a demand forecast
A faster production tool can lower the cost of putting a title into a catalog. It cannot make the title answer a reader’s question, maintain a story’s logic, or verify a citation. If output rises faster than attention, the average discoverability of an undifferentiated book can fall. Distinctive evidence, design, and trust become more important. A practical guide should include real worked examples; a story needs editorial continuity; a workbook needs exercises a user can complete without guessing.
What proportion of started books get finished?
The industry has no complete public denominator for “started books.” ISBN data count what reaches registration, not abandoned drafts. Surveys can ask writers to recall projects, but recall and recruitment bias are substantial. Tool vendors can measure project creation and export inside their own systems, yet the result cannot be generalized automatically to every author or genre.
Inkfluence AI publishes a 40.8% export-at-least-once rate for 102,854 projects it says were created through 15 September 2026. This is a vendor-reported platform metric, not an independently audited industry finish rate. The source page explicitly calls export a proxy. We cite it here only to illustrate the missing denominator problem. Export can mean “I want to proofread this draft,” and some finished books may never trigger an export event. Use the verb exported, not finished, when describing the measured event.
To increase your own chance of completion, define the smallest book that fulfills the promise. Write a one-page brief and a table of contents, then test a sample chapter with a real reader. Book projects stall when the promise keeps widening or the author cannot decide what “done” means. A bounded editorial checklist is more useful than an industry completion percentage whose population does not match your situation.
A practical launch dashboard for one author
Instead of chasing a general market figure, track a short sequence tied to your book’s purpose. Use the same definition each month and include both quality and distribution. The dashboard below works for a paid ebook, with adaptations for a lead magnet or service book.
| Stage | Measure | Review question | Possible action |
|---|---|---|---|
| Manuscript | Chapters reviewed by a person | Does each chapter fulfill a promised reader task? | Cut or rewrite vague sections. |
| Proof | Issues found on actual device or printed copy | Can readers navigate and use the visual material? | Fix layout and accessibility. |
| Listing | Page visits from relevant searches | Does the title match reader language? | Clarify metadata and sample pages. |
| Conversion | Paid copies divided by qualified visits | Does the preview justify the price? | Improve description or product. |
| Reader outcome | Feedback on tasks completed | Did the book deliver its promise? | Update confusing chapters. |
| Economics | Net receipts after direct costs | Is the project sustainable? | Adjust channel, price, or promotion. |
A consultant or educator should add downstream outcomes: subscribers who complete a useful exercise, prospective clients who cite a chapter, or students who apply a method. Do not infer causation from a traffic spike. Ask readers where they found the book, compare cohorts, and record uncertainties. Good measurement is modest and repeatable.
How should a first-time author choose a category?
Category counts can tell you what other publishers register; they do not tell you which niche has unmet demand. Start with a reader and a specific use case. Study the top competing books’ promises and reviews. Identify unanswered questions, stale information, poor examples, or weak visual explanations. Then test whether you can supply something credible that others cannot. A narrow topic with firsthand experience may be more defensible than a broad trend-driven title.
| Audience | Weak broad promise | Stronger, testable promise | Evidence to gather |
|---|---|---|---|
| New consultants | “Everything about business” | “Price a first advisory package and explain it to clients” | Interviews, sample proposals, objections |
| Local travelers | “The ultimate city guide” | “A car-free weekend with accessible routes” | Verified transit, opening times, route testing |
| Coaches | “Get fit fast” | “Plan four weeks of beginner movement safely” | Qualified review, clear adaptations, user tests |
| Parents | “Fun activities for kids” | “Ten low-prep rainy-day drawing sessions for ages 5–7” | Child testing, supply lists, readable layouts |
These are editorial examples, not claims that EarnDraft currently has specialized templates for every subject. For a human-led general book, start with the drafting flow and bring your own research and editorial judgment. The guide hub connects to format-specific publishing advice.
What should a good statistics page disclose?
When you publish your own analysis, give the reader a route back to the raw observation. State the producer of the data, the period counted, inclusion and exclusion rules, the unit of analysis, the sample size or population, missing data, and calculation method. Name the limitations where the headline appears. If you use another company’s figures, link and attribute them rather than laundering them into your own authority. If you publish a chart, label its axis and denominator.
For an income chart, specify gross versus net, book-only versus all author-related work, annual versus monthly, and mean versus median. For a publishing-output chart, specify ISBN records versus unique works and U.S. versus global coverage. For a tool-usage chart, distinguish created projects from exported manuscripts and store listings. Update the count date only when you recount the data. Preserve older values in a changelog if the page makes trend claims.
A citation rewrite exercise
Take the sentence “Three and a half million Americans published books last year.” Rewrite it as: “Bowker recorded more than 3.5 million self-published U.S. ISBN-bearing titles for 2025, according to Publishers Weekly’s March 2026 report.” The rewritten sentence avoids assuming one title equals one person, identifies the geography and counting unit, and tells readers where to check it. This is a small edit with a large gain in trust.
Methodology, source limits, and update policy
We read the linked public source pages in September 2026. Bowker’s 2025 figures are available to us through Publishers Weekly’s report of its compiled data and interview; we do not claim to possess Bowker’s underlying record-level dataset. The income figures come from the organizations’ survey publications. The NBER statement is a working-paper estimate within its studied Amazon e-book context. Inkfluence AI’s export count is a vendor claim about its private production records and is kept clearly attributed.
No EarnDraft account or book data were used. We did not calculate a global count, an average lifetime sale, or an author earnings forecast. Figures can change as their publishers revise or release newer studies. Before quoting them in a business plan, check the original page and observation year. This page should be reviewed when Bowker, ALLi, Written Word Media, the Authors Guild, or NBER publish materially newer comparable evidence, rather than receiving a date change without an evidence change.
Frequently asked questions
How many self-published books are released each year?
Bowker recorded more than 3.5 million self-published U.S. ISBN-bearing titles for 2025, as reported by Publishers Weekly. That is a registered-title count for one country. It misses books without an ISBN and may count multiple formats of one work separately.
Are self-published books outselling traditionally published books?
The Bowker output ratio does not answer a sales question. It compares registered title counts, not copies sold or revenue. Use retail sales data with matching coverage to analyze sales.
How much does the typical indie author earn?
There is no universal answer. The Authors Guild reported a $2,000 median book income for all respondents in its survey about 2022 and $12,800 for full-time self-published respondents. ALLi reported a $13,500 median gross annual income among qualified 2025 indie survey respondents. The populations, years, and income definitions differ.
Is it true that 90% of self-published books sell fewer than 100 copies?
We have not found a traceable public primary dataset supporting that exact market-wide percentage. Avoid using it as fact. Plan your project with channel-specific, transparent sales scenarios instead.
Did AI cause the 2025 increase in self-published titles?
The title count alone cannot establish causation. AI tools may reduce production friction, but classification changes, service providers, and other factors also affect registrations. Research on AI-containing titles is separate evidence and should not be used as a direct explanation for every Bowker change.
Does an ISBN mean someone bought the book?
No. An ISBN identifies a publication edition. Registration does not establish a sale, review, or read.
What should I measure for my first book?
Choose a reader outcome first. Then track a completed and tested manuscript, relevant listing visits, conversion where measurable, reader feedback, and net project economics. If the book supports a service, track qualified inquiries separately from royalties.
Can an author use AI and still self-publish?
Yes, subject to the current rules of the platforms and rights involved. KDP has disclosure requirements for AI-generated content; the U.S. Copyright Office distinguishes human-authored expression from material whose expressive elements were determined by a machine. Review both before publishing.
Primary reading and next steps
- Publishers Weekly: Bowker’s 2025 U.S. ISBN title output, March 2026. Direct report of the Bowker figures and Bowker representative’s explanation.
- Bowker’s description of title submissions and publisher categories. Background on the U.S. ISBN agency and its classification.
- Authors Guild 2023 Author Income Survey. Respondent sample and 2022 median earnings.
- ALLi’s 2025 Indie Author Data Drop. Qualified responses, gross income, and method.
- Written Word Media 2025 Indie Author Survey. Reported monthly income bands and marketing challenges.
- NBER working paper on AI and book quantity and usage. Scope and limits of the AI-containing-title estimate.
- U.S. Copyright Office AI initiative and KDP content guidelines. Rights and platform policy context.
The best use of these numbers is a better brief. Write down the exact reader, the decision your book helps them make, the evidence you can contribute, and how you will test the finished artifact. Then build a draft and use the publishing guides to choose an appropriate format and channel.