DeepDyve's AI-Powered Research Assistant

DeepDyve now has a powerful AI feature to help you understand and work with articles: Research Assistant. This article will help you understand how it works and how to use it.

Accessing Research Assistant

To launch Research Assistant, open up an article for reading. Click on "Research Assistant" in the white document bar:

In full text mode, click on the AI icon in the top black tool bar:

Using Research Assistant

Ask questions about the article you are looking at by typing them in the chat window. For best results, keep your questions short and targeted. Some examples you could try:

Summarize key points of the article

• Describe the key results from this article and their significance.
• What is the organism studied in this paper?
• Summarize the adverse events mentioned in this article
• What is the date range of data collected?

Create lists or tables of data in the article

• Please list the software mentioned in this article.
• Create a table of DNA primers found in this paper. Column 1 is the label. Column 2 is the nucleotide sequence with 5' and 3' specified.

After an answer is provided, you can continue to ask for refinements and further exploration. To begin a new conversation, click "Reset Conversation"

Using Research Assistant on multiple documents

In addition to querying the article you're looking at, you can choose to add up to four items from that article's references. To use this feature, click on "Include References":

Select up to four references and the click "Confirm Selection".

To see which articles are included, click the arrow. To select a different set of articles, click "Change References".

What articles does Research Assistant work with?

Research Assistant currently works with tens of millions of full-text papers from our OA and LitStream collections, as well as PDFs you upload to your My Papers repository, and for Enterprise customers, PDFs uploaded to the Research Archive.

Please note that not every publisher in our OA and LitStream collections have included all their works, so you may encounter a paper where this feature is unavailable. But rest assured that we are continuing to work with our partners to expand this feature to every paper in DeepDyve.

How confidential are my queries?

DeepDyve has developed its own AI features but also relies on the large language model (LLM) from OpenAI, Anthropic and other potential AI vendors, via their Enterprise API license. Under the terms of these licenses, the LLMs will neither store any data passed to them, nor use that data to train their models. Further, DeepDyve is not including any client or user-specific information in the API to ensure even greater confidentiality