
What it is
A book, read again as a set of posts
Most books hold more finished thinking than their author will ever publish anywhere else. The chapters are drafted, argued over, edited and set, and then they sit inside one object that a stranger has to buy before reading a line of it. Chapter to Blog begins from that fact. It takes a book that already exists and sets each of its chapters out again as a long post for the open web, written to be found by search, with the chapter it came from kept whole beside it.
The order of the work is plain, and the app keeps to it. A person brings the book in, by pasting its text, uploading a file or giving the address of a public Google document. An AI reads the whole of it and divides it into chapters. The person sees those chapters listed, names the book, chooses a length and decides whether each post should carry a picture. Then an AI rewrites each chapter as a post with its own title, meta description and keywords, its tags and its suggestions for internal links, and everything is kept in a library that holds every book the person has brought in.
Nothing it makes is sent to a blog, a network or a mailing list; every post leaves by a copy button or a downloaded text file, at the moment a person chooses.
What it is not matters as much as what it is. It does not publish. Nothing it makes is sent to a blog, a network or a mailing list; every post leaves by a copy button or a downloaded text file, at the moment a person chooses. It does not measure how a post performs, and it makes no claim about where a post will rank. The AI is instructed to write for search, and the app shows exactly what the AI wrote, field by field, so that a person can judge every line before any of it goes out.
Beside every finished post sits a second room, called AI Refine. There an AI will summarise a post into takeaways, expand it, rewrite it in another register, or draft posts for six social platforms at the counts a person sets. Two of those results can be applied to the saved post. The others are drafts to copy, kept only as long as the screen is open, which is the honest measure of what they are.

A day with it
Tuesday, an editor and twelve chapters
The managing editor of a small press of practical books arrives at half past nine with one job on her list. The author of the press's spring title, a careful guide to keeping bees, has sent the final manuscript as a Word file, and the press wants the book's chapters on its own site before the printed copies arrive. She opens Chapter to Blog and finds the first step waiting: an upload field that names the three kinds of file it reads, a line for a Google document's address, and a wide box for pasted text.
She chooses the file. The upload field says Processing file while the document is sent up and its text extracted, and then the whole manuscript appears in the box, with its length counted underneath it in characters, a little over three hundred thousand of them. She presses Detect Chapters. The whole book goes to an AI with a single instruction: divide it into chapters, keep every word as it is, and where there are no chapter markers, divide it by sense.
Twelve chapters remain, each with its post and its source.
The book returns as a list. Twelve Chapters Detected, the heading says, and beneath it twelve cards, numbered from 01, each with its title, its first two hundred characters, its length and a reading time estimated at two hundred words a minute. The chapter on swarming reads as an eleven-minute piece. The one on winter feeding, the shortest in the book, reads as four.
The title field has already filled itself from the file's name, which still carries the author's working title and the word final. She types the book's real title, because this is how she will find it in the library in a month. Then come the two choices. Target length is set to full, which tells the AI that each post must be as long as its chapter or longer and must keep at least ninety-five per cent of the text; she leaves it there. She turns on the switch for featured images, which asks the AI to describe a picture for each post and then has one made from that description.
She presses Process All, with the count of twelve beside the words. The screen becomes a single quiet progress page: a line that says the posts are being created, a count of chapters completed out of twelve, a note that images are being made, a tally of successes and failures, and a bar that fills as each pair of chapters finishes. With pictures switched on, the chapters go two at a time rather than three. She goes downstairs for coffee.
When she returns, the screen reads Your Blog Posts Are Ready. It shows eleven posts, the number of characters across all of them, and the change in length against the original chapters as a percentage. One chapter has not come through, and a card says so plainly, with a button that opens the book's own page to retry it. She opens the first post. Preview shows it rendered with its headings; Raw Text shows the Markdown she will paste into the press's site; Meta shows the title the AI wrote, its meta description and the chapter it came from. She copies it.
The book's page holds the rest. Every chapter is there in order with a badge for its state, and the one that did not come through carries its error in a line of plain text and a count of one retry. She presses Retry on that chapter alone. It runs at full length and without a new picture, because a retry from this page uses the default options, and within a minute its badge reads completed. She opens the chapter on swarming and reads the Original tab against the Blog Preview, paragraph by paragraph, to see what the AI has added: a hook in the opening lines, subheads every few hundred words, two calls to action she will cut, and the author's own story of a swarm settling on a churchyard yew, kept almost word for word.
In the Meta tab she finds the featured image and its alt text, the focus keyword, seven secondary keywords, ten tags and four topics the AI suggests linking to. Two of those topics are other books on the press's list, and she makes a note of them. Then she opens AI Refine. She asks for a summary and copies its executive summary into the draft of the press's newsletter. She chooses Mastodon and Bluesky at five posts each, and the button counts ten posts before she presses it; each draft comes back with its length counted against its platform's limit, and she copies three.
Last, she asks for an expanded version of the chapter on winter feeding, because the author has asked for more there. She reads the longer text and the list of what was added, cuts nothing yet, and presses Apply Changes, which replaces the saved post and updates the book's totals. Before she closes the app she goes to My Books and deletes a test upload from the week before. The app asks first and says that the book and all its chapters will go and cannot be brought back, and she confirms. Twelve chapters remain, each with its post and its source. Nothing has been sent anywhere. She will place the first post on the press's site on Thursday, by hand.

Bringing it in
Three doors for a manuscript, and a reading of its shape
A book can arrive in three ways, and the first screen holds all three at once, separated by a quiet rule and a line of small capitals. The first is a file. A plain text file is read directly in the browser, with nothing sent anywhere before the person asks. A PDF or a Word document, in either of its formats, is uploaded and its full text extracted by the platform's document reader, which is asked for one thing only: the complete text of the document as plain text, with its paragraphs and chapters intact. The name of the file appears beneath the field once it is read, so a person can see which draft they have chosen.
The second door is the address of a Google document. The app checks that the address is a Google document and that it carries a document's identity, and then it asks an AI with access to the web to fetch the document and return its complete text, exactly as written, without summarising it. There is no connection to a Google account here, and none is needed or claimed. The screen says plainly that the document must be set so that anyone with the link can view it, and if the text cannot be fetched, the app says so in the same words rather than leaving the box half filled.
Read down the column, the cards show a book's proportions at a glance: where the long chapters sit, where a short one might really be a preface, whether the AI has kept the introduction separate from chapter one.
The third door is the box itself. A person can paste a manuscript into it, or correct whatever the file or the document brought in, before anything else happens. Under the box a running count shows the length of the text in characters. It is the first moment a person sees the book as a measured object rather than a file on a desk.
Detect Chapters sends the whole of that text to an AI in a single reading. Its instruction is to return each chapter with its title and its full text, to keep every word exactly as it is and to shorten nothing, and, where a manuscript has no clear chapter markers, to divide it by logical sections instead. That last clause matters for essays, reports and long talks, which often have shape without headings. Because the whole book goes to the AI at once, a very long manuscript asks a great deal of that reading, and a careful person checks the list that comes back against the book.
The list is built for exactly that check. A heading states how many chapters were found. Each chapter is a card with a two-figure number in its corner, its title, its first two hundred characters and its length, and an estimate of the minutes it would take to read at two hundred words a minute. Read down the column, the cards show a book's proportions at a glance: where the long chapters sit, where a short one might really be a preface, whether the AI has kept the introduction separate from chapter one.
Above the cards is a field for the book's title in the library. The app fills it in advance from the file's name, or from the first chapter's title, or, failing both, from the date, and a person replaces it with the name they will recognise later. A Back button returns to the first step with the text still in the box. Nothing has been saved yet. The book enters the library only when the person decides to process it, so a first look costs nothing to the record.

The chapter is never overwritten. It waits in its own tab, whole, while the post is read against it line by line.

The post
What the AI is asked to write, and what comes back
Every post begins as a written instruction to an AI, and the instruction is specific enough to quote. The AI is asked for a title of fifty to sixty characters, with the focus keyword near its start, and for a meta description of one hundred and fifty to one hundred and sixty characters that carries the same keyword. It is asked to name one primary keyword and five to seven secondary ones, eight to twelve tags for the post's categories, and three to five topics the post could link to elsewhere on the same site. Inside the post it is asked to use the keyword in the first paragraph, in the subheadings and in the conclusion, and to set headings in proper order.
The same instruction asks for things a reader of the house would weigh with care. It asks the AI to preserve the author's voice and style, to add a hook to the opening, to place a subheading every three or four hundred words, to make the text easy to scan with bullet points where they fit, and to include two or three calls to action, one of them at the close. A person who wants the chapter's own quiet kept will read each post and cut what does not belong, and the app places the original in the next tab so that the cutting is easy.
A result under fifty characters is treated as a failure rather than accepted as a post.
Length is the first choice a person makes, and it is made once for the whole book. Full length tells the AI that the post must be as long as the chapter or longer, never shorter, and must keep at least ninety-five per cent of the original text. The other four choices aim at about fifteen hundred, two thousand, two thousand five hundred or three thousand words a post, and tell the AI to condense or expand towards that figure while keeping what is most valuable. Under the choice a single line says which it is doing: preserving and expanding the original, or aiming at the number. The register of the writing is set later, post by post, in the refinement room, where a person can ask for a different tone once they have read what the first pass made.
The second choice is the picture. With the switch on, the AI writes a description of a featured image and alt text for it as part of the same answer, and the app then has the image made from that description in the plain, clean style of a professional blog. If the image cannot be made, the post is kept without it, because a missing picture is not a reason to lose a finished chapter. The picture and its alt text appear later in the post's Meta tab, where a person can see whether the description fits before they use it.
The chapters are worked in small groups. The book and every chapter are saved first, each chapter marked pending, so that the record exists before a single word is rewritten. Then the chapters go to the AI three at a time, or two at a time when pictures are being made, and each moves through its own states: pending, processing, completed or failed. A result under fifty characters is treated as a failure rather than accepted as a post. A failure keeps its reason as a message on the chapter and adds one to that chapter's count of attempts, and the other chapters carry on without it.
When every chapter has finished, the app adds up the book: how many characters went in, how many came out, and whether the book is complete. A book whose chapters have all come through, with or without a failure among them, is marked completed, so that the posts that did arrive can be used while the one that did not waits for its retry. The person sees all of this on the results page, where the posts open one at a time to a rendered preview, the raw Markdown, and the fields that sit around a post when it is published.

The library
Every book kept, every chapter accounted for
The library is called My Books, and it holds every book a person has brought in, newest first. Each book is a card with its title, a badge for its state, the number of its chapters, where it came from (the file's name, a Google document, or pasted text), the date it was added and, once posts exist, the change in its length as a percentage. An empty library says that there are no books yet and offers a single way to add the first. A button at the head of the page starts a new book. The book a person uploaded on Tuesday is still there on Friday, in the same order, with the same badge.
A book's own page is where the work of a second reading happens. At the top sit the title, the number of chapters and how many have been processed, with Download All, Retry Failed and Process All beside them whenever there is something for each to do. Download All saves every completed post as one text file named after the book, each post set out with its title, its meta description, a rule and the post itself, so that a person can hand the whole set to a colleague or keep it in a folder with the manuscript.
Original holds the chapter exactly as it came in, title and text.
Every chapter appears in order beneath, as a row that opens. Its badge says whether it is pending, processing, completed or failed; a chapter that has been tried more than once says how many times; a chapter that failed says why, in its own words. A pending chapter has a Process button and a failed one a Retry button, and the same can be done for all of them at once, with a running count while the group is worked. Chapters processed from this page run at full length and without a picture, whatever was chosen on the first day, which is worth knowing before retrying a chapter in a book that was set to two thousand words.
Opened, a chapter shows its tabs. Original holds the chapter exactly as it came in, title and text. For a completed chapter the next tabs are Blog Preview, which renders the post as a reader would see it; Raw Text, which shows the Markdown for pasting; and Meta, which holds the featured image with its alt text, the post's title and meta description, its focus keyword, its secondary keywords and tags set out as small marks, and the topics the AI suggests linking to. The last tab is AI Refine. A copy button on the row itself copies the post without opening anything.
It is worth being exact about what can be changed here. The post's text changes when a person applies an expanded or a re-toned version from AI Refine, and the chapter's length and the book's totals are recalculated when they do. There is no editor in the app for typing into a post or a chapter by hand, and the book's title is set before processing rather than afterwards. Editing by hand happens wherever the post goes next, which is why the raw Markdown sits one tab away.
A book can be taken out of the library entirely. The delete button asks first, names the book, and says that it and all its chapters will be removed permanently and that this cannot be undone; on confirmation the app removes every chapter and then the book. There is no archive and no retiring a book to a back shelf. The confirmation is the whole of the safeguard, and it is stated in full, so a person deletes knowing exactly what goes.

Refining
Four second readings of a finished post
AI Refine sits in the last tab of every completed chapter, and it does four things, each behind its own tab and each done by an AI on the post as it now stands. The first is Summarize. It returns an executive summary of two or three sentences, five to seven key takeaways, and a condensed version of the post at about thirty per cent of its length. Each of the three has its own copy button, and each takeaway can be copied on its own, which suits a newsletter, a contents page or the back of a sales sheet. The summary is a draft for elsewhere. It is not saved to the chapter, and it is gone when the chapter is folded shut.
The second is Expand. The AI is asked to make the post at least half as long again while keeping its voice, adding fuller explanation of its main ideas, examples, further perspectives and practical advice, and supporting figures where it judges them appropriate. It returns the longer text and a short list of what it added, so a person can see the additions before reading the whole. Two buttons follow: Copy, and Apply Changes, which replaces the saved post with the longer one. Any figure the AI adds is its own and is not checked by the app, so a careful editor reads every one against the book.
Nothing is posted and nothing is scheduled.
The third is Tone. Six registers are offered: formal and professional, casual and friendly, persuasive and compelling, inspirational and motivating, educational and informative, and humorous and witty. The AI rewrites the post in the chosen register at about the same length, keeping its information and its points, and returns a new title for it along with a short note on what changed in the voice. The rewritten body can be applied to the saved post. The new title is shown with its own copy button and is not written into the record, so a person who prefers it carries it across by hand.
The fourth is Social. Six platforms are listed as marks that a person turns on or off: Instagram, Facebook, X, Pinterest, Mastodon and Bluesky. LinkedIn is not among them. For each platform the app carries a character limit and a hashtag limit, from two hundred and eighty characters on X to more than sixty thousand on Facebook, and the AI is asked to keep every draft inside them and to suit each platform's manner: more visual on Instagram, conversational on Facebook, brief on X, descriptive on Pinterest, thoughtful on Mastodon, casual on Bluesky. A person chooses between three and twenty posts a platform, and the button counts the total before it is pressed.
The drafts come back grouped by platform, each group folded until it is opened. Every draft shows its number, its own copy button and its length set against its platform's limit, so a draft that runs long is visible at once. Copy All gathers every draft for every chosen platform into one block of text, with the platform names as headings. Nothing is posted and nothing is scheduled. The drafts live on the screen until the person copies what they want and moves on.
Taken together, the four tabs are the old work of an editor's second pass, done quickly and set out in the open. A post can be condensed for one use, lengthened for another, moved to a different register for a different reader, and carried outward in short pieces, and in every case the person sees the result in full before deciding whether it touches the saved post. Only two of the four can touch it, and only by a button that says what it does.

Twenty drafts for one platform, each counted against its limit, and not one of them sent anywhere.

In the suite
Where it sits among the nineteen
Chapter to Blog is one of the nineteen applications of B-Corps, the house's own operating system, which the house runs on every day and now offers to organisations. It is installed on the house's own Mac by its own method, against its own cloud backend, on one desk that starts, opens and rests each application, and every two minutes restarts any that has stopped. A stock-take of every application's records and functions is kept with the suite, and this one's entry is short and exact: two kinds of record, a book and a chapter, and one function for reporting to the house's registry when that registry's address and key are set. The house's reading room reads it, as it reads every application and every book.
Like each of its siblings, it keeps its own sign-in today. A person signs in to it on its own terms, and nothing in its code passes data automatically to another application. What joins it to the rest of the suite is the way the suite is arranged to be used, and in the publishing layer that arrangement is close and practical, because the applications there take up a book at different moments of its life.
A press that runs all of them works one manuscript from several sides without retyping it, with each application doing the one thing it does.
A person who keeps ContentSpark open beside Chapter to Blog has the natural next step to hand. ContentSpark takes one source and sets it out across every channel, and into a course, so a post finished and applied here can become the source it works from. SquareDash, the marketing workspace for websites, is where a downloaded or copied post can be placed among a site's pages and articles, and from SquareDash a bridge turns articles into drafts on the house's Ghost publication, always as drafts. The Luminous Book, the brand studio at the centre of the suite, writes down a business's voice, and that written voice is the thing an editor holds a re-toned post against before applying it.
The other publishing applications meet this one at the same book. Luminous Bookflow carries the book through a publishing house in software; Chapter to Blog carries its chapters outward to the open web. Audiobook AI prepares the same manuscript for listening. GrowthWrite gives a studio a site whose blog the chapter posts can fill. OmniWriter keeps the writer's desk, with the writing, the posts, the people and the book's own page in one place. A press that runs all of them works one manuscript from several sides without retyping it, with each application doing the one thing it does.
None of this is a promise of wiring that does not exist. It is a description of a working room, where each application stays small and exact, and a person moves a finished piece from one to the next with a copy, a download or a paste, knowing at every step where the text is and who placed it there.

Who it is for
The person with a finished book and an empty blog
The app is for anyone who holds a finished book and a site that says too little. An author whose book took four years and whose blog holds three posts. An editor at a small press who wants each title's chapters on the press's own pages, under the press's own address, before the reviews arrive. A practitioner who wrote a book about their method and wants the method findable by the people searching for it at two in the morning. A consultant whose long reports are, in practice, books nobody outside the client has read. What they share is a body of finished work and a lack of hours to set it out again.
What it replaces is usually a slow ritual done by hand. A person opens the manuscript, copies a chapter into a chat window, asks for a post, copies the post back into a document, asks again for a title and a meta description, keeps a spreadsheet of which chapters are done, and loses the thread somewhere around chapter seven. Or the work goes to a freelancer, chapter by chapter, at a pace set by someone else's calendar. Chapter to Blog does the same work in one place, in the book's own order, with every chapter's state written down and every field the AI produced shown where a person can read it.
Every chapter keeps its original text in its own tab, untouched by the rewrite, so the comparison an editor has to make is one click away.
Three things set it apart from a general writing tool. The first is that it begins from the whole book rather than from a prompt. The chapters are found, numbered and kept together, so the posts remain a set with a source, and the library shows at any moment which of a book's chapters exist on the web and which do not.
The second is that the source is never lost. Every chapter keeps its original text in its own tab, untouched by the rewrite, so the comparison an editor has to make is one click away. The house reads this as a matter of respect for the author. An AI that rewrites a chapter will add things and remove things, and a person who can see both versions side by side can keep what the author wrote and cut what the instruction added.
The third is that it is honest about failure. A chapter that does not come through says so, keeps the reason, counts its attempts and waits for a retry, while the rest of the book carries on. Nothing is reported as finished that is not, and the results page names the number that failed beside the number that succeeded. For a press working to a launch date, that one property is worth more than speed, because it means the list of what remains is always true.
How it is offered
Installed for an organisation, and kept
Chapter to Blog is offered to an organisation in three ways: on its own, as part of the publishing floor, or inside the whole of B-Corps, where every application on these pages is installed together, set up in the organisation's own voice and with its own records, and kept running for as long as the organisation keeps it. The house carries each application into its suite for good, so what is set up now is still there, and still mended, in the years after.
The fee reflects the work of installing and shaping it, and the size of the organisation it serves. It is set with each organisation, after a conversation about what the application will carry for it.
Arranged in conversationChapter to Blog · B-Corps

The care
What it keeps, what it asks, and where it stops
The care in Chapter to Blog is mostly the care of keeping things. The book's full text is saved with the book, and each chapter's original is saved with the chapter, so that no rewrite ever stands where the source used to be. Every chapter carries its own state, its own count of attempts and, when something goes wrong, its own reason in plain words. A failed picture is not allowed to sink a finished post. A failed chapter is not allowed to stop the rest of the book. These are small decisions in the code, and together they mean that what the library says is true.
It is careful, too, about what it changes. The first pass writes each post once. After that, the saved post changes only when a person applies an expanded or re-toned version, by a button that says Apply Changes, and the book's totals are recalculated in the same moment. Summaries and social drafts never touch the record at all. Deleting a book asks first and states the whole consequence: the book and all its chapters, permanently, with no undoing. A person is never surprised by what the app has done to their work.
A book is written once and read for a long time.
When the app is set to require a sign-in, it sends a person to sign in before anything opens, and it returns them to the page they asked for. A person who reaches the app without being registered for it is told so directly, with the steps to take: check the account, ask the administrator for access, sign out and in again. It sends nothing in a person's name. Every word it makes leaves by a copy button or a downloaded file, at a moment the person chooses, which keeps the author's name in the author's hands.
And it names its AI at every step where one works. An AI divides the book into chapters, an AI writes each post and its picture's description, an AI fetches a public Google document's text, and an AI does each of the four refinements. The instructions it gives are specific, and the app shows their results in full rather than describing them. It does not claim that the AI has kept every word, or that any figure the AI adds is right, or that a post will be found. It gives a person the whole of what was made, beside the whole of what it was made from, and lets them decide.
A book is written once and read for a long time. Its chapters can be read again in another form, by people who will meet the book first as a post found on an ordinary evening, and the app exists to make that second reading possible without losing the first. The chapter stays where it was. The post goes out into the world, and the book is still on the shelf when the reader comes looking for it.
B-Corps
