Literature Search Using AI
The use of artificial intelligence (AI) can simplify the process of searching scientific literature and highlight connections, but it does not replace the expert assessment of quality, context and interpretation.
The Pros and Cons of AI Tools in Finding Literature
AI tools offer a quick way to start searching and can be a useful complement to traditional search tools such as library catalogues and databases. AI tools have limited ability to assess the relevance of literature to your research topic and can only access licensed academic databases and printed media to a limited extent – yet these are essential to the quality of your work. That is why sound research skills remain indispensable. To develop these skills, you can, amongst other things, attend our training sessions and courses.
If you possess these skills, AI tools can provide effective support with literature searches. AI tools can analyse large amounts of data, highlight correlations and semantically evaluate search queries. The results can often be automatically presented in the form of summaries.
Courses
You can find the latest dates in the events calendar or at the bottom of this page.
Selected AI-Powered Research Tools
If you are considering conducting a literature search using AI, it is advisable to use an AI tool that has been specifically developed for literature searches. General language models (Large Language Models, LLMs), such as OpenAI’s ChatGPT, are not yet suitable for systematic literature searches.
The following overview lists selected AI tools designed to assist students and researchers with literature searches. These AI tools take different approaches to analysing and presenting search results.
- Consensus (partly free, account required)
Search using a research question or keywords - Elicit (partially free, account required)
Search by research question or keywords - Keenious (partly free, account required)
Search using a research question, keywords or an uploaded document - ResearchRabbit (partially free, account required)
Search using a seed paper (a publication you have already found); visually presents connections between publications, authors, etc. - Semantic Scholar (free, no account required)
Search using keywords - Perplexity (free, no account required)
Search using a research question
Lizenzierte Datenbanken mit integrierter KI-Recherche
The databases licensed by SUB Göttingen, such as SciFinder, Web of Science, ProQuest Ebook Central, Statista and beck-online, increasingly feature integrated AI search options. Some databases offer chat functions, summaries of text passages or agent-based searches within the platform.
The advantage of these integrated AI functions is that these services may have access to the full texts in the databases, thereby providing more information than other tools. However, the same applies to these additional features: please always check the results, as these AI systems may also produce inaccurate results.
The Limitations of Literature Research Using AI Tools
Limited Data Set
The data sources used by AI research tools vary and are often not comprehensively documented. Many tools rely on freely available data and open-access content. Access to full-text articles from databases requiring a licence, on the other hand, is usually restricted and depends on the respective providers and licence agreements.
In particular, AI tools often fail to identify printed works and non-English-language literature, which are consequently not taken into account. For an academic literature search, however, all types of sources are relevant, including licensed databases, as many further academic publications can be found there. A literature search using AI cannot replace a traditional search, but can only complement it.
Quality Control Required
AI-powered literature search tools can produce misleading results in the form of so-called ‘hallucinations’, return irrelevant results or omit important information. Similarly, the choice of tool can introduce unintended bias. A thorough review of the results is essential, as you are responsible for the content of the work you submit.
Legal and Data Protection Aspects
The use of AI tools can raise issues relating to copyright and data protection. AI tools often impose extensive terms of use on the data provided by users. This applies both to search queries and prompts submitted and to uploaded documents.
Therefore, do not upload any copyright-protected documents to these tools. Do not enter any personal data belonging to third parties, and consider carefully which of your own data you wish to disclose.
The AI tools offered by the GWDG, such as ChatAI, place particular emphasis on data protection. Please note, however, that data protection is only guaranteed for internal open-source models and not for those that access external services.