Beta Readers: How to Find Them, and How to Read Their Feedback

The single most useful sentence about feedback comes from Neil Gaiman:

When people tell you something's wrong or doesn't work for them, they are almost always right. When they tell you exactly what they think is wrong and how to fix it, they are almost always wrong.

Everything below is an elaboration of that.

What a beta reader is for

Not to fix your book. Not to praise it. To be a sensor: a person who reads it once, the way a stranger would, and reports what happened to them while they did.

Where they were bored. Where they were confused. Where they stopped believing. Where they put it down and didn't pick it up for four days. That data is irreplaceable, because you can no longer experience your own novel — you know what you meant, and you supply it silently on every page.

Choosing them

Not your partner. Not your mother. Not because they'll lie, but because they will read you, not the book. They cannot help it, and neither could you.

Not only other writers. Writers are excellent at diagnosis and terrible at being readers — they will tell you about your comma splices and forget to mention that they didn't care about Anna. You want a mix: two or three people who read your genre for pleasure, and one or two who write.

People who finish books in your genre. A literary-fiction reader will tell you your thriller's characters are thin. They're right, and it doesn't matter; the book isn't for them.

Three to five. Fewer and one idiosyncratic reaction dominates. More and you drown, and you'll wait six months for the slowest.

When to send it

After you have fixed everything you already know is wrong.

Sending a draft you know is broken wastes the one thing a beta reader has: their first read. You get exactly one, and they'll spend it on a problem you could have found yourself. Do your structural pass, do your character pass, then send.

Send them the whole book. Chapter-by-chapter feedback produces readers who are critiquing rather than reading, and it's the reading you need.

What to ask

Do not ask "did you like it?" You will get "yes", and you'll have learned nothing.

Ask questions that report experience, not judgment:

• Where did you stop reading? Not "was it boring" — where did you physically put it down. That location is worth more than any opinion in the document.
• Where were you confused? Mark it, don't diagnose it.
• Who did you want more of? Less of?
• At what point did you know how it would end? If that's page 60, you have a problem, and it isn't a prose problem.
• What did you think the book was about? Compare to your answer. The gap is the most useful thing you'll learn all year.
• Was there anything you didn't believe? Not implausible — unbelieved. Different thing.
• What do you remember, a week later? Ask this a week later. Whatever they don't remember, you did not put on the page.

Notice that none of these ask for solutions.

Reading the feedback

Here's the rule, and it's Gaiman's:

Treat every report of a feeling as a fact. Treat every prescription as a hypothesis.

If four readers say the middle drags, the middle drags. That is now settled, and arguing with it is a way of not fixing your novel. But if four readers say "you should add a subplot for Tomas", they are guessing — they've noticed a problem and reached for the nearest tool. The real cause might be that nothing narrows in act two, and a Tomas subplot would make it worse.

So: collect the symptoms, then diagnose them yourself. You're the only person who can, because you're the only one who knows what the book is trying to be.

The counting rule

Where readers disagree, count.

• All five say it: it's true. Fix it.
• Three or four say it: it's probably true. Look hard.
• Two say it: possibly true, possibly a shared taste. Consider.
• One says it: information about that reader. Which can still be valuable — if the one is your ideal reader, listen closely.

The exception that overrides the counting: if a single reader says something that makes your stomach drop because you already knew, that note is true and you have been avoiding it for months. You'll recognise the feeling.

What to do about the notes you hate

Wait a week before touching anything. The first response to criticism is physiological, not editorial, and it is always wrong.

Then reread the notes cold. Most of what felt like an attack will read as reasonable, and some of what felt reasonable will read as wrong. Anything that still stings after a week is worth a very long look, because the sting is usually recognition.

Where a machine fits

An AI can do the reading a beta reader cannot: it can hold all ninety thousand words at once and tell you where a thread went quiet, where the antagonist vanished for eleven chapters, where chapter twenty-two contradicts chapter three. That's tireless, cheap, and genuinely useful.

What it cannot do is be bored. It will not put your book down. It has no taste and no stake, and it will agree with almost anyone about almost anything. Never use a model as a reader. The one thing you need — did this hold a human's attention? — is the one thing it cannot report.

Use both. Narrale's editorial lenses and continuity tracking do the mechanical read across the whole manuscript, so your five humans can spend their single irreplaceable first read on the only question that matters: did they want to turn the page.

Related: Self-editing a novel · Using AI to write a novel