How to Check Your Figures for Image Integrity Before You Submit
Nearly 4% of biomedical papers carry a problematic figure, and journals now screen for it at submission. Most of the flags are honest mistakes made while assembling figures, which is exactly why you want to find them first.
Readiness scan
Find out if this manuscript is ready to submit.
Run the Free Readiness Scan before you submit. Catch the issues editors reject on first read.
How to use this page well
These pages work best when they behave like tools, not essays. Use the quick structure first, then apply it to the exact journal and manuscript situation.
Question | What to do |
|---|---|
Use this page for | Getting the structure, tone, and decision logic right before you send anything out. |
Most important move | Make the reviewer-facing or editor-facing ask obvious early rather than burying it in prose. |
Common mistake | Turning a practical page into a long explanation instead of a working template or checklist. |
Next step | Use the page as a tool, then adjust it to the exact manuscript and journal situation. |
Quick answer: How to check your figures for image integrity: compare every panel for duplicated regions, inspect blots and gels for splicing without boundary lines, and confirm you have not over-processed brightness, contrast, or gamma in a way that erases or invents bands. Journals now run automated screens at submission, so the goal is to catch the honest mistakes yourself first. A screen of 20,621 papers found 3.8% carried a problematic figure, and most of those were not fraud. They were errors made while assembling figures under deadline.
That is the uncomfortable part. The most common image problem is not misconduct. It is a tired researcher duplicating the wrong panel at midnight.
Why this is worth an hour of your time
The mechanism is simple, and the stakes come straight from it. Image screening moved from a rare manual spot-check to routine automation. Journals now run tools like Proofig and ImageTwin on submitted figures, and those tools compare every region of every panel against every other, catching duplications and manipulations a human eye would miss. The screen does not read intent. It reads pixels.
So a duplicated control panel you reused by accident produces the same flag as a deliberately cloned one. The tool cannot tell them apart, and neither can the editor at first glance. What follows is an integrity query, and an integrity query is expensive even when you are completely innocent, because you now have to reconstruct what you did months ago and prove it was honest.
The Bik screen makes the scale concrete. Across 20,621 biomedical papers, 3.8% contained problematic figures, and roughly half of those were suggestive of deliberate manipulation. Read that the other way: about half were consistent with honest error, and every one of them still had to be explained. That is the population you do not want to join, and the entry fee is a check you can run before you submit.
What actually triggers an image flag
Not every adjustment is a problem, and treating them all the same paralyzes people. So here is where the line actually falls.
Issue | What it looks like | Verdict |
|---|---|---|
Whole-image brightness or contrast | Linear adjustment applied to the entire panel, disclosed | Allowed |
Selective or non-linear adjustment | Gamma or local edits that strengthen or erase bands | Not allowed |
Splicing without a line | Lanes from different gels joined with no visible boundary | Not allowed |
Inadvertent duplication | The same panel reused for two conditions by mistake | Honest error, still flagged |
Band or feature cloning | A region copied to fabricate or hide a result | Misconduct |
Adapted from journal image-integrity policies and COPE guidance (accessed August 2026).
The line that matters: you may adjust the whole image linearly and disclose it, and you may not do anything that changes what a specific band or feature appears to show. Everything hard sits between those two, which is why a second set of eyes helps.
The three common patterns behind an image flag
Here is what the inside of the review looks like. In our pre-submission review work with manuscripts, the figures that would trip a screen almost never come from misconduct. They come from three honest habits, and you can look for each of them in your own figures before an editor does.
The first is the assembly duplication. You are building a multi-panel figure, you copy a panel to use as a template, and you forget to replace the image inside it. What we see is that this is the single most common image flag, and it is entirely innocent, but the screen treats an identical control panel across two conditions as a duplication regardless of how it got there.
The second is the over-processed blot. You brighten a faint band so reviewers can see it, and in doing so you cross from linear adjustment into selectively changing what the blot shows. The intention was clarity. The result is a manipulation flag. The fix is to adjust the whole image, disclose it, and keep the raw file.
The third is the recycled control. You ran the same loading control for several experiments and used one image to represent all of them. That may be defensible if the experiments genuinely shared the control, but undisclosed it looks like duplication, and in practice it means you should show each control or state plainly that one representative image is shown.
None of these three is fraud. All three produce the same flag as fraud, which is why finding them yourself is the whole game.
How to actually check, from eye to tool
There are three levels, trading coverage for effort. Start with the first and escalate if your paper is figure-heavy.
Method | What it catches | Effort | Cost |
|---|---|---|---|
Careful visual comparison of every panel | Obvious duplications and splices you can see | Medium | Free |
Free image-forensics viewers | Contrast and clone artifacts a raw view hides | Medium | Free, partial |
Journal-grade screening (Proofig, ImageTwin) | The same automated check the journal will run | Low | Institutional or paid |
A pre-submission review that flags figure-integrity risks | The honest patterns above, and which check to run | Low | Free to paid |
The manual check works for a two-panel paper and falls apart for a figure with sixteen. This is the same access gap as plagiarism screening: the pixel-level forensic tools journals use, Proofig and ImageTwin, are sold to institutions, not to individual authors, so most people cannot run the exact automated check that will decide their paper. A pre-submission review is not a substitute for that forensic scan and should not claim to be one. What our pre-submission review does is read the manuscript for the three honest patterns above, the accidental duplication, the over-processed blot, the undisclosed recycled control, and point you to the raw files and disclosures worth fixing before the journal runs the scan you cannot.
Readiness check
Run the scan to see how your manuscript scores on these criteria.
See score, top issues, and what to fix before you submit.
When to worry, and when it is fine
Fix it before you submit if: any panel is identical or near-identical to another; a blot or gel shows a lane boundary that is not marked as a splice; a band looks unnaturally sharp against its background; or you adjusted part of an image rather than the whole. Keep the raw, unprocessed originals for every figure, always.
It is probably fine if: your adjustments were linear, applied to the whole image, and disclosed; your duplications are genuinely the same shared control and you say so; and you can produce the original files on request.
The one habit that protects you more than any tool is keeping the raw files. An integrity query you can answer with the untouched original is over in a day. The same query without the original can take months.
What we see across manuscripts
Avoiding an image query has nothing to do with having the prettiest figures. It comes from checking every panel for duplication and keeping the raw files. Use this guide when you are assembling a figure-heavy manuscript and want to catch the honest mistakes before an editor does. The figure-integrity policies and the Bik screen we reviewed are cited in full below. And if a reference in that same paper worries you, the companion guide on checking your references for retractions covers the other half of pre-submission integrity, alongside the journal directory for choosing where the cleaned-up paper goes.
Your figure-integrity checklist
Run this pass the day before you submit a figure-heavy paper:
- Open every multi-panel figure and compare each panel against the others for any region that repeats.
- Check every blot and gel for a lane boundary that is not marked as a deliberate splice.
- Confirm each brightness or contrast change was applied to the whole image and is disclosed in the methods.
- Make sure any shared loading control is either shown per experiment or labelled clearly as one representative image.
- Locate the raw, unprocessed file for every figure and keep it somewhere you can produce it within a day.
- Look for any band that appears unnaturally sharp against its background, and re-export it from the original if you are unsure.
If a panel fails one of these, fix it now. The same fix costs a five-minute edit before submission and a documented correction after.
Frequently asked questions
Compare every panel against the others for duplicated regions, inspect blots and gels for splicing without visible boundary lines, and confirm you have not over-adjusted brightness, contrast, or gamma in a way that erases or invents bands. Journals now run automated screens such as Proofig and ImageTwin at submission, so it is worth doing the same visual and tool-assisted check yourself first, while a fix is still a fix rather than a correction.
Two categories, and they are treated very differently. Beautification and honest error, like adjusting brightness across a whole image or accidentally reusing a control panel, are usually fixable if caught. Deliberate manipulation, such as cloning bands, splicing lanes without disclosure, or removing a feature, is misconduct. The trouble is that an automated screen flags the shape of both, so an honest duplication can still trigger an integrity review.
More common than most authors expect. A screen of 20,621 biomedical papers found that 3.8 percent contained problematic figures, and roughly half of those were suggestive of deliberate manipulation rather than honest error. That means a meaningful fraction of already-published work carries a figure that would not survive today's automated screening.
Yes, within limits. You may apply linear adjustments to the entire image, and you must disclose the adjustment. What you cannot do is adjust part of an image, apply non-linear changes that selectively strengthen or erase bands, or splice lanes from different gels without a clear dividing line. When in doubt, keep and submit the unprocessed original.
It is far worse than finding it before. A post-publication image flag can lead to a correction, an expression of concern, or a retraction, and it attaches to your name permanently in databases like Retraction Watch. Catching the same issue before submission turns a career risk into a five-minute fix at your desk.
Sources
- Bik, Casadevall, and Fang, The prevalence of inappropriate image duplication in biomedical research (2016): https://journals.asm.org/doi/10.1128/mbio.00809-16
- COPE, guidance on image manipulation in published articles: https://publicationethics.org/guidance/flowchart/inappropriate-image-manipulation-published-article
- Nature, image-integrity and figure policy for authors: https://www.nature.com/nature-portfolio/editorial-policies/image-integrity
Final step
Find out if this manuscript is ready to submit.
Run the Free Readiness Scan. See score, top issues, and journal-fit signals before you submit.
Anthropic Privacy Partner. Your manuscript is never used to train any model.