What's changing in AI & research integrity - reflections from WCRI 2026

Author: Simone Culurgioni

Team Lead, Research Integrity, Prevention

_{Highlights on the technological trends, trust and transparency from the 9th World Conference on Research Integrity, Vancouver}

Last month I had the opportunity to attend WCRI 2026 in Vancouver, joining colleagues from across the research integrity community to step back and reflect on how research integrity is evolving. As part of my role at Springer Nature, many of the conversations felt familiar, echoing challenges we’re already working on, while also offering additional perspectives on where things may be heading. 

AI was inescapable; not as a novelty, but as a thread running through almost every session I attended. What struck me most was how the conversation has matured.  Fewer people are now asking whether AI is changing research integrity, almost everyone now accepts that it is. The more interesting question, and the one that dominated the sessions I found most valuable, is what a meaningful, coordinated response could look like.  

One of the clearest signals of this shift came from the inclusion of sessions focusing on the development of AI-based detection systems that identify large volumes of submissions or analyse peer review activity for unusual patterns (we’re developing similar systems at SN). The scale of the challenge, and the need for responses that can operate systemically rather than case by case, was a consistent theme across the conference. 

Alongside this, specific integrity challenges highlighted how quickly the landscape is evolving. One widely discussed finding, from an analysis of BMJ Group journals, showed that just 5.7% of authors disclosed any AI use, even when they were required to. While many authors appear to be using these tools to improve writing, the gap between use and disclosure is hard to ignore.  It’s something we recognise too from our own work, and it highlights the need for clearer, more consistent guidance. 

Other sessions focused on the difficulty of identifying AI-generated or manipulated content. A pre-conference bioimaging workshop, for example, invited participants to distinguish genuine microscopy images from AI-generated ones, demonstrating how challenging this is in practice and how quickly generative technology is advancing.  

Throughout the conference, discussions reinforced that continuously updated detection systems, training and human expertise will all be needed to address these challenges effectively. 

In response to these pressures, there is a growing emphasis across the community on developing more coordinated approaches. The conference’s Focus Track (working toward a Global Reporting Standard for AI Disclosure in Research, supported by COPE and STM) felt particularly timely. The ambition is to produce what has been named the Vancouver Standard: a shared framework for AI disclosure that works across disciplines and publication cultures (more about this in Chris Graf’s opinion piece in Research Professional News). It is an initiative that will benefit from broad engagement across the research community. 

These discussions closely reflect work already underway at Springer Nature. Within the Research Integrity group, and particularly in the Prevention team, we have been developing approaches that align with many of the trends discussed at WCRI. One area of focus is the growing issue of hallucinated or unverifiable references. To address this, we have been working collaboratively across the company to expand our irrelevant reference detector towards detecting hallucinated or unverifiable references above a certain threshold. As with all AI-supported approaches, human oversight remains essential, particularly given the risk of false positives. We continue to refine our approach, taking care to improve accuracy, while carefully mitigating these risks.  

Alongside this, our work is increasingly centred on large-scale, data-led analysis that includes identifying patterns of misconduct through analysis of author networks and identifying peer-review duplication, which aren’t visible in a single submission.  This shift, from case-by-case investigation to system-level insight is helping to strengthen preventive measures and respond more effectively at scale. 

It was also encouraging to see colleagues from Springer Nature contributing directly to these discussions. Research integrity advisor Manisha Wardhwa emphasised the importance of not jumping to conclusions when assessing potential AI misuse, highlighting the need to look carefully at the metadata and carry out thorough investigations. Ed Gerstner addressed a related but equally important issue: the growing gap between how researchers are assessed and how they would prefer to be assessed, and the role that reform could play in shaping research culture. Together, these contributions highlighted the continued importance of careful judgement, support and education in strengthening research integrity and preventing misconduct. 

A plenary session on AI ethics added another important dimension, highlighting the need to consider global perspectives. Speakers Dr Titilola Olojede and Dr Retha Visagie reflected on the risks of developing AI ethics from a single cultural viewpoint, and the importance of accounting for differences in access, education and research contexts worldwide. 

What I found most encouraging at WCRI was a shared willingness to move beyond diagnosing challenges towards taking action. While tools are improving and policy frameworks are beginning to take shape, the conference reinforced that one of the most complex parts of this work is ensuring that researchers at every career stage and every part of the world can engage meaningfully with these developments and that the norms we build are ones they recognise as their own. 

If there is one overarching takeaway, it is that safeguarding the scholarly record will depend on how effectively we connect these efforts globally.  The challenge now is less about recognising risks, and more about aligning responses across systems, organisations and regions. Without that alignment, even the best tools and policies will fall short. 

Learn more about Springer Nature’s approach to research integrity here.