Artificial Intelligence


Artificial intelligence (AI) plays an increasingly important role as a tool for academic work. This webpage provides an overview of the responsible use of AI tools.
A general overview of AI competence at the University of Bayreuth is available on the central AI webpage of the University of Bayreuth.

Basics of Working with AI

Bayreuth University provides free access to ChatGPT and DALL-E through existing campus licences, integrated into Microsoft Copilot. Copilot uses ChatGPT for text generation and DALL-E for image generation.
The service is accessed using your own University ID. The data entered is not used for training purposes.
In addition, the Academic Cloud provides access to open-source AI models such as Apertus (a collaborative project of EPFL, ETH Zurich and the Swiss National Supercomputing Centre CSCS), Llama (Meta), Mistral (France) and DeepSeek R1.
Members of the University can use “ChatAI” free of charge via the Academic Cloud using federated authentication through UBT. Requests sent to the locally hosted services, which are operated on the servers of GWDG (Gesellschaft für wissenschaftliche Datenverarbeitung mbH Göttingen), are processed exclusively on GWDG systems.
Please consult the GWDG website for information about the available language models and data protection notices: Academic Cloud: 4. Services – AI Services – Chat AI.
The University of Paderborn also provides general information on AI language models.
Further information can be found on the website of the IT Service Centre.

TL;DR - The Most Important Points at a Glance

Typical use cases for AI:
  • Develope search terms for literature research
  • Summarise texts
  • Better understanding of texts in foreign languages
  • Correcting spelling and grammatical errors
  • Explain topics and clarify comprehension questions
  • Deepen understanding of subject-specific concepts
  • Develope initial outlines or structural ideas
  • Assistance and inspiration for problem-solving and decision-making
  • Generating practice exercises
Please note:
  • Information on whether AI use is permitted and how it must be documented should be obtained from the respective chair.
  • AI can make mistakes or generate inaccurate information.
  • Generated references are not always reliable.
  • You remain responsible for your content.
  • Do not enter sensitive or legally protected information.
  • Use the AI solutions provided through the University of Bayreuth that enable privacy-conscious and data-minimizing use.
  • As with literature research, the same applies to prompting: An initial result often serves as a starting point for further questions and refinements. Please keep in mind the limitations of the used AI (e.g., a tool’s data source)

Searching for Literature

Generative AI tools such as Microsoft Copilot can usefully complement literature searches, for example by assisting with the formulation of suitable search terms or helping to structure subject areas.
Example prompts:
  • ‘Give me an overview of subject X.’
  • ‘Which search terms are suitable for researching XY?’
However, such tools are not reliable for finding specific sources. Chatbots may invent references, provide incomplete bibliographic information or cite literature that does not exist in the stated form. You should therefore always verify any sources suggested by AI, even if they appear plausible or include a DOI, for example by conducting an additional search in library catalogues or databases.
This is due to the way language models work: they generate texts by statistically combining previously learned patterns and content rather than by accessing an actual database of verifiable sources. When AI systems produce incorrect information or information that is entirely made up, it is referred to as ‘hallucination’.
Some subject-specific literature databases also offer AI support, enabling summaries or natural-language searches based on the sources they contain. One example is the WISO database, which is licensed by Bayreuth University Library. In addition to its conventional search function, the CAS SciFinder database also provides an AI-assisted search function. These AI assistants enable targeted searches for content based on published scientific information contained in the respective database.
There are also freely available AI research tools that you can use. AI functions provided by these tools may include summaries or network-based visualisations of similar texts. Helpful overviews of AI research tools are available from the University Library of the University of Duisburg-Essen, the Frankfurt University of Applied Sciences Library and RheinMain University of Applied Sciences. However, searching for literature with the aid of AI tools cannot replace research in well-curated systems such as library catalogues, including the Bayreuth University Library catalogue, and subject databases available through DBIS. AI cannot conduct a well-founded critical assessment of literature and is susceptible to misinformation if the underlying data is incorrect or incomplete. Human expertise therefore remains indispensable.
Please note:
  • AI is not a substitute for a systematic literature search.
  • Bibliographic references may be incomplete or incorrect.
  • Always verify bibliographic references in the Bayreuth University Library catalogue or in subject databases.

Evaluating Sources

Whether content is produced by people or AI tools, careful evaluation of sources remains essential. For AI-generated content, the SIFT Method is recommended.
  • S = Stop and reflect: Does the statement seem plausible or emotionally manipulative?
  • I = Investigate the source: Who created the content and for what purpose?
  • F = Find Better Coverage: Can the claims be confirmed by independent sources?
  • T = Trace Claims, Quotes, and Media to their Original Context: Where does the information originate? Are primary studies, data, methodology, and bibliographic references transparently accessible?
These steps help identify misinformation, oversimplified representations, or AI-related errors at an early stage.

Summarising Texts

AI tools can be used to summarise longer texts, articles or documents and to identify key points. This can be particularly helpful when first familiarising oneself with a subject area or when seeking a quick overview of extensive texts.
Example prompt:
  • “Summarise this article in five bullet points.”
It is important to note that summaries always represent an interpretation of the original: Content may be abridged, emphasised or unintentionally distorted. AI-generated summaries therefore do not replace reading the original text but serve only as a support tool. Responsibility for the accuracy, completeness and contextualisation of the content remains with the user.
From a data protection perspective, the use of the ChatAI service within the Academic Cloud is recommended. Depending on the language model selected, personal data and content are stored only locally and deleted once the session ends (federated login via the University of Bayreuth using a University ID is possible). When using institutional access (login with a University ID), inputs entered into Microsoft Copilot are likewise not used for training purposes; however, uploaded files are stored within Microsoft's cloud environment (OneDrive).
Please note:
  • AI-generated summaries are interpretations. Content may be simplified or distorted.
  • Ensure that neither the information you enter nor the results you receive infringe copyright.
  • Use AI for orientation and support, not as a substitute for reading the original text.
  • Do not present AI-generated texts as your own work.
References:
University Library Duisburg-Essen: AI Tools for Literature Research and Academic Work (last accessed: 5th May 2026)
Academic Cloud: AI chatbot ChatAI: AI Chatbot ChatAI (last accessed: 20th August 2026)

Translating Texts

AI-based translation tools enable the rapid translation of texts between different languages, for example from English into German or vice versa. This can support the understanding of academic literature written in foreign languages and the preparation of your own texts.
Example prompt:
  • “Translate this paragraph into German.”
However, particular caution is required when working with academic texts: technical terminology, nuances of meaning and stylistic features may be lost or translated inaccurately. Translations should therefore always be reviewed and, especially in the case of key passages, compared with the original text.
Data protection aspects are also relevant in this context. For external services such as DeepL, it should be noted that only the paid version guarantees that entered texts will not be used for training purposes. Institutionally licensed tools such as Microsoft Copilot or services provided through the Academic Cloud often operate under stricter data protection frameworks. Please familiarise yourself with the relevant policies before using these tools.
Please note:
  • Technical terms may be translated inaccurately.
  • Nuances of meaning may be lost.
  • Always check important passages against the original text.
  • Do not enter sensitive or protected content into AI tools.

Understanding Texts

AI tools can be used to engage in a dialogue with texts or documents, for example by asking questions about the content, requesting explanations of concepts or generating summaries of relationships and connections. This can make it easier to understand the text, particularly when dealing with complex or technically demanding material.
Example prompts:
  • “Explain method X in an accessible way.”
  • “What does term Y mean in the context of Z?”
Such functions are particularly useful for supporting the processes of reading and understanding texts. However, they do not replace in-depth analysis or critical engagement with the material. AI-generated responses should always be checked for plausibility, accuracy and subject-specific correctness.
For this type of use, the ChatAI service provided through the Academic Cloud is particularly recommended. Depending on the language model selected, personal data and content are stored only locally and deleted once the session ends (a federated login via the University of Bayreuth using a University ID is available). Only texts for which appropriate usage rights exist, or which are freely accessible (e.g. Open Access publications under a suitable Creative Commons licence), should be processed.
Please note:
  • Verify the results carefully. Content may be shortened, simplified or distorted.
  • Use AI strategically as a tool to support your own reading, not as a substitute for reading the original material.
  • Do not enter sensitive or protected content into an AI tool.

References:
German Research Ombudsman: FAQ Artificial Intelligence and Good Research Practice (last accessed: 5th May 2026)
TU Berlin: Designing Higher Education Teaching with Artificial Intelligence (last accessed: 5th May 2026)

Ethics and Legal Issues

Ethics and Bias in the Use of AI

As the use of artificial intelligence becomes increasingly widespread, questions of ethics, transparency and the responsible use of AI are gaining growing importance. AI systems often appear objective; however, they can inherit existing social biases and stereotypes from their training data and may therefore generate discriminatory or distorted results. This may occur, for example, through one-sided source materials, problematic datasets or the reproduction of stereotypes in image generation (see UNESCO: AI and Discrimination).
AI-supported tools can be valuable for learning, research and analysis, but they do not replace a critical and reflective approach to information, as AI-generated outputs are never completely neutral. Key prerequisites for the responsible use of AI are therefore traceability, explainability and transparency, as discussed, among other sources, in the White Paper of the German Federal Office for Information Security on the Transparency of AI Systems (press release of 5th August 2024).
At the same time, the resource-related dimension of AI must also be considered. Every AI query relies on energy-intensive data centres that consume substantial amounts of electricity, water and raw materials (see Artificial Intelligence from a Sustainability Perspective, German Bundestag, 2024). Ethical engagement with AI therefore extends beyond issues of fairness and bias to include questions of sustainability, climate protection and responsible resource use (Keeping AI in Perspective – Pathways Towards Climate-Friendly and Socially Responsible Use of Artificial Intelligence).
Further learning resources, such as the free learning module “Bias and Fairness in AI Systems or the guide Fair AI Prompting, can help users develop a reflective understanding of the opportunities and risks associated with AI.

Legal and Data Protection Considerations When Using AI Tools

Before using an AI tool, it is recommended that you familiarise yourself with the relevant legal framework and data protection policies in advance.

Important considerations:
  • Personal, confidential or sensitive data must not be entered into AI tools (this includes, for example, health data, unpublished research results, contractual data, patent applications, prototypes and similar information).
  • Servers located outside the European Union may be subject to lower data protection standards than those within the EU or allow access by public authorities and are not GDPR (=General Data Protection Regulation)-compliant.
  • Licensed content, such as e-books or journal articles licensed by Bayreuth University Library, must generally not be uploaded, as these materials are protected by copyright.
  • Text and Data Mining (TDM) is a part of the license agreement for many publishing contracts. If you are planning projects involving TDM, we therefore recommend that you contact the University Library in advance to determine what—and, if applicable, to what extent—TDM is permitted in this context, as well as the technical requirements.
  • Literature made available under open licences, such as Creative Commons licences, may be used in accordance with the relevant licence conditions. In addition to copyright, ethical issues, data protection rights and privacy rights are also relevant when using AI (see also the German-language Creative Commons FAQs and presentation slides by Brehm, Elke (2024)).
  • Check whether suitable AI solutions are provided by your institution. Institutional licence agreements often offer greater security with regard to data protection (GDPR compliance) and copyright requirements. Information on AI solutions available at the University of Bayreuth can be found on the IT Service Centre webpages.

AI and Academic Work

Good Research Practice

AI systems are increasingly being used as supporting tools within the research process. At the same time, the principles of Good Research Practice (GRP) and established academic standards also apply to the use of AI systems. This includes:
  • the traceability and verifiability of AI-generated results,
  • documentation requirements and transparency regarding the use of AI applications in academic work,
  • as well as compliance with research ethics standards.
Responsibility for compliance with Good Research Practice and scientific integrity therefore remains with researchers, even when AI is used within the research process. Researchers should carefully assess the potential opportunities and risks of a chosen AI application before using it.
As guidance for an informed engagement with AI from the perspective of Good Research Practice, the German Research Ombudsman (OWID) recommends the article by Resnik and Hosseini: “The Ethics of Using Artificial Intelligence in Scientific Research: New Guidance Needed for a New Tool”.
In the context of academic publishing, it is also advisable to familiarise yourself at an early stage with the requirements of journals and funding organisations regarding the use of AI (see also the sections “Publisher Policies” and “Use of AI in the Context of Third-Party Funding Applications” on this page).
To support researchers and provide guidance in this rapidly evolving field, OWID has published a collection of frequently asked questions on AI and Good Research Practice. These FAQs contextualise existing recommendations from the perspective of Good Research Practice: Frisch, Katrin (2025). FAQ Artificial Intelligence and Good Research Practice. Version 2. Zenodo.

Opportunities and Limitations of AI Tools for Supporting
Academic Publishing and the Research Process

AI tools are evolving rapidly and can provide targeted support to researchers at various stages of the research cycle. For example, AI tools can assist with the analysis of experimental errors, the understanding of complex content, or in the context of scientific publishing. They are therefore becoming an integral part of academic work. However, they can only serve a supporting function and cannot replace researchers’ scientific judgement or peer review by academic colleagues.
In research practice, both generative AI tools and specialised analytical tools are used. According to a Nature Career Feature article, researchers already use AI for:
  • literature research (subject to the limitations outlined above) and support in preparing systematic reviews,
  • summarising and structuring complex texts,
  • supporting the writing of manuscripts and funding applications,
  • data analysis, statistical evaluation and code generation.
Particularly useful are tools that combine multiple functions, such as text comprehension, logical structuring and the formal evaluation of academic arguments, thereby increasing the efficiency of scholarly work.
In its AI Guide, Georgetown University Library emphasises the responsible and reflective use of AI in research and teaching. AI should be understood solely as a support tool and never as a substitute for independent academic work. Researchers are encouraged to use AI, for example, to:
  • develop stronger research questions,
  • critically revise texts,
  • identify methodological alternatives,
  • improve the transparency and reproducibility of their work.
At the same time, the guide emphasises that AI-generated content must be appropriately disclosed, critically reviewed and properly cited in accordance with the principles of Good Research Practice.
A particularly illustrative example of a specialised AI tool is described in Laborjournal using the example of q.e.d. Science. This tool analyses academic manuscripts, breaks them down into logical statements (“claims”) and identifies potential gaps in argumentation. It provides specific recommendations for text revisions, suggests additional experiments where supporting data are lacking and proposes alternative approaches to data analysis. Many researchers report that q.e.d. Science works more precisely and systematically than a superficial preliminary review by colleagues. At the same time, the limitations of AI, such as the lack of contextual knowledge, feasibility assessment and disciplinary intuition, demonstrate clearly that AI cannot replace peer review conducted by researchers.

Publisher Policies (Selected Examples)

Many publishers have now published overarching AI principles as well as specific guidelines for researchers on the use of AI tools in academic publishing:

Springer Nature

Taylor & Francis

Wiley

Use of AI in the Context of Third-Party Funding Applications

DFG (German Research Foundation)

In March 2026, the DFG published its Guideline on the Use of Artificial Intelligence in the Review Process
(DFG form 4.04 – 03/26). At the same time, the white paper Artificial Intelligence in the Review Process ─ DFG Position and Perspectives was published.
The new guideline came into effect on 16th April 2026. Further information and guidance are available on the DFG website:

European Research Council (ERC)

At the end of March 2026, the ERC published guidelines on the use of AI in the evaluation of grant proposals. The document Guidelines for ERC Panel Members and Remote Reviewers also includes an FAQ section addressing specific questions such as “Can I use an AI tool to generate a first draft of my review, which I then edit and finalise myself?” and “Can I upload a proposal, or parts of it, to online AI tools such as ChatGPT, Claude, Gemini etc. to get assistance?”. The ERC had already published a general position statement on the use of AI in December 2023.

Further Resources and Support

Disclosure of Delegation to Generative AI
The authors declare the use of generative AI in the research and writing process for selected content on this website. According to the GAIDeT taxonomy (2026), parts of the following tasks were delegated to GAI tools (Microsoft CoPilot, ChatGPT-5) under human supervision:
Text generation, Proofreading and editing, Summarizing text, Translation
Responsibility for the content of this website lies entirely with the authors.