What you can do that a rewrite cannot
Three things, and they are the three that matter most.
Add information. The single most effective edit in flat prose is replacing an abstract noun with something specific — a number, a date, a name, an incident. That specificity has to come from outside the text, from what you know about the subject, and a tool operating on the words in front of it has no access to any of it. It can make an abstract sentence sound better. It cannot make it concrete.
Delete. You can decide a paragraph should not exist. A rewrite will faithfully improve a paragraph that was never needed, which is the most common way an automated pass produces a worse document — every section survives, better dressed.
Notice that the argument fails. Reading your own draft with fresh attention is how you find the claim you never supported and the objection you never answered. Those are properties of the whole piece, and nothing at sentence level can see them.
What “editing” actually consists of
Most of the argument about tools versus people is muddled by the word editing doing three jobs at once. They are separate passes, they run in a fixed order, and only one of them is a job a rewriting tool can attempt.
The structural pass asks whether the piece argues something, in a sequence a reader can follow, without a hole in the middle. Its characteristic action is deletion, and its characteristic output is a shorter document with a different shape. This is the pass that decides whether the piece is any good, and no tool operating on sentences can perform it, because the fault is never inside a sentence.
The line pass works at paragraph and sentence level: rhythm, transitions, the buried verb, the sentence that says the same thing as the one before it. This is the only pass where an automated rewrite is a genuine substitute for part of the work, and even here it is doing a subset — it can vary your rhythm, and it cannot notice that two paragraphs are making the same point.
The copy pass is spelling, grammar, consistency of style, whether the citation format holds. Software has been good at most of this for twenty years and it is not what anyone means by a humanizer.
Put that way, the honest claim for a rewriting tool is narrow: it automates part of one of the three passes. That is worth paying for when the other two are done. It is worth very little when they are not, which is the situation most drafts are actually in.
A paragraph, both ways
Take a paragraph that is grammatically fine, factually correct, and dead on the page — the exact case people reach for a tool to fix.
The implementation of the new policy resulted in a significant improvement in student outcomes. It is important to note that the effects were observed across multiple departments. Moreover, faculty members expressed positive sentiments regarding the changes that had been introduced.
What a rewrite does with it. It varies the sentence lengths, drops Moreover, unpicks the implementation of into a verb, and hands back something that reads considerably better. Three sentences in, three sentences out. What it cannot do is ask the question the paragraph is begging for, because the answer is not in the text.
What you do with it. You notice that significant improvement has no number attached, and that you know the number. You notice multiple departmentsmeans three, and that naming them is more convincing than counting them. You notice that faculty expressed positive sentiments is a sentence protecting itself, and that what actually happened is that two department heads asked to extend the pilot. Then you delete the third sentence entirely, because a fact about people’s feelings adds nothing next to a fact about what they did.
The rewritten version is better prose. Yours is a better paragraph, and the difference is entirely information the tool never had. That is the whole comparison in one example, and it is why the order in the next section is not a preference.
What the tool is genuinely better at
Hearing rhythm you have gone deaf to. After the fourth read of your own paragraph you cannot tell whether it flows, because you know what it is going to say. A rewrite gives you a second version to compare against, and the comparison is often more useful than the output — you accept two of its changes and reject the rest, and the two you accepted are ones you would not have found.
Volume under time pressure. Eight paragraphs that are correct and leaden, due tomorrow, is exactly the situation where an hour of manual work is not available and a rewrite is better than shipping them as they are.
Consistency across a long document. Sections written weeks apart drift in register, and a pass at one tone setting pulls them closer together than most people manage by hand.
The order that works
Structure first, by hand. Read the whole thing and fix what it argues, what it leaves out and what should be cut. Do this before any automated pass, because a rewrite makes weak prose read smoothly and smooth prose hides structural faults you needed to notice.
Then facts, also by hand. Check the numbers, open the sources, confirm the citations support what the sentences claim.
Then the rewrite, on the paragraphs that survived and are still stiff. Then a final read of the whole document in one sitting, for register — because a rewritten paragraph sitting between two of your own is exactly where a reader hears the join.
Where the hybrid goes wrong
Recommending a hybrid is easy. The two ways it fails in practice are worth knowing before you adopt it.
The join. A rewritten paragraph sitting between two of your own is where a reader hears the seam — not because the rewrite is bad, but because it is differently good. Its register is slightly more even, its transitions slightly more explicit. One such paragraph in a page is invisible. Every third paragraph rewritten and the piece develops a pulse, alternating between two voices. The fix is the final read-through in one sitting, and it is the step people skip because the document already looks finished.
Rewriting first. Running the tool before the structural pass is the single most common mistake, and it is expensive in a way that is hard to see. Smooth prose hides structural faults. A paragraph that should be deleted reads defensibly once its rhythm is fixed, and you will not delete it, because the thing that made it obviously disposable was how badly it read. You have spent the tool’s value making a document harder to edit.
How to hear your own draft again
Most people reach for a rewrite because they have gone deaf to their own prose, which is a real problem with several free solutions worth trying first.
Read it aloud. Not silently — aloud, or at least subvocalised. Your ear catches a metronome your eye reads straight past, and it catches breath: any sentence you cannot say in one go is a sentence a reader is also struggling with.
Change how it looks. A different font, a narrower column, or the draft on your phone breaks the familiarity that stops you seeing words you have read forty times.
Leave it overnight. The cheapest and least available option, and the one that works best. Most of what a second opinion gives you is distance, and time gives the same thing for free.
Read the first sentence of every paragraph, in sequence. If that is not the argument, you have a structural problem and no amount of line editing will rescue it.
If those fail — and after the fourth read of your own paragraph they often do — a second version to compare against is a legitimate reason to run a rewrite. Use it as a mirror rather than as an answer: accept the two changes that make you think oh, yes, and reject the rest.
When editing it yourself is the wrong call
This page argues for doing it by hand, so the cases where that advice fails belong on it rather than somewhere more convenient.
When you do not know the subject. Every advantage claimed above — adding a specific, deleting what is unnecessary, noticing the argument fails — depends on knowing more than the text does. Edit a piece on something you have just read about and you have none of that. You will polish sentences, which is the one part a tool does competently, and you will not catch the error that matters. The fix is not a tool; it is finding someone who knows the subject to read it.
When you are editing into a register you are not fluent in. This is the case nobody in our industry writes about, and it deserves care. The advice to break conventions on purpose — start with a conjunction, use a fragment, let one sentence run long — assumes enough command of the language to know which rule you are breaking and why. A writer working in a second language who applies it can end up with prose that reads as error rather than as style, and will be marked down by a human for it. There is a related unfairness in how detectors read that population, which we go through in why non-native writers are flagged more often. Careful, conventional prose is a legitimate register. It is not a fault to be edited out.
When the piece does not warrant it. An internal status update, a product description in a batch of two hundred, a routine reply. The hour is real and so is its opportunity cost, and spending it on something nobody will read closely is not craft, it is avoidance. Several of our use-case pages are about exactly this kind of volume work, where the honest recommendation is to spend your attention on the pieces that carry weight and let the rest be adequate.
The time and the money, honestly
A serious editing pass on a 2,000-word piece is closer to an hour than to ten minutes, and the gap between those two numbers is most of why this market exists. Split across the three passes, the structural one is usually the shortest and the line pass the longest.
Against that, our Writer plan is $18 a month for 25,000 words — see pricing for the rest. Whether that is good value depends on a calculation worth doing honestly: it buys back part of one of three passes. If your drafts are structurally sound and merely stiff, that is a real hour saved and the tool is cheap. If your drafts have the usual problems, you are paying to have the wrong pass automated, and the hour you needed to spend is still waiting for you.
There is also a compounding argument against the tool that we would rather state than leave for someone else. Editing is a skill that improves with use, and the improvement transfers to your first drafts — people who edit their own work for a year write better first drafts, which is a return no subscription pays. A tool used to skip the learning keeps you at the level you started.
Where this leaves the product
A rewriting tool is a late-stage instrument. It is worth money for the mechanical pass on prose that is already correct, and it is the wrong instrument for a draft with a structural problem, which is most first drafts.
That is a smaller claim than this category usually makes, and it survives contact with anyone who has actually edited something. The neighbouring comparison — rewriting versus paraphrasing — is about two tools that look similar and do different jobs, which is a different question from this one.
If what you are actually deciding is whether to edit at all, the honest version of the argument is on what a rewrite cannot fix. And if you want to see which sentences a detector is reacting to before you touch any of them, our detector reports the score and the signals behind it — the flat, evenly-weighted lines it highlights are usually worth rewriting on their own merits, whatever any detector thinks.