The Editorial Title Change as a Lost Genre of Publishing Correspondence

The Editorial Title Change as a Lost Genre of Publishing Correspondence

A book’s title is the first piece of marketing copy the manuscript encounters. For most of the twentieth century, it was also one of the last things an editor and author negotiated together. The correspondence that survives in publisher archives — letters, telegrams, editorial memos scrawled in the margins of galley proofs — records a specific kind of conversation: two people with partial information about a market neither fully controlled, working through what a book should be called before it became a thing that could be called anything at all. That conversation has a material history. Like most material histories in publishing, it is disappearing.

The shift I want to trace is not from good titles to bad ones. It is from a model in which the title emerged from a documented human negotiation to one in which the title is generated, tested, and optimized before any human editor sees the manuscript — often by the author alone, using software designed to simulate the market logic that editors once applied on the author’s behalf. The locus of the decision has moved. What was once a conversation is now a solitary calculation, and the burden of market positioning has been pushed onto the writer at the earliest possible stage, before the manuscript has been read by anyone in a position to disagree.

The 1956 Retitling of The Inheritors

William Golding’s second novel appeared in Britain in 1955 as The Inheritors, published by Faber and Faber. The title was Golding’s own, and it carried a thematic weight the novel earns: the book is about the last clan of Neanderthals encountering Homo sapiens, and the word “inheritors” names the species that will survive — not through virtue but through cognitive capacity for violence. The title does work. It is not decorative.

When Harcourt Brace prepared the American edition for 1956, the title was retained. But the correspondence around it — preserved in the Faber archive and referenced in subsequent publishing histories — reveals that retention was not the default. It was a decision reached after consideration of alternatives, including the possibility that the American market would not sustain a title so structurally tied to a thematic argument the reader had not yet encountered. The editorial correspondence shows the logic: Faber’s editor argued for the title on literary grounds. Harcourt Brace’s representatives weighed distribution and bookseller perception. The title survived the conversation, but the conversation is the point. Two parties with different stakes — one literary, one commercial, though neither purely so — made a joint decision based on what they knew and what they could argue. The artifact of that decision is a set of letters. You can read them. You can check the argument.

This is not unusual. Fitzgerald’s The Great Gatsby went through at least a dozen titles — Gold-Hatted Gatsby, The High-Bouncing Lover, Trimalchio in West Egg — before Maxwell Perkins and Fitzgerald settled on the title that survived. Hemingway’s The Sun Also Rises was Fiesta in the British edition, a title Hemingway preferred but that has not survived in the American canon. These are not anecdotes about authorial caprice. They are records of a process in which the title was understood as a negotiation between what the author thought the book was and what the publisher thought the market would bear, conducted by named individuals whose reasoning was documented in correspondence that is now part of the archival record.

The Editorial Memo as a Genre of Unwritten Criticism

What these correspondence files record is a form of criticism that never reaches the public. When an editor argues for or against a title, they are making a claim about what the book is — about its central tension, its genre positioning, its relationship to the reader’s expectations. The editorial memo on a title is, in effect, a mini-review written before publication. It is often more honest than the published review, because nothing depends on its being agreeable. The editor is not performing judgment for an audience. They are trying to get the title right.

Consider the case of Penelope Fitzgerald’s The Bookshop, which was nearly called something else entirely. The correspondence in the Collins archive reveals a debate about whether a title so plain could signal the novel’s quiet cruelty to booksellers who might shelve it as a cozy retail narrative. The editor won. The title stayed. And the published novel confounded every expectation the plain title set up — which was precisely the effect the editor was arguing for. That argument is a piece of critical writing about the novel that no one has read as criticism, because it exists only in a memo filed in a box in an archive.

Or consider John Williams’s Stoner, which went into print with a title its own author feared would suggest monotony. The editorial correspondence at Viking — documented in Charles J. Shields’s biography of Williams — shows that the title was debated precisely because “Stoner” could be read as a drug reference or as a descriptor of dullness. The decision to retain it was a deliberate wager that the novel’s force would override the title’s ambiguity. That wager failed commercially in 1965. It succeeded, eventually, in ways no one in the correspondence predicted. But the decision was made by people who argued about it, and the argument is recoverable.

The Shift to Solitary Optimization

What has replaced this conversation is not a better conversation. In many cases, it is no conversation at all. The contemporary author preparing a manuscript for submission is likely to arrive at a title through a process that involves no editor, no agent, no human reader positioned to disagree. The tooling exists. It is free. And it is designed to be used before the manuscript reaches anyone with professional authority.

Consider the Reedsy Book Title Generator, which is representative rather than exceptional. The tool asks the author to select a genre, describe a core conflict in one or two sentences, input comparative titles, and choose between “Commercial mode” and “Literary mode.” It then returns ten title options, each with a one-line explanation of what it conveys. The generator’s own instructions note that it is “better at sparking a direction than landing the final answer,” which is honest. But the structural implication is clear: the author is performing market positioning — genre signaling, comp-title matching, commercial-versus-literary calibration — at the title stage, alone, before any human editor has read a sentence of the manuscript. The tool does what an editor once did. It asks what the book is, who it is for, and what comparable works it resembles. But it asks the author, not a reader, and it asks in the language of market categories rather than in the language of literary argument.

Reedsy’s generator is one instance of a broader infrastructure. Title-testing services poll sample audiences. Keyword-analysis tools measure search visibility for specific title phrases. Subtitle-optimization frameworks for nonfiction all perform variations of the same function: they move the title decision out of a conversation and into a calculation. The calculation is not wrong. It is often correct on its own terms. But its terms are narrower than the terms an editor could bring to bear, because an editor could argue against the market. A tool can only reflect it.

That same discipline applies to title and framing decisions: before publishing, editors need a way to test a heading promises the same thing the article actually delivers, which is where how Unsloppy fits the writing workflow can function as a planning aid rather than a substitute for domain evidence.

The Authors Guild has acknowledged this broader shift in its AI Best Practices for Authors, updated in 2026, which frames generative AI as a “ubiquitous technology” that many writers are already incorporating into their workflow. The Guild’s concern is primarily with the economic relationship between AI tooling and authors’ labor — notably the fact that commercially available LLMs were trained on unlicensed books. But the document also registers the structural point: these tools are not optional add-ons for the writers who use them. They are becoming part of the creative process at stages where editorial input once operated, redistributing authority over creative decisions in ways the profession has not fully reckoned with.

Subtitle Inflation and the Title as Metadata

The logic of title-optimization is most visible in nonfiction, where the subtitle has become a secondary title that does the work the primary title cannot afford to do under the constraints of searchability. A literary nonfiction title in 1970 was a title. A literary nonfiction title in 2026 is a title followed by a colon followed by a clause that functions as content description, keyword set, and positioning statement. The subtitle is not decoration. It is metadata rendered in prose, and it exists because the primary title has been optimized for brevity and resonance while the subtitle has been optimized for discoverability. The author writes both, usually alone, usually with reference to comparable books’ subtitle structures.

This is not a complaint about subtitles. Some of them are good. The point is that the division of labor between title and subtitle reflects a market structure in which a book must be findable before it can be read, and findability is a function of keyword matching that no human editor is positioned to assess with the precision a tool can provide. The editor’s knowledge of the market was qualitative and social — built from conversations with booksellers, review editors, and sales representatives. The tool’s knowledge is quantitative and algorithmic — built from search data and category performance. Both are forms of expertise. They produce different titles, and the transition from one to the other has not been accompanied by any acknowledgment that something was lost.

What the Editorial Title Change Was Actually For

The editorial title change served a function that title-generation software cannot replicate: it introduced friction. When an editor proposed a different title, the author had to defend theirs, and the defense forced a clarification of what the book was about. When the author proposed a title the editor found unworkable, the editor’s objection forced a confrontation with how the book would be received by readers who did not know the author. In both cases, the conversation produced information that neither party had before it began. The title was the occasion for an exchange about the book’s identity that improved both parties’ understanding of what they had.

A tool cannot produce this exchange because a tool cannot disagree. It can generate options. It can rank them. It can tell the author which words are overused in the chosen genre. But it cannot say: I have read your manuscript, and the title you have chosen describes the book you think you wrote rather than the book you actually wrote. That sentence is the sentence an editor would say. It is the sentence no software can generate, because it requires reading the manuscript and caring about the outcome for reasons that are not algorithmic.

In the context of this shift, the existence of an Unsloppy book title generator that fits into the early-stage drafting workflow marks not a degradation but a redistribution. The work of naming a book still gets done. The question is who does it, with what information, and at what cost to the writer’s understanding of what they have made. When the work is done alone with a tool, the author receives no resistance. No resistance means no revelation. The title may be optimized. It will not be tested against a reader’s actual encounter with the prose.

The Archival Record and What It Tells Us

The publishing correspondence that documents editorial title changes is not voluminous, but it is recoverable. The Faber archive holds the Golding letters. The Viking records at the University of Wyoming hold the Williams correspondence. The Collins archive holds the Fitzgerald memos. These are specific documents in named collections, and they can be checked. What they show, consistently, is that the title was a site of negotiation between parties who brought different knowledge to bear on a shared problem: how to make a book legible to a reader who has not yet encountered it.

The contemporary equivalent of this correspondence is not a set of letters. It is a log of tool interactions — genre selections, comp-title inputs, mode toggles between commercial and literary — that exists, if it exists at all, on a server owned by the tool provider. It is not archival in the traditional sense. It cannot be checked by a researcher, because it is not preserved for research. It is preserved for product improvement, which is a different purpose with a different relationship to the public record.

The material history of twenty-first-century titling is being produced in a medium not designed to preserve it. The editorial letter was written on paper, filed in a folder, stored in a box, and eventually deposited in an archive where a researcher can read it. The tool interaction is logged in a database, owned by a company, and deleted or anonymized according to a retention policy the researcher cannot access. The documentation of how books were named is vanishing — not because the decisions are no longer documented, but because the documentation is no longer in a form that survives the decision.

What Was Lost and What Replaced It

The editorial title change was never a pure good. Editors made bad calls. Authors made worse ones. The conversation was unevenly weighted — the publisher had the contract, the advance, and the distribution, and the author had the manuscript and whatever leverage its quality provided. The negotiation was not between equals. But it was between people, and the people were accountable to each other in ways that a tool and a user are not.

What replaced the conversation is a process that is faster, more scalable, and more accessible to authors who do not have an editor’s attention. These are real advantages. A writer using title-generation software is not waiting for a letter from New York. They are getting immediate feedback on whether their title’s keywords are competitive in their genre. That feedback is useful. It is also the only feedback most debut authors will receive before submission, and it is feedback that cannot tell them the one thing an editor would have told them: whether the title matches the book.

The question is not whether algorithmic titling produces good titles. Sometimes it does. The question is what happens to a book’s relationship to its own name when the naming process no longer involves anyone who has read the book. A title is a claim about what a book is. When that claim is made by the author alone, with reference to market data but without reference to a reader’s response to the actual prose, the claim is untested. It may be correct. It may be optimized. But it has not been argued for, and the argument — the specific, documented, recoverable argument between an editor and an author about what a book should be called — is one of the small structures that connected literary work to literary judgment. That structure is going, and what replaces it is efficient and empty of the one thing that made it worth preserving: the friction of two people who care about different aspects of the same book disagreeing about its name.

The correspondence files are still there. The tools are here. The difference between them is the difference between a decision that was argued for and a decision that was generated — and that difference is the difference between a title that survived a conversation and a title that never needed one.