Proofreading: What It Catches, Why It Happens Last, and What It Costs

What is proofreading?

Proofreading is the final quality check, carried out after a book has been designed and laid out and before anything is printed or uploaded. Two documents sit side by side while it happens: the finished pages, and the manuscript everyone signed off. The job is to find what slipped through earlier and what the design process has since broken.

The defining feature is the object being read. Every earlier stage works on your manuscript, a file of text. A proofread works on pages: designed, numbered, with running heads and margins and line breaks. Those pages didn’t exist when your copy editor was working on your book, and they contain a whole class of errors nobody upstream could have found.

It is also, and this matters, not another edit. A proofreader is not scoped to improve a sentence they merely dislike, and on a traditionally published book they would not be permitted to. At this stage, changes cost money and can reflow the text onto a new page. The discipline is to fix what is wrong and leave alone what is merely different from how you’d have done it.

What does a proofreader catch that a copy editor can’t?

Everything that only exists once the book has been designed, plus whatever slipped through earlier. The first category is the reason the stage exists, and none of it is findable in a Word file.

• Bad breaks. A word hyphenated in the wrong place at a line end. Software hyphenation is confidently wrong several times in an average book.
• Widows and orphans. A lone short line of a paragraph stranded at the top or bottom of a page, or a heading sitting by itself at the foot.
• Running heads and folios. Still naming chapter four three pages into chapter five. Page numbers missing, repeated, or on the wrong side.
• Lost formatting. Emphasis that vanished somewhere in the conversion. A blank line marking a scene division, silently closed up, so two separate moments now run together as one continuous passage.
• Cross-references. “See page 000” is real, it happens in nonfiction constantly, and it reaches print more often than anyone likes to admit.
• Spacing and alignment. A paragraph that lost its indent, a stack of hyphens down a right margin, uneven word spacing in justified text.
• Art and captions. A caption attached to the wrong image, or an image placed before the paragraph that introduces it.

Then there is the second category: the errors that were always in the text and that everyone has now read past a dozen times. That’s not incompetence upstream, it’s how reading works. Familiarity makes text invisible, which is why you should put some time between your copy edit and your proofread, to come at it with fresh eyes, or have the very last read be done by someone who has never seen the book.

Three documents make this stage work, and the third is the one people forget: the laid-out pages, the approved manuscript, and the style sheet your copy editor compiled. Hand over the first two only and you have hired someone to impose their own conventions on a book that already chose its own. They will do it conscientiously, several hundred times, and every one of those changes will need reversing.

What is a real-word error, and why does spellcheck never catch it?

A real-word error is a correctly spelled word in the wrong place. Spellcheck cannot flag it because nothing is misspelled, and grammar software catches only a fraction because most of them are grammatical. They are the single most common thing a proofreader finds in an otherwise clean book.

Here is a passage with five of them. Every word in it is a real English word, and a spellchecker returns nothing:

The manger had been clear about the trail date. She read the letter form the board twice, looking for a casual link between the complaint and the suspension, and found none. By the time the untied delegation arrived, she had already decided.

The intended words were manager, trial, from, causal, and united. Read it quickly and most people catch one or two. Read it aloud and you catch more.

What makes these dangerous is that they are invisible in exactly the situations where they matter. Your brain corrects them as you read, and it corrects them most reliably in your own writing because you know what you meant. A stranger reading slowly, on paper or on a device, at a pace that feels uncomfortably deliberate, is the only reliable detector anyone has found.

Two related classes a proofreader watches for. Homophone slips, where their, there, and they’re or its and it’s go wrong, which grammar software does usually catch. And transposed proper nouns, where a character or place name from an earlier draft survives into a scene it no longer belongs in. Software has no view about that at all because it doesn’t know your book.

How much does proofreading cost?

Note that the following are the industry’s average proofreading rates. Senior editors with in-house publishing backgrounds and, better still, verifiable multiple bestseller and award-winner credits should charge above these bands.

In 2026, proofreading typically runs around 1.2 to 2.0 cents a word for fiction and 2.0 to 3.0 cents for nonfiction. On an 80,000-word novel, that’s roughly $960 to $1,600, and on the same length of nonfiction, about $1,600 to $2,400.

Those rates come from the Editorial Freelancers Association 2026 rate chart, the medians reported by more than 1,100 working editors for what they charged during 2025, at the standard 250 words to a page.

Table 1. Proofreading rates by genre, 2026 Editorial Freelancers Association medians.

A note on the wider figures you’ll meet. Search for proofreading rates and you’ll be quoted 1 to 5 cents a word, and the summary at the top of Google’s results builds its table from marketplace and gig-platform averages rather than from what editors report charging. The narrower survey band is the more useful planning number for a book. As a cross-check, one large marketplace publishes its own observed average at 2.0 cents a word, which sits exactly where the survey puts the top of fiction and the bottom of nonfiction.

At the very cheap end, ask what is actually being delivered. A proofread of a laid-out book at a few hundred dollars is a spellcheck pass with a person’s name attached, and it will not catch a single one of the layout errors that are the reason for the stage.

Why is proofreading the cheapest stage, and what does that tell you?

Because it’s the fastest, and it’s the fastest because the decisions have already been made. A proofreader reads 10 to 15 or more pages an hour on fiction against a copy editor’s 6 to 10 pages an hour and a line editor’s 5 to 9.5 pages an hour. Nobody is weighing whether a sentence could be better; they’re checking whether it’s right.

That has a practical consequence that people miss. The low price is not a signal that the stage is optional or that anyone can do it. It’s a signal that the work is narrow, and narrow work is exactly where a specialist beats a generalist. A good proofreader is not a cheap editor. They are someone who has trained themselves out of reading for meaning, which is a genuinely unnatural skill and the reason most writers cannot proofread their own books at any price.

It also tells you where the bargain is. Proofreading is the least expensive stage and it protects every dollar you spent on the ones before it. Spending $4,000 on structural and line work and then skipping a $1,000 proofread is the worst allocation available because errors on page 1 are what a reader uses to decide how seriously to take everything after it.

Can software or AI replace a proofreader?

Not yet, and the reason is specific rather than sentimental: the errors that matter most at this stage are either invisible to software or invisible to a text-only tool. Software is genuinely useful here, more useful than at any other stage, but it still doesn’t close the job.

What software does well, and you should absolutely use it: misspellings, doubled words, inconsistent hyphenation, straight quotes left among curly ones, spacing anomalies, and most homophone slips. Running a check before you hand over pages is free and it lowers your bill.

What it doesn’t do:

• It cannot see the page. Bad breaks, widows and orphans, incorrect running heads, and misplaced captions are properties of the layout. A tool-reading text has no access to any of it.
• It does not know your book. It has no view on whether the character in this scene should be Ruth or Rachel because both are valid names and only your manuscript says which.
• It has no style sheet. It will flag hundreds of deliberate decisions as errors and stay silent on the one place where you departed from your own convention.
• It is confident about real-word errors it has no way to resolve. Given the passage above, current tools catch some and invent objections to correct sentences, and you cannot tell which is which without already knowing the answer.

The honest summary: software has made proofreaders faster but hasn’t made them unnecessary. Use it to clear the countable errors, then have a person read the pages. If your budget genuinely allows for only one of the two, the person is the one who catches what embarrasses you.

When should a proofread happen, and on what file?

After typesetting, on the laid-out pages, and after every other change is finished. If anything substantive is still moving, you’re too early, and you will pay for a second pass.

The sequence that works:

• Finish the copy edit and your review of it. Every tracked change accepted or rejected, every query answered.
• Typeset the book.
• Proofread the pages, with the edited manuscript and style sheet at hand.
• Make the corrections.
• Check the corrections. This is a real step, not a formality: fixing one line reflows the ones around it.

The most common expensive mistake is proofreading the manuscript instead of the typeset pages, usually because it feels more convenient. You end up paying for a read that cannot catch the layout errors, then discovering them after publication, then paying again.

The second most common mistake is treating the proofread as a chance to revise. Every author feels the urge, seeing their words set as a real book for the first time. Resist it. Substantive changes at the proofreading stage are expensive, they reflow pages, and they introduce fresh errors at exactly the moment nobody is scheduled to look for them.

Do you still need a proofread if you’re publishing an ebook only?

Yes, and the reason is different from print. An ebook has no fixed pages, so the classic page-break errors don’t apply, but reflowable formats introduce their own faults and they only appear on a device.

What to check on an ebook proof:

• Chapter breaks that don’t start a new screen, or that start one in the wrong place.
• A table of contents whose links go nowhere or that skip a chapter.
• Italics and small caps lost in conversion, which is extremely common.
• Scene breaks that vanish because a blank line is not a reliable instruction in a reflowable file.
• Tables and images that are unreadable at mobile width.
• Front matter and copyright page details, which nobody ever rereads and which are frequently wrong.

Check it on an actual device, at more than one text size. A file that looks correct in a desktop preview can break at the largest accessibility setting, which is precisely the setting a substantial part of your readership uses.

If you’re publishing in both formats, they need separate checks. They are different files with different failure modes, and passing one tells you very little about the other.

How do you get the most out of a proofread?

Give the proofreader what they need, then leave them alone to be slow. The stage is cheap enough that the main way people waste money on it is by making it harder than it needs to be.

• Send the style sheet. If your copy editor produced one, and they should have, it goes to the proofreader. Without it, you’re paying someone to guess at (and possibly undo) decisions you already made.
• Send the edited manuscript too. The proofread is a comparison, not just a read. It answers “does the page match what was approved,” which requires both.
• Freeze the text first. Every late change is a new opportunity for an error and a fresh set of reflows.
• Tell them what’s deliberate. Invented spellings, dialect, a character who is illiterate in a specific and consistent way. Flag it or watch it get corrected.
• Don’t ask for a rush. This is the one stage where speed directly destroys the product. The whole value is a slow, deliberate read, and a proofreader who agrees to do your book overnight is telling you what kind of read you’ll get.

Do a pass of your own first, always. My ten tips for catching your own errors covers the technique. It replaces nothing, but it lowers the bill and it teaches you which mistakes are specifically yours.

Frequently asked questions

What’s the difference between proofreading and copy editing?

A copy edit changes your manuscript, working on grammar, usage, and consistency, and it produces a style sheet. A proofread checks the typeset pages against that edited manuscript and fixes only what is wrong. They look at different objects at different stages, and one is not a lighter version of the other.

How long does a proofread take?

Keeping in mind that each editor has unique demands on their schedule (and that schedules slide forward), a 320-page book should take roughly 21 to 32 working hours at a fiction pace, spread across one to two weeks of calendar time. Nonfiction is slower, 7 to 12 pages an hour, because of notes, tables, and cross-references.

Can I proofread my own book?

Yes, but along with your proofreader. Don’t rely on yourself as the sole set of final eyes. Familiarity makes your own errors invisible, and that effect gets stronger the more times you read the text, not weaker.

Is one proofreading pass enough?

For most self-published books, one careful pass on the final pages plus a check of the corrections is enough. Trade books usually get two passes of proofs, sometimes three on illustrated or heavily designed titles, and the second pass exists mainly to catch errors introduced by fixing the first set.

What if the proofreader finds a real problem, not a typo?

They flag it rather than fix it, and you decide. Occasionally, a proofread surfaces something substantive: a continuity break nobody caught, a sentence that no longer makes sense after an earlier cut. Fixing it is your call and it needs to be weighed against reflowing the pages, which is why the flag comes to you rather than being silently resolved.

My book is already published and has typos. What now?

Fix them and re-upload, if it’s print on demand or an ebook. That’s one of the genuine advantages you have over traditional publishing, where a correction waits for a reprint. Get the pages proofread properly, upload the corrected file, and the next copy anyone buys is clean.