University Library: Catalog plus SehKon - Online catalog for the visually impaired Service for the blind and visually impaired of the Dortmund University Library To path indicator Subpages of “Search & Media“ To navigation by target groups To navigation by topic To quick access To footer with other services
To content
Logo der UB Dortmund
Students in summer on north campus, red gears in background © Roland Baege​/​TU Dortmund
AI tools for scientific literature research

Questions and answers

  • Conventional literature databases query structured, systematized data using, among other things, defined keywords/title keywords/abstracts. In order to find relevant publications, users must formulate precise queries with exact search terms and make manual adjustments (e.g. using Boolean operators).
  • AI-supported search tools can mainly process natural language and thus create semantic links between different sources even without keywords assigned in the metadata.
  • Many tools create new content, e.g. short summaries for publications in the hit list (often marked with the note "TLDR"). Some AI tools can automatically "learn" from the user's previous search queries and recommend personalized publications.
  • AI tools do not currently replace database searches, but they can supplement them.

Note: Conventional databases are also increasingly integrating AI-supported functions or developing their own AI tools, e.g. Web of Science (Research Assistant) and Scopus (Scopus AI).

  • Semantic search: By using LLMs, the context of the search query is "understood". Results are displayed that are not only based on the exact search terms, but also include the context of meaning and synonyms - unlike in traditional databases. With some tools, search queries can also be formulated as research questions.
  • Focus on the scientific field: The AI tools presented are geared towards the scientific field. The search query is therefore placed in a scientific context - even without extensive prompting and without the use of specific technical terms.
  • Additional functions: Some of the AI tools offer additional functions (e.g. short summaries, chat with paper, copilot) that simplify the review of literature and can help with the understanding of the researched texts.
  • Graphical presentation: Many tools prepare the search results graphically (e.g. in network diagrams). This can accommodate different types of learners and support citation analysis.
  • Data protection and copyright: The user's input is sometimes used to train the AI. In addition, the operators of the tools are often granted extensive rights of use. Therefore, no sensitive, personal data should be entered and the uploading of copyrighted works should be avoided.
  • Scope of the data basis: The tools usually only have a limited data source (e.g. only open access publications). In some cases, publications from important publishers are missing. Due to the underlying data source, not every tool is equally suitable for every subject. The data basis should therefore be checked and, above all, not just one tool should be used.
  • Quality of the results: Generative AI provides the statistically most probable answer in certain contexts. The models are therefore language-based, but not fact- or knowledge-based. This can lead to "hallucinations". In addition, the quality of the results depends on the training data (topicality, stereotypes/bias, etc.). These possible effects on search results and scientific neutrality should be taken into account. The results should always be checked.

Note: The challenges regarding ethics and sustainability should also not be ignored. The high energy and resource consumption of AI is just one example.

  • Limitations: AI chatbots such as ChatGPT/Campus AI, Gemini, Perplexity AI & Co. are currently not geared towards literature research and therefore often do not provide any (scientific) sources. In some cases, bibliographic information is "invented". The original source should always be checked: Does it even exist? Has the chatbot reproduced the content correctly?
  • Possibilities: Chatbots can help you prepare for a literature search by suggesting relevant search terms and their synonyms. The development of the deep research functions of some chatbots that are specifically geared towards research also remains to be observed.
  • When using the research tools to find literature, these do not have to be cited. This process is comparable to an internet or database search.
  • Direct use / adoption of AI-generated content should be labeled. In order to be able to prove the content, the chat history/prompt etc. should also be documented.
  • There are currently no standards regarding the citation of AI-generated content at TU Dortmund University. The use and citation of AI should therefore be discussed with the examiners beforehand.
  • Course dates on request: We offer introductory courses on AI-supported literature research for groups, chairs or seminars of 5 or more people. We are happy to tailor the content to your wishes and needs. Please contact us to arrange a date!
To top of page