The Manuscript That Sounds Like AI: Signs Your Book Has Lost Its Human Voice

A manuscript can be grammatically clean and still feel as if nobody in particular wrote it. The sentences are balanced, the chapters end on uplifting notes and nothing is obviously wrong, yet the reader finishes a page without having seen a single person, place or decision clearly. Authors and editors increasingly describe this as writing that "sounds like AI".

That label is not very useful on its own. Stock phrases such as "overcame adversity", "shaped the person I am today" and "a powerful reminder that we all have a story to tell" filled manuscripts long before chatbots existed, and nobody can reliably tell from a sentence like that whether a tool wrote it. The useful question is what the pages let a reader see, understand and believe. If every difficult event is smoothed into the same lesson, the manuscript has lost the particular mind that was supposed to tell it, whoever or whatever produced the words.

Sotrane Publishers' position on AI in book writing puts it simply: we would rather read the author's thinking than a perfect imitation of it. The checks below are ways to find where that thinking has gone missing and bring it back.

Look for the same lesson at the end of every section

Read the last sentence of five consecutive sections. If each declares that the author learned resilience, embraced change or discovered the strength of community, the repetition is flattening the events themselves. A book can contain those ideas, but they mean little when attached to every experience regardless of what happened.

Suppose a memoirist recalls being sent home from school because fees had not been paid. The paragraph could end, "This challenge taught me that perseverance is the key to success." That line skips the moment when the child has to explain the situation at home and the parent decides what can be paid first. If the author remembers waiting outside the classroom, or what happened on the walk back, those details may carry the meaning more honestly. If the memory is uncertain, say what can be recalled and leave out invented dialogue.

Underline every sentence that tells the reader what an experience "teaches us" and ask whether the scene before it has earned the lesson. Keep the insights that change how the reader understands the event, and cut or replace the rest with the action and thought that led to them.

This is a writing problem whether or not a tool was involved. AI can help organise a book, but it becomes especially easy to produce a chain of polished lessons when an author asks a tool to "make this inspiring" instead of supplying the events and judgements that only the author knows.

Find what the smooth sentences are covering

A chapter in a business book states that "successful entrepreneurs must adapt to changing market conditions through innovative strategies". Nobody would argue with it, and it does not help a trader decide whether to keep ordering a product that has stopped selling.

What did the author actually observe? Perhaps customers in a Lusaka market began asking for smaller packs while the trader's larger stock stayed on the shelf. What did the trader try, did it work and what did it cost? If the author has a real case and permission to use it, that material makes the point. Otherwise a clearly labelled illustrative example is better than a fabricated success story.

In a practical book, circle words such as "effective", "transformative" and "strategic" and ask what a reader could do differently after reading the paragraph. If the answer is nothing, the writing may be covering for a missing example, step, limit or decision. The repair is rarely a stronger adjective. It is the knowledge the author has not yet put on the page.

A chapter about starting university in Zambia might advise students to "practise responsible financial management". A more useful passage could show a hypothetical student dividing a fixed allowance between transport, meals, data and a small reserve, then asking what happens when transport costs rise. Any figures should be marked as illustrative, but the passage now discusses a real trade-off instead of decorating a general statement.

A book becomes forgettable when it tells the reader what to feel and leaves out what happened.

Listen for a voice that never changes

In some novels, a head teacher, a teenager and an older market trader all speak in the same balanced sentences. Each states a feeling politely, explains a reason and ends with a lesson. The dialogue is clear, but without the names attached it is impossible to tell who is speaking.

Cover the character names on a page of dialogue and see whether you can still tell the speakers apart by what they notice, avoid, misunderstand or want. Voice is more than slang. A child may ask a blunt question at the worst possible moment, an adult may answer indirectly because the truth would embarrass someone and a trader may calculate while talking. Those differences come from character and circumstance, not from a list of "authentic" expressions pasted into the text.

In non-fiction, compare the opening chapter with the author's own conversation or earlier writing. A lecturer who explains research methods through the questions students actually ask should not suddenly sound like a conference brochure. A pastor who uses plain stories in person need not turn every page into an abstract sermon. An editor's job includes making prose clearer without replacing the person who thinks behind it. When Sotrane edits a manuscript, changes are made in tracked changes so that the author can accept or reject each one, which keeps the final decision about voice with the author.

Do not make every line eccentric to prove that a human wrote it. The aim is a consistent voice that can become quiet, funny or formal when the material calls for it.

Make the Zambian setting do some work

Some manuscripts mention Lusaka, nshima and a chitenge in the first chapter and then carry on as though those words had supplied a setting. Readers still do not know what decisions people face, how they move through the place or why any detail matters to the scene.

If a character needs to send books from Lusaka to a reader in Chipata, the distance can affect when the order arrives and what the buyer pays. If the scene is a family meal, the conversation at the table may matter more than naming the food. If the writer describes a church gathering, the reader needs to see what the gathering changes for the people in the story, not "African community" added as atmosphere.

A Zambian reader should never be made to feel like a tourist in an ordinary Zambian scene, so there is no need to explain familiar things as if they belonged in a museum. Equally, a local word should not be added for decoration when its use is uncertain. Ask a fluent speaker where needed and pay attention to who would naturally say it.

One quick test is to remove every place name from a page. If nothing about the action, relationships or decisions changes, the setting is not yet doing any work. That does not mean filling every paragraph with geography. It means letting the real place shape the story where it matters.

Check where the draft sounds more certain than the author is

Generic prose often hides doubt behind confident claims. "Research conclusively proves that young people no longer read books" may appear without a source, a definition or a time period. "Every parent knows..." may present one person's opinion as universal experience. A memoir may state exactly what another person felt, when the author can only know what that person said or did.

Mark every claim that involves numbers, research, a person's motives or current rules, and check whether a reliable source supports the actual wording. If it does not, narrow the claim or remove it. A book does not become weaker by acknowledging its limits. Readers can trust a writer who says "several of the people interviewed for this book described..." when that is what the author can show. They cannot check a sweeping statement that floats above the evidence.

If a tool has suggested a quotation or source, trace it to the original before printing, because a plausible reference can be wrong. The same check applies to notes copied from an old presentation or a friend's message. This is an author's responsibility, not an exercise in catching machines.

Rewrite one paragraph from what only you know

Choose a paragraph that feels polished but forgettable. Without looking at it, answer a few questions in your own notes. What did you see, hear, decide or learn that led you to write it? What would a sceptical reader ask? Which fact can you check? Which details are yours to tell, and which belong to someone else?

Then write the paragraph again from those answers. In an illustrative first version, a trader writes: "My business taught me the importance of adaptability and determination in the face of challenges." The account underneath might be: "I ordered the same school exercise books in January that had sold in December. By March, the last carton was still under the counter. When a customer asked for smaller packs, I finally looked at what people were taking home rather than what I had hoped to sell." If those events actually happened, they give the reader a decision and its consequence. If they did not, do not borrow them because the second version sounds vivid.

Some paragraphs genuinely need to stay simple and general. An introduction can state a book's scope without a dramatic anecdote, and a transition can quietly move the reader on. The aim is to find the places where abstraction has replaced the material that gives the book a reason to exist.

It also helps to ask a reader who does not know you to mark three passages they still remember a day later and three they could not explain back to you. Their reasons will help far more than a score from an AI detector. A polished line is not evidence of tool use, and a dull line is not a verdict on the writer.

Repair the draft at the right level

If the same vague lesson closes every chapter, review what each chapter is for before polishing sentences. If examples are missing, return to interviews, records and lived experience. If local language has been flattened, read passages aloud with people who know the setting and can challenge a false note. If a claim has no evidence, research it or narrow it. Only after that is it worth correcting rhythm and grammar.

A proofreading pass cannot supply a scene that was never written, and a cover cannot make an empty promise substantial. This kind of revision takes longer than asking a tool to "humanise this text", but it gives the author back control of what the book actually says.

The clearest sign of a human voice is not imperfect spelling or forced slang. It is a recognisable person making choices on the page about what to notice, what to question, what they know and what they refuse to pretend to know. Which line in your own draft could be moved into another writer's book without anyone noticing, and what experience was it meant to describe?