Publishers can integrate AI image integrity tools into their editorial workflows in a COPE-aligned way by using them as a screening layer that surfaces suspected issues for human review, not as a replacement for editorial judgment. The key is clear internal policy, documented process, and author communication that does not assume harmful intent.
Why Systematic Image Screening Is No Longer Optional
The scale of image integrity problems in scientific publishing has outpaced what manual review can address. The pressure is most visible at the submission stage: drawing on her experience as an image integrity analyst screening manuscripts before publication, Jana Christopher reports that roughly 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 call for human review rather than automated judgment.
These are not marginal numbers. With submission volumes rising across disciplines, no editorial team can visually inspect every figure in every manuscript. A study at the University of Split found that researchers and medical students correctly identified only a minority of image duplications when reviewing figures by eye. The human eye, even a trained one, is not sufficient.
Publishers need a structured, principled approach to image screening. That approach must align with the ethical framework most journals already follow: the guidelines set by the Committee on Publication Ethics (COPE).
What COPE Guidelines Require of Publishers
COPE’s Core Practices place specific obligations on editors and publishers regarding research integrity. Editors are expected to investigate concerns about the integrity of published or submitted work, to act on those concerns in a timely manner, and, critically, to avoid assuming guilt before an investigation is complete.
COPE’s flowcharts for handling suspected image manipulation provide step-by-step guidance for editors who encounter problematic figures. The process begins with identifying the concern, proceeds through author communication and institutional notification where appropriate, and concludes with an editorial decision informed by the evidence gathered. At every stage, the emphasis is on investigation and judgment, not on automated determination.
This framework has a direct implication for how publishers should think about AI tools. Any technology integrated into the editorial workflow must support the investigation process. It must not replace it. AI tools that surface suspected issues for editorial review are consistent with COPE’s model. Tools positioned as making final determinations about misconduct are not.
Where AI Screening Fits in the Editorial Workflow
AI-driven image integrity screening adds value at specific stages of the editorial process. The table below maps those stages to screening actions and corresponding COPE obligations.
| Workflow Stage | AI Screening Action | COPE Obligation |
|---|---|---|
| Submission (pre-peer review) | Automated scan of all figures for duplication, manipulation, AI-generated content, and image plagiarism | Editors should have processes to identify integrity concerns before committing reviewer resources |
| Peer review | Flagged results available to editors alongside reviewer reports | Editors must investigate concerns raised during review and not ignore potential problems |
| Acceptance (pre-publication) | Final confirmation that flagged issues have been resolved or explained | Editors should ensure that published work meets integrity standards |
| Post-publication | Retrospective screening if concerns are raised after publication | Editors must investigate post-publication concerns and issue corrections or retractions as warranted |
The highest-value integration point is at submission. Screening manuscripts before peer review begins gives editors actionable information early, before reviewer time is invested and before a manuscript moves toward acceptance. If a figure contains a suspected duplication or a region flagged for possible manipulation, the editor can query the author, request original files, or place the manuscript on hold for further review.
This is precisely the approach adopted by leading publishers. The Science family of journals adopted Proofig AI across all six journals in 2024 to screen for altered images before publication, after a multi-month pilot demonstrated the tool’s effectiveness at surfacing problems pre-publication. MDPI similarly integrated Proofig into its workflow, with Milos Cuculovic, Head of Technology Innovation at MDPI, noting: “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.”
Building a COPE-Aligned Screening Process: Practical Steps
Integrating an AI screening tool is not just a technical decision. It requires internal policy, clear communication protocols, and documentation. The following steps outline a practical model.
1. Define what a flagged result triggers. Not every flag warrants the same response. A minor duplication between two panels in a supplementary figure may require a simple author query. A figure with regions flagged for suspected splicing or content-aware editing may warrant an editorial hold and a request for original, unprocessed image files. Establish a tiered response protocol that distinguishes between low-risk flags (likely honest errors or figure preparation artifacts) and high-risk flags (patterns consistent with possible intentional alteration).
2. Communicate with authors neutrally. COPE’s guidance is explicit: do not assume intent. When contacting authors about flagged images, use neutral, factual language. Describe what the screening tool detected, for example, “overlapping regions were identified between panels A and C of Figure 3.” Ask the author to provide original files or an explanation. Many issues are introduced during figure preparation, by collaborators, or through workflow errors. Even researchers who uphold the highest standards may unknowingly submit figures with problems introduced by others involved in data handling.
3. Document the screening and response. Maintain an audit trail. Record which manuscripts were screened, what was flagged, how the editorial team responded, and what the outcome was. This documentation is essential if a post-publication concern arises, if an institutional investigation is initiated, or if COPE itself reviews the journal’s handling of a case.
4. Train editorial staff on interpreting results. AI screening tools surface suspected issues. They do not make diagnoses. Editorial staff need to understand what different types of flags mean, what constitutes a likely figure preparation error versus a pattern requiring deeper investigation, and when to escalate to a research integrity officer. Proofig’s reports are designed to support this kind of informed human review.
The Cost of Reactive Approaches
Publishers who rely on post-publication scrutiny, whether from platforms like PubPeer or from readers and integrity investigators, are operating reactively. Post-publication discovery of image problems is more damaging and more costly to resolve than pre-publication detection. Retractions carry significant financial and reputational consequences for journals, and the rate of retractions is increasing: approximately 0.2% of articles published in 2022 are now retracted, roughly triple the rate from a decade earlier.
Proactive screening shifts the burden from damage control to prevention. It protects reviewer time, editorial credibility, and the authors themselves, many of whom would prefer to catch and correct an honest error before publication rather than face a correction or retraction afterward.
What COPE Corporate Membership Means for Publishers
Proofig AI is a Corporate Member of COPE. In practical terms, this means Proofig has made an organizational commitment to operating within the ethical and procedural framework that COPE sets for the publishing community. It is not a certification of Proofig’s outputs, and it is not an endorsement of any specific detection result. It reflects alignment with the principle that tools serving the publishing ecosystem should support, not circumvent, the human judgment that COPE guidelines require.
For publishers evaluating image integrity screening partners, COPE Corporate Membership is one signal, among others, that a vendor operates within the norms of responsible publishing practice.
Practical Takeaway for Publishers
Integrating AI image integrity screening is not about automating editorial decisions. It is about giving editorial teams better information earlier, so that human judgment can be applied where it matters most. Publishers who build systematic screening into their workflows, with clear internal policy, COPE-aligned author communication, and documented audit trails, are better positioned to protect their journals, their authors, and the integrity of the scientific record.
Publishers ready to explore how automated screening fits their workflow can learn more about Proofig’s image integrity platform or explore PubShield, a manuscript quality assurance hub that brings image integrity screening together with text similarity, reference checking, and other checks in a single dashboard.
