Cover Testing · 9 min · September 24, 2026

How to Choose a Book Cover with Reader Votes

A finished CoverCrushing test names a leading cover, a Crush Score, and a separate AI panel read. Here is how to turn that vote split into a publish, revise, or keep decision without treating the score as a sales forecast.

By The CoverCrush team · Editorial

Choosing a book cover with reader votes means reading the comparison you actually bought, then deciding. On CoverCrushing, the leading cover is the one with the most recorded votes in that test. The Crush Score describes how wide that lead was. The AI panel is a separate score. None of those outputs publish the book for you, and none of them predict sales. This page is the decision order the report supports.

Start only after the test is marked complete. A test stays open until the recorded vote count reaches that test’s target. You receive an email when it is marked complete. If you decide from a partial count, you are reading a test that has not finished. Wait for the target unless you are only checking that votes have started to arrive.

Read the vote split first

Open the results and find which cover has the most recorded votes. That cover leads. In an A/B test there are two images. In an A/B/C test there are three or four. The leader is not “the prettiest cover” and not “the cover the AI liked most.” It is the cover voters chose more often in this test.

Write the split down in plain counts before you interpret it. If one cover has a clear majority of the recorded votes, you have a leader. If the counts sit close together, you have a near tie. The Crush Score is the site’s way of expressing that margin as a 0–100 figure. A wide margin produces a higher score than a near tie. It is not a measure of time spent looking at a cover, and it is not a forecast of sales or rank.

Sample size is the vote target on the test you launched. The new-author free A/B test targets 200 votes on two covers. A paid test uses the target set for that test. A larger target can take longer than a smaller one. Do not compare your Crush Score with another author’s score from a different target, a different genre, or a different pair of covers. The score only describes the files inside your test.

Keep the AI panel in its own column

The AI panel scores every cover on a 12-criterion rubric from five perspectives: Genre Reader, Art Director, Retailer, New Reader, and Casual Browser. A Series Strategist replaces the Casual Browser when the book is marked as part of a series. Each perspective scores the criteria from 1 to 10, applies its own weights, and returns a composite, a love-or-pass vote, and a short comment. The panel composite combines those five scores.

Read that column after the vote split, not instead of it. The results view keeps AI panel output and reader votes separate on purpose. If readers pick Cover A and the panel composite prefers Cover B, you do not “average” them into a third cover. You have a disagreement between two labeled inputs. Common cases:

Readers and the panel pick the same cover. That is the simplest publish case, if you are willing to ship that file.
Readers pick one cover by a wide margin and the panel is mixed. The vote still names the leader. Use the panel notes as revision comments on typography, contrast, thumbnail legibility, or genre signal, not as a veto of the count.
The vote is a near tie and the panel prefers one file. Treat the vote as “either can lead” and use the panel to see whether one file is harder to read at a small size. You may still keep the cover you prefer.
You dislike the leader. You do not have to use it. Some authors ship the winner, some brief a designer with the split, and some keep the cover they preferred.

The CoverCrush Report writes this up in one place. AI panel text is labeled as AI. Reader votes are counted on their own. The report names a leading cover and a score based on that vote split. It is an aid while you decide, not a certificate.

Turn a clear lead into a decision

When one cover leads by a wide margin, decide among three moves.

Publish the leader if the file is actually finished: title readable, author name readable, and the genre you chose on the test matches the book you will list. The test compared those images only. It did not check your blurb, price, or categories.

Revise the leader if the vote lead is clear but the panel, or your own eye, shows a fix that does not throw away the direction readers picked. A contrast or type-size fix on the winning direction is a different job from starting over. If you change the image enough that it is a new cover, it is a new test, not the old result.

Set the leader aside if you have a reason you can name: a rights problem, a series look you must keep, or a file that does not match the book. The report does not override that constraint. Record why you set it aside so the next test does not repeat a file you cannot ship.

Turn a near tie into a decision

A near tie means the readers who saw these files did not hand you a strong leader. That usually means one of two things. Both covers are acceptable for this audience, so you can ship the one you prefer without pretending the data crowned it. Or the covers are too similar, so the test could not see a difference. Look at the images side by side at thumbnail size. If you have to squint to tell them apart, the next test needs a real difference: a different image, a different type hierarchy, or a different genre cue. Retesting the same two files will spend another test on the same split.

Do not invent a tie-break the product does not have. There is no hidden sales model under the Crush Score. Price, blurb, category, and the book itself still decide how a book sells. If you need another comparison, run another test with two covers that differ in a way you can describe. An additional A/B test is $12. An A/B/C test of three or four covers is $19. CrushPass ($19 per month) includes one A/B test credit per paid invoice, not unlimited tests. Pro Monthly ($29 per month) includes two A/B test credits plus unlimited single-cover feedback.

What not to do with the numbers

Do not send the report to a retailer. CoverCrushing does not send results to Amazon, KDP, or IngramSpark, and a test does not change rank or sales. Do not treat a single-cover feedback score as a head-to-head win. Single Cover Feedback is $7 and rates one cover. A preference between two covers only exists when both were in the same test.

Do not decide from your writing group and call it the same thing as this vote. The vote page is people who picked genres and chose a cover, with a rate limit per IP. Friends who know you are the author are a different room. You can still ask them. Just do not merge their opinions into the recorded vote count.

Do not wait for a perfect score. The product does not define a Crush Score that “guarantees” a launch. A wide lead is a clearer comparison than a tie. It is still a comparison of the covers you uploaded.

When you are ready to run the comparison, start a cover test. If you do not have two finished files yet, read A/B testing versus a designer or a contest first.

Frequently asked questions

Q: How do I choose a book cover from reader votes? A: Wait until the test is marked complete. The leading cover is the one with the most recorded votes. The Crush Score describes how wide that lead was, from 0 to 100. Read the AI panel separately. Then publish the leader, revise it, or keep another cover. The report is decision support, not a sales forecast.

Q: What if the vote is close? A: A close split is a near tie. You can ship the cover you prefer, or you can test again with covers that differ in a way readers can see. The Crush Score is lower when the margin is narrow. It does not hide a second ranking.

Q: Should I follow the AI panel if it disagrees with the votes? A: Keep them separate. Reader votes name the leader. The AI panel is a 12-criterion read from five perspectives, labeled as AI. Use panel notes as revision comments. Do not average the panel into the vote count.

Q: Do I have to publish the winning cover? A: No. You can ship the winner, brief a designer with the split, or keep the cover you preferred. Uploading the files does not transfer copyright, and the test does not change your Amazon rank.