Cover Testing · 9 min · September 24, 2026

Book Cover A/B Testing vs Hiring a Designer or Running a Cover Contest

A designer or a cover contest produces files. A reader A/B test records which of those files voters prefer. Here is how the jobs differ, when to commission art, and when to put the finalists in front of readers.

By The CoverCrush team · Editorial

Hiring a designer, running a cover contest, and A/B testing a book cover are three different jobs. A designer makes a cover. A contest asks several makers to submit concepts so you can pick among their files. A reader A/B test does neither. It takes images you already have and records which one voters choose. CoverCrushing is the testing site. It is not a design studio, and it does not run design contests.

Authors get into trouble when they treat those jobs as substitutes. A beautiful file from a skilled designer can still be the wrong signal for the genre. A contest can produce a stack of options and still leave you guessing which one a reader would click. A vote test cannot invent a cover you have not uploaded. The practical order is to get real cover files first, then test the ones you are willing to publish.

What you hire a designer to do

A cover designer turns a brief into a file: typography, image, color, and a composition that can be read at thumbnail size. The site’s cover cost calculator treats a designer hire as a way to produce that file, separate from a vote. It groups mid-range marketplace routes such as 99designs and Reedsy together, and it groups premium custom work separately. Those calculator ranges are estimates for making a cover. They are not CoverCrushing test prices.

You still need a brief. Genre, comp titles, whether the book is a series, and whether the title must stay readable at search-result size all belong in that brief. The designer’s job ends when you have files you can actually publish. Their taste is informed, and it is still one person’s taste. It is not a count of how readers voted.

CoverCrushing does not replace that hire. The upload step expects a cover that already looks like a cover. Title and author name should be readable. A blank title area changes how people vote. Testing finished or near-finished options produces a clearer comparison than testing rough sketches. Many authors commission two or three near-final directions, then test those files instead of asking the designer to guess.

What a cover contest usually produces

A cover contest is a way to buy several concepts at once. You post a brief. Designers submit entries. You pick a winner, sometimes after revisions. The site’s own design notes describe 99designs as one place authors run that kind of contest, with quality that varies and with you acting as the creative director. A contest is useful when you do not yet have a designer and you want to see more than one visual idea before you spend on a final.

A contest vote, when the marketplace offers one, is not the same thing as a CoverCrushing reader vote. Contest feedback often comes from other designers, from friends you invite, or from people who are judging craft. CoverCrushing voters pick the genres they read, then choose a cover on the public vote page. Those are different audiences. A design that wins a contest because the illustration is skilled can still lose a reader test because the genre signal is muddy at thumbnail size.

Use a contest to get candidate files. Use a reader test on the two, three, or four finalists you would actually ship. Do not upload every sketch from a long contest. An A/B test compares two covers. An A/B/C test compares three or four. More options than that are outside those products.

What a reader A/B test measures

On CoverCrushing, an A/B test compares two images for the same book. You sign in, upload Cover A and Cover B, name the book, and choose a genre. Readers who pick A/B Compare see both covers and choose the one they prefer. The cover with more recorded votes leads. The same test also gets a five-perspective AI panel score, and that panel output is labeled separately from reader votes.

The leading cover is a comparison of the files in that test. It is not a certificate, and it is not a promise of sales. The Crush Score is a 0–100 figure based on the leading cover’s share of the votes. A wide margin scores higher than a near tie. It does not measure how long someone looked, and it does not forecast rank.

An A/B test of two covers is $12. An A/B/C test of three or four covers is $19. Single Cover Feedback, for one cover only, is $7. New authors get one free A/B test during the 48 hours after signup. That free test is two covers with a target of 200 votes, and it does not ask for a card while the trial is still open and unused. Prices are in US dollars and are charged through Stripe. The full list is on the pricing page.

CrushPass is $19 per month and includes one A/B test credit per paid invoice, plus 20% off additional tests. It is not an unlimited-test plan. Pro Monthly is $29 per month and includes two A/B test credits plus unlimited single-cover feedback. A membership is a way to repeat tests across a series. It is not a substitute for paying a designer to draw the covers.

When to design first, then test

Design first when you do not yet have two publishable files. A reader cannot vote on a mood board that is still in your head. Commission the designer, or run the contest, until you have two to four covers with the title and author name on them. Then start a cover test.

Test before you pay for a full paperback wrap, a hardback jacket, and a set of ads that all use one image. Changing direction after the ebook front cover is chosen is cheaper than changing it after the printer has the wrap. The test still will not tell the designer how to illustrate. It will tell you which of the directions readers picked, and the AI panel will tell you, separately, how the rubric scored typography, thumbnail legibility, genre signal, and the other criteria.

Bring the report back to the designer as a brief, not as an insult. If readers split evenly, the covers may be too similar, or both may be acceptable. If one cover leads by a wide margin and the panel also flags a weak thumbnail on the loser, that is a concrete revision note: hierarchy, contrast, or genre cue. The panel comment is labeled as AI. The vote count is the reader result. Do not average them into a single “truth” and do not ask the designer to chase a score the test did not produce.

What neither path owes you

A designer does not owe you a bestseller. A contest does not owe you a reader’s click. A CoverCrushing test does not owe you a sales lift, and it does not send the winner to Amazon. The test does not change your KDP dashboard. You keep the copyright in the files you upload. You can ship the leader, revise, or keep the cover you preferred. The report is decision support. The cover you publish is still your choice.

If the test does not finish within 72 hours because of a platform failure, you can get a full refund or a free re-run. A finished test is not refunded because you disliked the winner.

Frequently asked questions

Q: Should I hire a cover designer or run a CoverCrush A/B test? A: Hire a designer when you need a cover file made. Run a CoverCrushing A/B test when you already have two images and want reader votes on them. The test does not draw the cover. An A/B test is $12. An A/B/C test of three or four covers is $19.

Q: Is a cover contest the same as reader A/B testing? A: No. A cover contest asks designers to submit concepts. A CoverCrushing A/B test shows images you upload to readers who pick a genre and choose a cover. Contest feedback and reader votes are different audiences. Use a contest to get files, then test the finalists you would publish.

Q: Can I test rough sketches? A: You can upload a cover before the book is finished, but a missing title or a placeholder author name changes the votes. Testing finished or near-finished options produces a clearer comparison than testing rough sketches.

Q: Does the winning test cover replace my designer’s judgment? A: No. The winner is the cover with the most recorded votes in that test, next to a separate AI panel score. You can ship it, use the split to brief a revision, or keep another cover. The report does not transfer the design decision off your desk.