Answers · Cover feedback

How can authors test a book cover with real readers before publishing?

Short answer

Put two to four versions of the cover in front of readers of your genre and see which one they pick. CoverCrushing, an author tool in the Taciturn network, does this: upload 2 to 4 cover variants, genre-matched voters pick a winner, and you get a full analytics report.

Why test a cover at all

A cover has one job: tell a browsing reader, at thumbnail size, what kind of book this is and make them curious enough to click. Authors are usually the worst judges of that, because they already know the story and see details a stranger never notices. Readers of the genre react to type, colour, and mood in a second or two, and their first pick is the signal that matters.

Testing before launch is cheaper than a redesign after. Once a book has reviews and ads pointing at one cover, changing it means new images, new ad creative and sometimes a confused audience.

How to run a fair test

  • Show every variant at the same small size a store search page uses, not full screen.
  • Ask readers of your genre, not friends and family, who will try to be kind.
  • Change one big thing between versions, such as the image or the title treatment, so you know what made the difference.
  • Keep the title and author name identical across versions unless the type is what you are testing.
  • Wait for enough votes that one result is clear; a handful of responses can swing either way.

How CoverCrushing works

  • Upload 2 to 4 cover variants.
  • Genre-matched voters pick a winner, with live vote tracking.
  • Hybrid Panel scoring combines AI analysis with Covercrusher votes.
  • Get your full analytics report, then re-run or refund.

Price

New authors get one free A/B test. First Crush is $5 for your first paid test, then $12. Details at covercrushing.com.

What a thumbnail test is measuring

A store search page does not show the cover at the size you designed it. It shows a small rectangle beside a title and a price, and the reader decides in the time it takes to scroll. The test that matches that moment shows every variant at that small size, not as a full-bleed mockup on a laptop. KDP's cover image guidance is the retailer's own rule for the file you upload. This page does not restate pixel sizes from memory; the help topic is the source, and it changes when Amazon changes it. What does not change is the job of the thumbnail: tell a stranger the genre and give them a reason to stop.

Authors are a poor panel for their own jacket because they already know the plot. They see the symbol they insisted on. A stranger sees a dark blob or a title they cannot read. That is why the voters have to be people who buy the genre and who have not heard you explain the image. Friends and family try to be kind, and kindness is not a click. A handful of kind votes will swing either way. Wait until one variant is actually ahead, then stop. Running the test again because you preferred the loser is how the test becomes a rubber stamp.

One change at a time

If two variants differ in the photograph, the type, and the colour of the title, a win does not tell you which of the three did the work. Change the one thing you are willing to reprint. Keep the title words and the author name the same unless the type treatment is the question. Keep the crop the same unless the crop is the question. Write down, before you upload, the single sentence the cover is supposed to say ('this is a quiet historical novel,' 'this is a fae romance with a knife in it'). After the vote, ask whether the winner says that sentence at small size. A winner that says a different sentence is a different book, and you should notice that before you lock the file.

Testing before the listing is public is cheaper than testing after reviews and ads point at one image. A later swap means new creative and a stretch of readers who think the book changed. It can still be the right call when the first jacket was wrong. It is an expensive call if you never asked a stranger. CoverCrushing's report is the record of that ask: the variants, the votes, and the hybrid score the tool describes on its own site. The report is not a promise of sales. It is evidence about a picture.

After you have a winner

Take the winning file into the upload path the retailer documents, and check it at the size their preview uses, not only at full size on your screen. Then leave it alone through launch unless the preview shows a real defect (type cut off, a required element missing). A voice problem in chapter one is not a cover problem. Do not reopen the test to soothe a bad review of the prose. The tool for the prose is a different one, described on the voice page, and the choice between the two tools is on author tools. The reader sites in the directory are not part of this test.

Visit CoverCrushing