Perspectives on AI in scholarly communications
A discussion with libraries and publishing professionals
Perspectives on AI in scholarly communications
A discussion with libraries and publishing professionals
Foreword
Saskia Steinacker
Chief Digital Officer, Springer Nature
It is not new to say that artificial intelligence (AI) is having a dramatic impact not only on scholarly communications, but the entire process of research. It continues to provide many new opportunities for the academic community – enabling us to collectively accelerate discovery, protect integrity and promote equity, ultimately delivering better outcomes for research and for wider society.
Technology has always played an important role at Springer Nature - it has been central to our growth and development as a publisher, enabling us to deliver on our commitments to the research community. We have been exploring how we can best use AI to help researchers get published faster and find relevant content more quickly, to make sure research can be trusted, and to reduce barriers to equity. We have been using AI, always with human-centred values, for over 10 years already in support of these goals, with some encouraging results.
AI development remains, though, a fast-moving area. As use cases continue to evolve, we apply the same considerations in our approach as we have to date: Does it truly benefit the research ecosystem? Does it meet our AI governance principles of fairness, transparency, accountability, and respect?
This means that working collaboratively with key stakeholders to better understand needs, challenges and opportunities is more important than ever. Libraries, in particular, have an important role to play within their institutions regarding AI adoption,1 leading not only on information discovery but also increasingly supporting researchers in their usage of AI tools. Strong partnerships between libraries and publishers are essential in an AI-driven world – combining deep expertise in information stewardship with innovative publishing and discovery practices, to responsibly harnessing emerging technologies for the advancement of research and knowledge.
Strong partnerships between libraries and publishers are essential in an AI-driven world – combining deep expertise in information stewardship with innovative publishing and discovery practices, to responsibly harnessing emerging technologies for the advancement of research and knowledge.
This perspectives report brings together the views and experiences of select library staff and some of Springer Nature’s AI experts to provide a top-level overview of current AI implementation in scholarly communications. In sharing these perspectives, our aim is to build on our collaborative approach to provide a view on AI use across publishers and libraries, share case studies, and discuss challenges. We thank all interviewees for their invaluable insights on this critical topic.
1 CILIP. (2021). The impact of AI, machine learning, automation and robotics on the information profession.
Introduction
AI is changing the landscape of scholarly communications, and of wider society, at an unprecedented pace.
A Nature survey conducted in 2023 found that researchers saw many positive impacts of AI, such as providing faster ways to process data, speeding up computations, and saving scientists time and money.2 In 2025, an unpublished survey conducted for Springer Nature (n=2,021) revealed that 57% of researchers have used an AI tool to help them stay up to date with published research or to read research papers, while 52% have used an AI tool to write research papers or grant applications. Of those who have used AI tools, 80% expressed their intention to continue using them.
In 2025, an unpublished survey conducted for Springer Nature (n=2,021) revealed that 57% of researchers have used an AI tool to help them stay up to date with published research or to read research papers, while 52% have used an AI tool to write research papers or grant applications.
Earlier this year we interviewed several library partners, as well as several stakeholders across Springer Nature, to get their insights around approaches to AI usage throughout the research lifecycle including research support and publishing. This perspectives report is a presentation of those discussions, with views ranging from the enthusiastic and experimental, to cautious scepticism.
2 Van Noorden, R., & Perkel, J. M. (2023). AI and science: What 1,600 researchers think. Nature, 620, 672-675.
Interviewees
Letícia Antunes Nogueira
Head of Section Resources & Digital Service, Norwegian University of Science & Technology Library, Norway
Dr Santhosh KV
Deputy Director Research, Manipal Academy of Higher Education, India
Beth Montague-Hellen
Head of Library & Information Services, Francis Crick Institute, United Kingdom
Keith Webster
Dean of University Libraries and Director of Emerging and Integrative Media Initiatives, Carnegie Mellon University, United States of America
Heather Devereaux
Former Vice President AI Product Delivery, Springer Nature
Chris Graf
Director of Research Integrity, Springer Nature
Henning Schoenenberger
Vice President Content Innovation, Springer Nature
Harald Wirsching
Executive Vice President Data & Analytics Solutions, Springer Nature
How is AI currently being used?
To begin, the interviewees were asked to explain some of the ways they are using AI tools, with use cases ranging quite broadly.
As a publisher and solutions provider, Springer Nature’s focus is on a broad set of capabilities designed to assist or automate tasks within the research, editorial and scientific publishing workflow – using AI to augment human capabilities rather than replace them. For libraries, the view is more cautious, with concerns around “hype” regarding AI, and progress observed in some areas but not others.
The most common use cases from the discussions are outlined below.
Improving information discovery
As mentioned in the introduction, over half of researchers have used an AI tool to help them find or read research papers, so it comes as no surprise that this was a common point in many of the interviews. Letícia Antunes Nogueira, Head of Section Resources & Digital Service at the Norwegian University of Science & Technology Library, observed a shift within her institution: “We are seeing now, especially with students who are coming from high school, they have been used to searching using AI systems [...] so these tools become ever more ubiquitous in making recommendations and finding literature.”
Keith Webster, Dean of University Libraries and Director of Emerging and Integrative Media Initiatives, Carnegie Mellon University added that AI is often “used in synthesising literature, either auto-summarisation or auto-synthesis.” He continued, “A few of us joke that we will never read a whole article again [and] from the perspective of researcher efficiency, that’s great.” Within his library, Keith has deployed AI discovery tools such as Keenious and Scite, which have received positive feedback from library users.
Also related to aiding researchers in being more efficient, information discovery is a key area for AI development within Springer Nature. Harald Wirsching, Executive Vice President Data & Analytics Solutions at Springer Nature, explained: “Many of our data and analytics products are extracting data from scientific literature and putting it in a structured format, or combining content from different sources.”
Harald gave two examples of this work in practice:
- AskAdis, an AI-chatbot enhancement to the AdisInsight drug discovery database that enables users to ask questions in natural language and return relevant information.
- Methods Muse, a platform that supports experimental work by streamlining protocol design, implementation, validation, and optimisation.
Streamlining manual processes
Both inside and outside the research community, it is widely accepted that AI can help automate manual tasks, enabling people to focus on the parts of research where their skills offer the most value – and this was confirmed by our interviewees. In line with Keith’s comments above, Dr Santhosh KV, Deputy Director Research at Manipal Academy of Higher Education, noted that AI has proven to be incredibly beneficial for his students, particularly when conducting literature surveys and background research.
AI can help automate manual tasks enabling people to focus on the parts of research where their skills offer the most value.
Beyond using generative AI models for literature research, Beth Montague-Hellen, Head of Library & Information Services at the Francis Crick Institute, observed other AI-enabled processes being used at her institution to save researchers time, for example AlphaFold to provide protein structure predictions, and Microsoft Copilot to assist with coding. Keith agreed, citing use cases from writing grant applications to analysing datasets to translating research for local audiences.
Publishers, too, are using AI-assisted tools to support a more streamlined and integrated publishing experience, whilst maintaining the quality and integrity expected of the research they publish. Springer Nature has embedded AI tools within its peer review platform, Snapp, to streamline processes which typically involve manual effort for authors and editors, such as finding the right journal to submit a paper, or finding the right reviewers.
Heather Devereaux, former Vice President AI Product Delivery at Springer Nature, elaborated on the example of the reviewer finder tool. “The editorial teams were spending upwards of 90 minutes trying to find appropriate reviewers,” she explained. “The reviewer finder is now one of the most requested editor features because they really struggle with finding the right reviewers, and having technology help them with that is absolutely valued.”
Supporting research integrity
Understandably, there has been a lot of discussion within the academic community about the impact of AI on research integrity,3 and this was mentioned by all of the library staff we interviewed. Alongside concerns, there was also an acknowledgement that AI presents many opportunities across research when used ethically and responsibly. Keith reflected, “I am increasingly worried about where AI may be used to create content that is not top-quality [...] but perhaps there is a silver lining if publishers can use AI to pinpoint likely candidates that are in breach of research integrity standards.”
AI presents many opportunities across research when used ethically and responsibly.
Chris Graf, Director of Research Integrity at Springer Nature, described how Springer Nature's Research Integrity Group is investing in the latest technologies to help identify unethical behaviour: “In the integrity space, we found that paper mills or networks of bad actors were using large language models to generate fake articles. And we thought it was reasonable, therefore, to use a large language model to help us spot this.”
There are several tools that have been successfully developed and implemented in the past few years, including tools to detect nonsense AI-generated text, identify irrelevant references, and look for problematic images. All of these tools work with human oversight, providing signals as to whether further human assessment is needed. Chris explained that while nonsense AI-generated submissions only make up a small subset of submitted articles, and human decision-making is still required, these tools are making “a small but meaningful difference” by alleviating the workload on editors and peer reviewers, allowing them to focus on high-quality submissions.
3 Chen, Z., Chen, C., Yang, G., He, X., Chi, X., Zeng, Z., & Chen, X. (2024). Research integrity in the era of artificial intelligence: Challenges and responses. Medicine, 103 (27), e38811.
What needs to be considered when implementing AI?
All interviewees highlighted that while AI brings immense opportunity, it also requires thoughtful, transparent implementation, grounded in ethics and critical thinking, and always aligned with the needs and values of its users.
We asked interviewees to discuss some of the considerations they have taken into account when implementing AI technologies, the key points of which are summarised below.
Facilitating ethical usage of AI tools
All interviewees highlighted the importance of ensuring AI is implemented with ethics principles at the core, with both publishers and libraries playing a pivotal role in educating and supporting researchers in using these tools. As Santhosh reflects, “Good or bad, [AI has] already penetrated in a large sense. So now the only thing is educating and creating awareness of what are the good things in terms of AI, what are the do-nots.” Beth agreed: “We just have to make sure that we’re including it in all our training [...] and pointing people to the right things in the same way that we discourage people from just using Google to look for sources.”
Both publishers and libraries play a pivotal role in educating and supporting researchers in using AI tools.
In some cases, this type of education is supported by institutional committees in partnership with the library – for example, Keith noted that Carnegie Mellon University benefits from having three centres for ethical or responsible use of AI, and Santhosh works closely with Manipal Academy of Higher Education’s ethics committees.
Interviewees from Springer Nature also stressed the need for publishers to ensure the sustainable and ethical use of AI by authors, across the research community, and internally as developers of AI based solutions. Henning Schoenenberger, Vice President Content Innovation at Springer Nature, pointed towards Springer Nature’s AI Principles of fairness, transparency, accountability, privacy and minimising harm. “As a publisher,” he said, “we are in the position to provide an infrastructure and services for researchers to create reliable content faster and more efficiently, while also providing guidance for best practice which sustains trust, integrity and reputation in line with AI safety. This is only possible through human oversight of technology.”
Ensuring there is a human in the loop
Following from Henning’s point above, it was unanimously agreed amongst our interviewees that ethical use of AI is only possible with human intervention; it should never be used as a replacement for human intelligence. Both Beth and Letícia expressed concerns about overreliance on AI as a substitute for critical thinking, and stressed the importance of human ownership in the scholarly process. Letícia summarised, “AI is important, it’s useful, but it’s not everything, and it definitely will not substitute the need for making judgements and the need for taking responsibility for what we publish.”
Ethical use of AI is only possible with human intervention; it should never be used as a replacement for human intelligence.
Springer Nature interviewees also echoed this from a publishing and product development standpoint. Henning shared his approach to integrating AI tools into publishing workflows, focusing on streamlining more admin-heavy tasks while ensuring that the integrity and quality of research and services is not compromised, and researchers remain in full control of their work.
Harald agreed, stating that human quality control is particularly paramount as AI tools continue to develop: “We always called it AI-assisted humans, where we used AI to assist humans to do their processes more efficiently. However, we are now coming into the area where we can also talk about human-assisted AI: where we can extract information automatically, but always have humans in the loop to make key decisions.”
Assessing data quality
AI tools are typically trained on vast amounts of data which, without proper oversight and curation, can perpetuate existing biases. Beth expressed her concerns about this: “I am bothered about the kind of doubling down effect on any biases that are already in training material. Large language models tend to the mean, that’s what they’re designed to do, [and] the internet is a really dreadful place a lot of the time.”
Many interviewees talked about the significance of robust and diverse data quality when developing, training and implementing AI tools. Keith underscored, “We are training AI tools on as diverse and representative a dataset as possible. I think that, in the research space, is the big one.” Heather echoed this from a publishing perspective: “There was a moment a couple of years ago when these AI systems came out where companies thought, oh good, we don’t have to fix our data problems, you can feed it anything. But it’s still ‘garbage in, garbage out’. At Springer Nature, we have a data competence centre and a team that’s focused on AI-ready data, so we’re making quality, ethics and sustainability of data absolutely essential for our AI systems.”
Santhosh approached this issue from another angle, acknowledging the need for libraries and institutions to ensure their researchers are properly assessing the quality of data when using AI tools. “We make sure that we talk to the researchers, ask them what is the source of the data, how is it validated,” he said. “Then in one or two cases we also verify it on a random basis to check the authenticity of the data.”
Developing tools collaboratively and with purpose
Since the launch of ChatGPT at the end of 2022, the growth of AI tools on the market has been explosive, and for some it may seem as if AI is increasingly regarded as the answer to every problem. Beth remarked, “Often people see the shiny thing and think, how can I integrate the shiny thing into whatever it is that I’m pushing, without thinking about whether it’s actually the right tool for the job.”
From their experiences developing AI tools, Chris and Heather emphasised that it is essential to truly understand the use cases and problems that you are trying to solve. As Chris said, “We all kind of expect AI to come charging over the hill like a knight in shining armour, but it requires both technical and subject matter expertise. Do you really understand the problem, and can you help to therefore validate that the tool is doing what you want it to do?” Heather added, “It’s not AI for AI’s sake, it’s AI for reason, it’s AI for purpose. It’s got to drive your community’s success, or you shouldn’t use it at all.”
Interviewees agreed that this is a key area for collaboration between publishers and libraries, as well as researchers – AI tools should be co-created in order to truly benefit the research ecosystem. Beth and Keith highlighted the importance of libraries engaging early with the community to best understand their needs, and then working with publishers and product developers to test new solutions. Keith expressed that his library is in conversation with many publishers and AI companies: “Being able to represent some of the researcher needs we hope can help product development more broadly.”
AI tools should be co-created in order to truly benefit the research ecosystem.
This sentiment was also evident on the publisher side. Harald shared that Springer Nature prioritises continuous communication with universities, libraries and researchers to explore new AI use cases. He explained, “We are seeking feedback about the usefulness of what we are developing, while being open to feedback about concerns that people might have.” This ensures that user needs are fully understood and that new tools actually serve the people they are created for.
"We are training AI tools on as diverse and representative a dataset as possible."
Keith Webster
Dean of University Libraries and Director of Emerging and Integrative Media Initiatives, Carnegie Mellon University, United States of America
"It’s not AI for AI’s sake, it’s AI for reason, it’s AI for purpose. It’s got to drive your community’s success, or you shouldn’t use it at all."
Heather Devereaux
Former Vice President AI Product Delivery, Springer Nature
“Good or bad, [AI has] already penetrated in a large sense. So now the only thing is educating and creating awareness of what are the good things in terms of AI, what are the do-nots.”
Dr Santhosh KV
Deputy Director Research, Manipal Academy of Higher Education, India
What does an AI-enhanced future look like?
AI has immense potential to benefit the academic community, and improve the ways research is conducted and disseminated, however in many cases the tools are not quite where they need to be in terms of quality of output and time saved.
This section summarises our interviewees’ hopes for the future, and where they see AI having impact – alongside those use cases that have already been discussed above.
Changing the role of the library
Keith reflected that the digitalisation of research content has made it less likely for researchers to come to the library, but AI tools provide a new role for librarians, as information specialists, to play within researcher workflows: “I can see scope for us to be the experts in AI-powered tools and workflow solutions.”
This sentiment was shared by the other library staff interviewed, who all spoke to the ways libraries can support the research ecosystem as technologies continue to change. As Letícia summarised, “AI doesn’t change the mission of libraries, but it does change the environment where knowledge is produced [...] That also gives us the opportunity to renew our role as a library in the university system. We are not just a deposit of books, we are people with expertise on, for instance, assessing the relevance of sources or finding the right information.” Santhosh concurred that in the future, “the library may just not be the content provider, they may be there for content validation or content security.”
Keith also expressed that he would like to see AI used more within library processes, to help take librarians’ time away from administrative tasks (such as answering questions about library opening times and cataloguing metadata) to allow them to play a more strategic role within the university, for example with predictive analytics. “Where I see potential is the administrative efficiency side,” he said, “where we can perhaps free our colleagues from some of the more routine work that allows them to concentrate on the value-add activities.”
Working together to protect the scientific record
Chris reflected that there are many rightful concerns around AI and research integrity, but with the developments being made in this space, his overall view for the future is positive: “In three to five years’ time, there will be a lot less fear and uncertainty and denial, and it will be much more normal to use AI in everything.” But he stressed that the only way to reach this future is for all stakeholders within the research ecosystem to collaborate.
There are various ways we can all collaborate across stakeholder groups. For example, Springer Nature sits on various committees such as the Committee on Publication Ethics, the UK Committee on Research Integrity, the STM Integrity Hub and the STM Standards & Technology Committee, working with academics, institutions and other publishers to solve integrity challenges and protect the scientific record. This allows the community to share ideas and innovations, which is essential in such a fast-moving landscape.
According to Letícia, each stakeholder group within the research community approaches AI with a different perspective, but they should use these diverse perspectives to work towards a common goal: “Publishers, libraries and researchers have different roles and we have to be aware of that. But it’s important that we all think about how we make sure that the integrity of the information is kept.” She stressed that this goes beyond just AI to address larger systemic issues such as research assessment: “As far as we have tools that offer a more efficient way of doing things and people are rewarded on the quantity of publications [...] they are going to prioritise that efficiency.”
Democratising access to knowledge
In several interviews, participants highlighted that they see potential for AI to improve accessibility and inclusivity within research, but more work needs to be done both to develop suitable tools, and ensure they are used responsibly. Beth noted, “When [generative AI] first came out, I had quite a lot of hopes that it would help level the playing field with people who don’t have English as a first language, or people who are dyslexic, things like that.” However, she continued that this has not necessarily been the case – text produced by large language models can often be formulaic, and when readers recognise overused phrases in research, it may undermine the perceived credibility of said research and, in turn, contribute to further bias or discrimination.
There is potential for AI to improve accessibility and inclusivity within research.
Santhosh reinforced the need for better guidance and guidelines from publishers on these kinds of use cases, which he described as a “grey area”: “[In India] not everyone has good English, so students often use tools like Grammarly or Paperpal – but what is the level in which they can do so? Is it allowed, and where is the limit?”
Keith also added that the “transformative potential” of AI may lie in its ability to make research content more accessible to non-specialist audiences. “For 20 years now we’ve been talking about open access as a way of opening up research to the world, [but] the citizen who has paid for the research through taxes, and may be interested in the latest medical conditions that their loved ones suffer from, can’t understand the articles. Can we use AI to make it more understandable?”
"AI doesn’t change the mission of libraries, but it does change the environment where knowledge is produced"
Letícia Antunes Nogueira
Head of Section Resources & Digital Service, Norwegian University of Science & Technology Library, Norway
“In three to five years’ time, there will be a lot less fear and uncertainty and denial, and it will be much more normal to use AI in everything.”
Chris Graf
Director of Research Integrity, Springer Nature
Key takeaways
The views expressed within this perspectives report are by no means exhaustive, but provide a snapshot of the discussions around AI which are taking place between libraries, publishers and wider organisations reflective of the academic community.
To summarise the main points:
Responsible implementation is necessary
AI use cases are currently centred around improving information discovery, streamlining manual or laborious tasks within the research process, and identifying research integrity issues. Libraries and publishers have key roles to play, working together to support the community and make sure that AI is implemented responsibly.
Ethical considerations are critical
It is vital to keep ethical and sustainability considerations at the forefront when developing and implementing AI technologies – in particular, ensuring human oversight and critical thinking, analysing data quality and training and supporting researchers.
Cross-stakeholder collaboration can be impactful
There are opportunities to use AI to support publishing workflows, protect research integrity and improve equity, but this has the greatest impact when done collaboratively and always with human decision-making.
Stay informed
To receive regular news and resources on AI and other topics, sign up to The Link Alerts. Our communications bring together the latest knowledge from Springer Nature and other community experts to provide valuable insights to those who work within a research environment.
For wider context and the latest updates on Springer Nature’s use of AI:
Stay informed
To receive regular news and resources on AI and other topics, sign up to The Link Alerts. Our communications bring together the latest knowledge from Springer Nature and other community experts to provide valuable insights to those who work within a research environment.
For wider context and the latest updates on Springer Nature’s use of AI:
Share your thoughts
We'd love to hear your feedback about this topic. Please share your thoughts by completing this short poll.
How to cite this report
Springer Nature (2025, October). Perspectives on AI in scholarly communications: a discussion with libraries and publishing professionals. https://stories.springernature.com/AI-perspectives/index.html
Built with Shorthand