Trust the Author, Not the Tool
Your report comes back. Forty flags. Maybe more. Your stomach does the thing stomachs do, and some quiet voice says great, so the book is broken in forty places.
Slow down. Read the flags again, and this time read them as questions. Every one is the tool tapping you on the shoulder and saying "hey, take a look at this." Not one of them is saying "change this or you've failed." The difference between those two readings is the difference between a report that helps you and a report that quietly talks you into sanding the life out of your own book.
So let's talk about who's actually in charge of your manuscript. (Spoiler: it's you. It was always you.)
A good manuscript tool points at places where a craft principle might be in play. It does not decide whether your choice was wrong. That call is yours, and the flags worth acting on are usually the ones that make you stop and think, not the ones you can explain away in a sentence.
"Trust the reader" has a cousin
If you've read much on craft, you know "trust the reader." Don't over-explain. Don't spell out the subtext. Don't tack an emotion onto a line that already carries it. Give the reader credit and let them meet you halfway. Browne and King built half of Self-Editing for Fiction Writers on that one idea.
There's a cousin to it that nobody says out loud: trust the AUTHOR.
You made choices. Some of them were instinct, some were deliberate, some you agonized over at two in the morning. A tool reading your manuscript doesn't know which is which. It sees a fragment sentence and flags it, not knowing you fragmented it on purpose for punch. It sees a repeated word and flags it, not knowing the repetition is doing rhythmic work. The tool is pattern-matching against a rule. YOU know why the line is the way it is.
So when a flag lands on something you did on purpose, and it works, the correct response is to leave it exactly as it is. That's not you ignoring feedback. That's you doing your job.
Should I change everything an AI editor flags?
No. And a tool that expects you to is training you badly.
Here's a sorting method that works. Every flag falls into one of three buckets:
- The obvious catch. A real typo. A genuine continuity error (her eyes were green on page 12 and brown on page 200). A timeline that doesn't add up. Fix these. No debate. This is the stuff you're genuinely glad it caught.
- The one that makes you pause. You read the flag, you look at your line, and something in you goes "hm... is that actually right?" THAT hesitation is gold. That's the flag earning its keep. Go investigate it properly.
- The one you can answer instantly. You read the flag, you know exactly why you did it, and you know it works. Keep your line. Move on. Don't argue with the tool, don't second-guess yourself into a worse sentence.
Most writers get in trouble by treating all three buckets like bucket one. They march down the list changing everything, and forty edits later the book reads flatter than it did before, because half those "fixes" sanded off things that were working.
Why a tool that makes the call for you is dangerous
A prescriptive tool, one that rewrites your line or tells you flatly "this is wrong," has a bias baked in. It's pulling your prose toward the average. Toward the safe, smooth, unobjectionable version of every sentence.
The problem is that the average is where books go to be forgettable. Your voice lives in the choices that a rule would flag. The odd rhythm. The deliberate fragment. The scene that runs long because it needs to breathe. Strip all of those out to satisfy a checker and you get clean, competent, invisible prose that reads like everyone else's. Nobody ever fell in love with a book because it had no flags.
This is exactly why FirstReader doesn't rewrite your words (I wrote about that whole stance in An AI Fiction Editor That Doesn't Rewrite Your Book). It names the principle, shows you the excerpt, and stops. What happens next is your call, because it's your book.
How to actually use the report
So the method, start to finish:
Read the whole report once without touching anything. All of it. Get the shape of what it's seeing before you fix a single comma.
Then go through and sort. Obvious catches get fixed. Pauses get investigated. Instant-answers get left alone with a clear conscience. And when a pattern shows up over and over (say the tool keeps flagging the same thing across chapters), pay attention, because a repeated flag is more likely to be real than a one-off.
That's it. You end up acting on the flags that deserve it and confidently skipping the ones that don't, instead of grinding through all forty like a to-do list.
Where this cuts the other way
Let me be honest, because "trust the author" can turn into a very comfortable excuse.
If you're waving off ninety percent of every report, stop and ask yourself a hard question: is that craft judgment, or is that ego? Because they feel IDENTICAL from the inside. The flags that annoy you most are often the ones sitting closest to something you don't want to look at. "Trust yourself" is a tool for keeping the choices you can defend, not a shield for the ones you just don't feel like fixing.
And the newer you are, the less earned that confidence is. A writer on their first novel hasn't built the instinct yet to know a working fragment from a lazy one. If that's you, weight the flags a little heavier and your gut a little lighter, at least until you've got some reps in. (I say this as someone who has confidently defended plenty of lines that were, in fact, just bad. It's humbling. It's also how you learn.)
The tool is a second set of eyes, not a boss and not a rubber stamp. Use its flags to think harder. Then make the call yourself, every time.
If you want a read that hands you the flag and the reasoning and then gets out of your way, run a free chapter through FirstReader. You decide what to do with every single note.
Common questions
Should I change everything an AI editor flags?
No. Fix the obvious catches (typos, real continuity errors), investigate the flags that make you genuinely pause, and keep the choices you made on purpose that work. A flag marks a spot to look, not a mistake to correct on sight.
What does "trust the author" mean in feedback?
It means the writer, not the tool, decides whether a flagged choice stays or changes. A tool points at places a craft principle might apply. It can't know what you did on purpose, so the final call on your voice and your intent belongs to you.
How do I know which feedback to ignore?
Keep a flagged choice if you can immediately explain why you made it and it works on the page. Investigate anything that makes you hesitate. Be suspicious of your urge to dismiss a flag that irritates you, since irritation often marks a real weak spot.
Isn't ignoring feedback just ego?
It can be. Craft judgment and ego feel the same from the inside. The honest test: can you defend the choice on its merits, to someone other than yourself? If yes, keep it. If you're only defending it because changing it sounds like work, that's ego.
If this was useful, drop a comment and let me know. If it wasn't, I want to hear that too. And if you're not on the list yet, subscribe so you don't miss the next one.