AI tools are entering scholarly submission and review workflows to handle quality checks at a scale that manual processes can no longer sustain. Image integrity is a distinct, specialized part of this pipeline. Proofig AI plays a specific role here, surfacing suspected image issues for editorial review before manuscripts reach peer review or publication.
Why Is the Pressure on Editorial Workflows Growing?
Submission volumes across scientific publishing continue to rise, while editorial capacity remains constrained. Editors and reviewers cannot manually inspect every figure in every manuscript. This is not a failure of editorial diligence. It is a structural reality. The consequences of missed image issues are well documented: approximately 0.2% of articles published in 2022 are now retracted, roughly triple the rate from a decade earlier. Some of these retractions involve image-related concerns, including manipulation or duplication, as documented in case reports and Retraction Watch’s image manipulation archive.
The problem is even more visible at the pre-publication stage. According to image integrity analyst Jana Christopher, quoted in a UKRIO expert interview, about 20 to 35% of submitted manuscripts are flagged for image-related issues during editorial screening. Many of these flags reflect honest error rather than misconduct, which is precisely why they warrant human review before a manuscript advances.
These figures make the case that systematic, automated screening is not optional for journals operating at scale. It is a practical necessity.
Why Does Image Integrity Deserve Dedicated Attention?
Most discussion of AI in peer review focuses on text. Plagiarism detection, AI-generated writing identification, and reference checking have all received significant investment. Image integrity, by contrast, has historically received less systematic attention in editorial workflows, even though image-related problems are a leading cause of corrections and retractions.
A peer-reviewed study published in Research Integrity and Peer Review (2025) tested 831 medical students and 26 researchers on their ability to detect image duplications. Students correctly identified a median of only 29.4% of duplications, and researchers identified just 32.4%. The difference was not statistically significant. This underscores a critical point: unaided human review misses the majority of image integrity issues, regardless of training level.
Image screening is a separate, specialized problem. It requires tools purpose-built for detecting visual irregularities across scientific image types, not general manuscript quality assurance tools that treat images as an afterthought. This is the gap that Proofig AI was designed to fill.
What Does Image Integrity Screening Actually Catch?
Proofig AI provides five core capabilities, each addressing a distinct category of image integrity risk. The table below summarizes what each capability does in practical editorial terms.
| Capability | What It Detects | Why It Matters for Editors |
|---|---|---|
| Image Duplication & Reuse Detection | Duplicated or reused regions within a manuscript, including scaled, rotated, flipped, or partially overlapping images | Catches figure-preparation errors and accidental reuse before publication |
| Image Manipulation Detection | Alterations within a single image, including cloning, splicing, deletion, and content-aware edits | Surfaces suspected modifications that may affect data interpretation |
| AI-Generated Image Identification | Images created by AI models across microscopy, histology, Western blots, medical scans, and more | Addresses the growing challenge of AI-generated scientific images that may not accurately represent underlying data |
| Image Plagiarism (PubMed Source) | Reuse of images from previously published manuscripts, checked against tens of millions of images in PubMed | Identifies potential image plagiarism from the published literature |
| Self-Plagiarism Prevention (My Database) | Reuse of a researcher’s own previously published images via a personalized repository | Helps authors and editors prevent unintentional self-plagiarism |
Proofig surfaces suspected issues for human review. It does not make final determinations about intent or misconduct. Editors remain in control of every decision. Many image issues are unintentional, introduced through complex workflows, collaborator handoffs, or figure-preparation errors. Pre-submission and at-submission screening protects both the journal and the author.
How Serious Is the False-Alarm Problem?
One legitimate concern editors raise about automated screening is false positives. If a tool generates excessive noise, it wastes reviewer time and creates friction with authors. This concern is worth addressing directly.
Proofig’s false-positive rates have been validated as very low across large real-image datasets, including both microscopy and Western blot image types. The practical implication is that when Proofig flags an item, editors can treat it as a meaningful signal worth investigating, not as background noise. This level of specificity matters for editorial trust in the tool.
How Does Proofig Fit into an Existing Editorial Workflow?
Proofig integrates with Aries Systems, which is widely used for editorial workflow management in scholarly publishing. This means journals can add automated image integrity screening to their existing submission infrastructure without rebuilding their systems.
For institutions and publishers seeking a broader quality assurance layer, PubShield serves as a manuscript quality assurance hub. PubShield brings Proofig AI’s image integrity checks together with text similarity screening (powered by iThenticate), AI-generated text detection (powered by Pangram), reference integrity analysis (powered by RefGuard), and data reporting compliance (powered by DataSeer) into a single dashboard. This consolidation eliminates scattered logins and fragmented workflows. Importantly, each of these checks is powered by a specialized partner tool within the PubShield ecosystem. Proofig AI handles image integrity. The other checks are handled by their respective partners.
Who Has Already Adopted This Approach?
Proofig AI is not experimental. The Science family of journals adopted Proofig across all six journals in 2024, after a multi-month pilot demonstrated that problematic figures could be detected before publication. In his editorial announcing the adoption, Science Editor-in-Chief H. Holden Thorp described Proofig as analogous to iThenticate, which Science had used for seven years for text-plagiarism detection.
MDPI also committed to Proofig AI, implementing the tool in selected journals to reinforce its editorial process. Milos Cuculovic, MDPI’s Head of Technology Innovation, stated in an MDPI Blog “From the experts, it seemed Proofig AI was by far the best choice, and here we are: further ensuring research integrity and guaranteeing the high quality of research output published by MDPI.”
These are institutional commitments from major publishers, not pilot programs. They signal to editors at other journals that systematic image screening has moved from aspiration to operational practice.
What Should Editors Take Away?
Systematic image screening is no longer a task that scales with manual effort. As submission volumes grow, the gap between what editors can inspect by eye and what actually needs checking will only widen. AI tools like Proofig AI exist to close that gap by surfacing suspected issues before they reach peer review or, worse, post-publication scrutiny.
The goal is not to automate editorial judgment. It is to ensure that editors and reviewers spend their expertise on the judgment calls that require human assessment, rather than on the detection work that AI can handle more consistently and at greater scale. Editors who integrate image integrity checks earlier in the workflow are better positioned to maintain the standards their journals depend on, while protecting both their reputation and the researchers who submit to them.
