AI Tools from JSTOR
Semantic Results
"Semantic results" uses AI to interpret a natural language query to find content conceptually related to the query.
"Semantic Results" is available in: JSTOR JSTOR 19th Century British Pamphlets and JSTOR Struggles for Freedom: Southern Africa
The semantic results tool is available on the results screen after an initial standard search, just below the search box.
- On the results screen you can enter a natural language query, and then click on the "semantic results" button.
More information: Searching — Keyword versus Semantic Results
AI Research Tool
The "AI Research Tool" identifies the main points of a document relevant to the query, and also provides a document summary if an author abstract is unavailable.
"AI Research Tool" is available in: JSTOR JSTOR 19th Century British Pamphlets and JSTOR Struggles for Freedom: Southern Africa
- From the results, screen click on a document title to bring up the content/full-text page. On this page the AI Research Tool toggle is available.
- if it is not already "ON" then click the toggle to generate the main relevant points.
Below the generated text are buttons with additional options:
- What is this about? — generates a summary of the document
- Show related content — identifies other documents related to the one you are viewing
- Recommended topics — suggests new searches to expand on the topic
More Information: JSTOR's AI Research Tool
Under the Hood
What powers JSTOR's AI tools?
JSTOR’s AI tools—including the AI Research Tool and semantic results—are powered by large language models such as OpenAI’s GPT-4o-mini, Anthropic models (accessed via Amazon Bedrock), and the open-source all-MiniLM-L6-v2 sentence transformer. The system uses Retrieval-Augmented Generation (RAG), grounding every response in content retrieved from specific JSTOR documents, never from the open web. These models generate summaries, answer questions about open documents, and suggest topics or related JSTOR content, while ensuring every output is directly linked to and verifiable from JSTOR’s trusted, curated collection.
AI-assisted, human-verified text