OpenEvidence is a tool designed for the rapid search, synthesis, and delivery of up-to-date medical information within the clinical workflow. It displays the information sources on which its answers are based, allowing users to verify their accuracy. Users can also ask follow-up questions, either based on the tool’s suggestions or of their own choosing. In addition, it offers a “Deep Consult” option for more in-depth searching, as well as the ability to upload files.

Information sources: Articles from high-impact journals indexed in PubMed (approximately 30% of PubMed content, according to company data—details not available on the website) and U.S. clinical guidelines (specific sources of the guidelines are not disclosed).

Special features: The company signs content agreements with publishers and medical associations, giving it access to the full text of articles behind paywalls and thereby improving the accuracy of its answers. Dotflows allows users to save fixed prompts throughout the workflow. The tool can also provide practice questions for the USMLE exam (via the side menu), clinical trials data  and even games (via the discovery option). In addition, it supports patient case management and integration with institutional systems, and is HIPAA compliant.

Apps: iPhone, Android  

Usage Costs: The system implements a restricted access model for the general public, while offering unrestricted usage to healthcare professionals. Free access can be obtained through authentication with various license numbers, including Israeli MOH numbers. Medical students may register for free by uploading a photo of their student card (via the “other” option. In cases where this identification method is not successful, users are directed to contact the provided email address for further assistance: help@openevidence.com 

Link: https://www.openevidence.com/

Additional information: https://www.openevidence.com/user-guide

Update regarding the tool’s models (3.09.2026)

Last update: 19.09.26

PubMed has recently introduced a new experimental AI-based tool called Linked Discoveries. A link to the tool can be found on (almost) every article record in the database. It essentially performs a semantic search for articles similar to the selected source article. Up to 200 articles can be displayed at a time, and the results can be filtered by various parameters and viewed either by similarity to the source article or along a timeline. At this stage, the results can only be exported as a CSV file.

Additional information: Link

Last update: 27.9.2026

UpToDate is a clinical decision-support resource based on continuously updated content written and reviewed by medical experts.

AI-enhanced search: Users can submit clinical questions in natural language. In addition to traditional search results, the system retrieves relevant  passages from UpToDate topics rather than generating a new summary. Each passage links to its location in the complete topic and to the supporting references. This feature is available as part of UpToDate Enterprise Edition.

UpToDate Expert AI: A separate product that provides a conversational interface and supports follow-up questions. It generates context-sensitive answers grounded in UpToDate content, states the assumptions used in constructing the response, and provides inline links connecting individual claims to supporting UpToDate topics. Responses may include clinical reasoning, comparison tables, differential diagnoses, and drug and dosage information also grounded in UpToDate Lexidrug. Voice input is supported.

Information sources: Expert-authored and reviewed UpToDate clinical content, supplemented by UpToDate Lexidrug information for relevant medication questions. The system does not independently search the entire scientific literature and should not be considered a substitute for bibliographic database searching or a literature review.

Cost and availability: Both standard and AI-powered search are available through the institutional subscriptions of Rambam and the Technion. ExpertAI is currently available only for a trial period through the Technion.

Additional information about UpToDate Expert AI
Information about AI-enhanced search

Last updated: 19.9.26

EvidenceHunt, developed in the Netherlands, offers multiple modes of interaction. Users can pose questions in natural language via a chat interface (including support for Hebrew/Arabic input), conduct searches using PICO framework fields or disciplinary categories, or utilize PubMed’s algorithmic search methodology in the Search interface. The chat interface generates summarized information with accompanying references in response to queries, while also facilitating follow-up questions. The search results can be filtered by journal ranking (it appear as Impact Factor but the tool does’nt use the original IF by Clarivate), year, type of study or sample size. Focus on articles related to cost-effectiveness  is only available in the search interface based on PICO. Filters must be selected before the query is used in the chat.

Information sources: PubMed, synced daily. Since April 1 2026 coveres also data from Semantic Scholar and OpenAlex. Adding personal sources is available for paid users only.

Special features: CSV or RIS files of the selected articles can be downloaded (RIS only in the paid version).
Each result includes tags representing PICO characteristics.
After getting an initial answer in the chat, it is possible to choose specific sources and run the response generation again (available only in pro searches).
It is also possible to generate tables from the results and download them as a file. Answer formatting into different styles is available in the paid version only. Alerts can be created for new articles matching queries in the tool. This feature is called Evidence Signals.

Usage costs: Basic use of the chat is free (limited use of advanced model queries per week  for non-paying users). There is an option for a paid premium account (personal, teams or institutional). In the premium account, advanced model is activated for more accurate and comprehensive summaries. The search feature is free of charge.

Link: https://evidencehunt.com/

Additional video tutorials: https://www.youtube.com/@EvidenceHunt

Last update: 19.09.26

IncitefulMed is a second tool developed by Inciteful. The original tool is based on citation network analysis and is designed for interdisciplinary literature discovery. IncitefulMed is intended for searching and summarizing medical information.

At present, the primary focus of the tool is on patients; however, it is also suitable for healthcare professionals and caregivers (e.g., family members). The interface includes dedicated tabs for each target population as described above, although currently there is no difference in the responses generated based on the user type selected.

Search Mechanism:
Work within the tool begins by entering a free-text question into the chat interface. Queries may be submitted in multiple languages. The tool decomposes the question into topic-based sub-questions and identifies relevant seed articles for each through a combination of semantic search and keyword matching. It then expands the search using citation graph analysis (as implemented in the original Inciteful tool) for each sub-question.

After retrieving the relevant articles (abstracts or full text, when publicly accessible), the content is processed by a large language model (the specific model may vary over time) to generate a summary of the information. Users may also upload files to provide additional context for their query.

Sources of Information:
The tool synchronizes daily with PubMed. It also incorporates metadata from OpenAlex. If the query includes a request for clinical trial information, an additional search is conducted in ClinicalTrials.gov. Integration of drug-related information from RxNorm is planned.

Special Features:
The tool suggests follow-up questions, including recommendations for issues that may be important to discuss with a physician. Users can create dedicated project/patient folders for different individuals or topics, within which all relevant questions are stored.  Memory is maintained at the project/patient level, so the project folders retains memory of files uploaded across all threads within that folder. Therefore, if uploaded files are not relevant to all threads, it is advisable to create a separate project/patient folder.

Usage costs: Limited free account.

Link: https://incitefulmed.com/

Additional Information: https://incitefulmed.substack.com/

Last update: 21.09.26

A dedicated version of ChatGPT for healthcare professionals. It is currently available to physicians, physician assistants, nurses, and pharmacists in the United States, and is expected to become available in additional locations in the future. Among other features, the system offers access to advanced models; reusable Skills for clinical workflows, such as referral letters, prior authorizations, and patient instructions; search based on peer-reviewed medical sources; in-depth literature reviews across medical journals, with the ability to define trusted sources and generate a cited report; optional support for HIPAA compliance; and account security and privacy protections, including not using conversations to train models and safeguards such as multi-factor authentication.

Usage cost: free

Additional information: https://openai.com/index/making-chatgpt-better-for-clinicians/

Last update: 3.5.26

A National Library of Medicine (NLM) tool that enables users to formulate search queries in the form of a sentence or paragraph and retrieve sentences or paragraphs with similar content from scholarly articles. The first version of the tool was released in 2018, with version 2 launched in 2025. It is designed to support literature reviews, fact-checking, and targeted information extraction.

The search process is conducted in two stages: an initial filter based on keywords, followed by ranking of the results according to semantic similarity. For sentences, the minimum similarity threshold is 60%; if no matches are identified, the threshold is reduced to 30%. If no matches are found even at this level, the system does not return results. For paragraphs, the minimum similarity threshold is 30%, which may be lowered to 10% if necessary.

Information source: Abstracts from PubMed (over 38 million) and full-text articles from PubMed Central (over 6 million). The database is updated weekly with newly published articles.

Special features: Search results can be filtered according to the section of the article in which the sentence appears (e.g., abstract, introduction, conclusions, and so forth). Entities within the text—such as diseases, drugs, genes, and others—can be highlighted.

Limitations: The tool does not filter out duplicate articles in the results, and the results can only be downloaded in TSV format.

Usage costs: Free

Link: https://www.ncbi.nlm.nih.gov/research/litsense2/

Additional Information:

https://academic.oup.com/nar/article/53/W1/W361/8133630?login=false

Last update: 3.11.25

Elicit is a search engine that utilizes semantic search to extract information from articles. Available since 2021. 

Information sources: Semantic Scholar, PubMed – weekly update. Paid accounts also include information from ClinicalTrials.gov. Alerts (available only with a paid account) also make use of data from OpenAlex.

Special Features:

Elicit includes several built-in tools for literature review and data analysis. The tool find papers allows users to search for academic papers and extract specific information from them in table format. After formulating a research question, the tool suggests possible improvements to the query. Once the search is run, in addition to the results table, Elicit displays a chat tab for interacting with the papers. 

Users can also upload PDF files to a personal library for analysis and “conversation.” The library supports organizing files into collections and detecting duplicates.

Elicit also includes a reserch agent. The agent has access to additional information sources beyond papers, including general web search, information on clinical trials, and more. It also allows the creation of customized outputs. Presentations and visuals creation ia available in the paid plans. Additional information about the agent is available at the following link.

The Research Report feature that generates a report based on a research question. The report covers a limited number of articles and cannot be edited in the free plan (but can be downloaded as a PDF). In the paid version, users can generate extended reports, edit them, and add chosen papers. The number of reports that can be created per month in the free plan is limited.

Search results can be exported as CSV, RIS, or BibTeX files — available only in the paid plan.
Elicit can sync with Zotero for uploading articles only (configured via account settings, not a live sync).
It also supports article screening for systematic reviews (paid plan only) and offers email alerts for new papers based on past searches  in the paid version.

Elicit now supports connecting via MCP or API within ChatGPT and Claude. This connection,  together with ollaborative work and the use of skills, are available only with a paid account.

Usage cost: A free  account allows unlimited article searches, as well as PDF analysis of up to 20 files per month. Research reports include analysis of a limited number of articles and cannot be edited.  The dedicated tool for systematic reviews is available only with a paid plan, as are data export and alerts for new articles. 

Link: https://elicit.com/

Research reports video tutorial

Last update: 15.09.2026

ORKG Ask is a tool developed by The Open Research Knowledge Graph (ORKG) project. It is designed for searching and summarizing scientific information based on text analysis. To access full functionality and save search history, users are required to create an account.

Information source: a dataset sourced from the British multidisciplinary open-access article repository, CORE. As of today, the data is updated annually, but there are plans to transition to continuous updates in the near future. The tool’s statistics page provides the latest coverage information.

Special features: The information summary is generated by the Mistral Instruct 7B v0.2 chatbot. After submitting a question in the chat, a brief summary of the topic is provided, along with a table containing various characteristics of the articles, which can be edited as needed. A CSV file with the table’s information can be downloaded. By default, five articles are shown in the results, but more can be added by clicking the “Load more” button at the bottom of the page.

Disadvantage: It is not suitable for searching for very recent information, as the current update is delayed by approximately one year.

Usage costs: Free

Link: https://ask.orkg.org/

Last update: 3.11.25

Consensus is an AI-based search engine where users can enter a free-text question in the search box. It uses semantic search along side keyword search  and citation search, and also supports Boolean operators (AND/OR/NOT). The tool offers features such as information summarization, outline creation, presentation content, and summary tables. There are three modes: Quick (free, up to 10 articles, no text generation, available from the side menu), PRO (up to 20 articles, 10 monthly credits for free users), and DEEP (up to 50 articles, 3 free credits per month or 100 for paying users). After performing a search, results can be displayed in a table view with columns such as Population, Intervention, and others. In the free plan, only the first three articles are shown in the table.

In the Pro and Deep modes the full text of the paper is analysed if available. Consensus now has also many content agreements with publishers which gives them access to the full text of papers behind paywalls.  When writing a prompt in DEEP mode, it is possible to apply filters within the prompt—for example, to request articles published up to a certain year, from a specific journal, and so on. In PRO mode, users can also search by author name to receive a summary of their work and relevant references. In Pro mode and in deep search, it is also possible to obtain various visualizations, such as citation networks, research gaps, and strength-of-evidence rankings.

Support and focus on medical research:
The search supports MeSH terms and allows users to limit results to medical journals only by selecting the medical corpus option (this is the list of medical journals included, in addition to ~50,000 clinical guidelines, in total of about 8 M papers).

Integration with large language models: Consensus can now be integrated with large language models through a variety of tools and methods. Full details, including ready-to-use Skills for a range of workflows, are available here.

Information source and update frequency:

Semantic Scholar, updates monthly according to Semantic Scholar’s update frequency.

Special Features:  It is possible to locate a specific article by its DOI number by pasting the number into the search box.  Results can be downloaded in CSV or RIS format. Filters include evidence quality, sample size, SJR ranking, and research field. Recently, filters for study duration and country have been added. The tool allows searches in multiple languages, including Hebrew and Arabic, and provides search history. In the personal profile, users can select their preferred language for receiving information, which currently includes Arabic and Russian, though Hebrew is not yet supported. Integration with  Zotero is available.  LibKey integration is also available in Cosensus (not applicable for the full text analysis).  Additionally, the platform now enables the generation of citation network graphs.

Usage cost: A  free account includes unlimited quick searches and monthly renewable credits for PRO and DEEP modes.

A link to the tool: https://consensus.app/search/

A Link to Video tutorials

Last update: 21.09.26

ScienceOS  is an AI search engine based on semantic search, developed in Germany.  The tool includes a Chat-based interface, and for each query in natural Language the users receive a well-formulated answer supported by citations from articles within the text. SCImago rank is available for each paper. To obtain a list of articles or their citation network, users can click the appropriate button below the response.

It also enables the use and creation of AI Actions, saved prompts for various recurring tasks such as creating presentations and summarizing articles. There are several ready-made prompts, and users can create additional custom prompts.

Information source and update frequency: Semantic Scholar, updated every two weeks to once a month, depending on the update frequency of Semantic Scholar.

Special features: certain answers or chats can be saved by clicking the star icon, and then it can also be annotated. Users can create tables and diagrams and upload up to 4000 PDF files (or RIS/BIBTEX files, or direct from Zotero) for further work. Deep research option is available in the paid version only. Citation networks can be built in the library too. Shareable projects are also available. 

Usage costs: A chat based on less-advanced model is available for free, with a maximum of four questions per conversation for free users. More advanced models are available for paid users only. Chat with PDFs is available for paid users only.

Link: https://scienceos.ai/

Link to video tutorials

Last update: 21.04.26

The Scite tool began as a platform that characterizes the sentiment of citations in articles as neutral/supporting/contradicting the cited paper. It later evolved into a tool that also allows article searches and includes a chatbot to assist with text writing. The searches performed by the tool are carried out by converting the user’s questions into Boolean queries.

Main information sources: Not based on a specific database, but on articles for which the company has access to the full text for information mining. Scite has signed agreements with various publishers to obtain access.

Special features: The tool offers a free browser extension that provides information from anywhere on the web, including the number of citations an article has, their sentiment, and whether the article has been retracted. The search results can be formatted as a table.  Additionally, there is a Zotero plugin and a tool for checking the quality of references in an article. Available now also via an app within the ChatGPT interface and/or MCP connection (for subscribers). Integrates with LibKey for full text instetutional access.

Cost: Paid (except for the browser extension). The Pro version, available for an additional fee, enables access to information on patents, grants, and clinical trials through the MCP connection.

Tool link: https://scite.ai/home

Tutorial videos: link

Last update: 09.5.26

Keenious was created by a company based in Norway. operates by analyzing the text of an article uploaded to the tool via a PDF file or text paste. The recommendation is based on term similarity, semantic analysis (based on meaning) of the text, the number of citations, and the publication date (newer and more highly cited articles are prioritized). Highlighting a specific part of the text gives it higher priority in the search algorithm. Highlighting a specific part of the results text enables an option of explenation or new search.

At the top of the results page, there is a visual map that organizes the search results by topic. Each point on the map represents an article; its size reflects the number of citations, and the intensity of its color reflects the article’s relevance to the topic. The map can be adjusted using various parameters and downloaded as an image file. Additional information about the map can be found here.

Information source: OpenAlex

Special features: It offers add-ons for Word and Google Docs and ensures user privacy. Integrates with LibKey for full text instetutional access. Results export is available via different citation styles and file formats.

Usage costs: A free basic account allows  5 AI responses per conversation, 3 conversations per day and
3 MB max file upload size.

Link: https://keenious.com/?landing

Additional information: Video tutorial

Last update: 19.09.26

Undermind.ai is a tool for searching and analyzing scientific information based on deep search technology.

Information source: Semantic Scholar.

Special features: Classic interface – the search is carried out using Deep Search technology in multiple stages and takes several minutes. Before the search begins, the tool’s  agent assists in building a precise search query. The results are organized into categories by topic, timeline, foundational articles in the field, and articles developed based on them. After the search, it is possible to interact with the results via chat, including ready-made prompts for common questions. Search results can be shared and downloaded as a PDF file. RIS or BibTeX files can only be downloaded with a paid account. Email alerts about new articles matching the queries are sent once a month by default and can be disabled if needed. The classic interface can be accessed from the side menu by clicking Manage Workspaces.

The new interface, called Workspaces, includes the ability to search for external articles, work with PDF files stored in the workspace – including interpreting images and graphs, generate customized reports, and more.

MCP integration enables Undermind to connect with Claude, ChatGPT, and other AI tools.

Usage costs: The Pro plan is available at no cost during the first month. The free plan allows for up to five searches per month.. There is an option for a paid personal or institutional account. Creating an account is optional with an institutional email only.

Link: https://www.undermind.ai/home/

Last update: 20.9.26

ASTA is a research platform developed by the Allen Institute for AI that enables to find papers, generate reports and analyze data. The tool is based on a corpus of more than 12 million open-access articles and over 108 million abstracts in the fields of medicine, life sciences, and computer science. Its search mechanism is agentic (Deep research), allowing it to retrieve even relatively niche articles that might not appear in less in-depth searches. If further depth is required, users can type “Work Harder” in the chat window after the search is complete.

A new experimental tool called AutoDiscovery is a platform that enables the discovery of findings and the testing of hypotheses in research data. Users upload a dataset to the tool, add relevant information about the file, and define the number of hypotheses they would like to generate (it is recommended to start with around 10 in the initial stage). Link to AutoDiscovery ,Video tutorial

Another new experimental tool, enables interaction with a specific article, including its images, tables, and text. The tool is called Paper+Figure QA and accepts open access articles via links from the Semantic Scholar database.

At the start, a pop-up window appears where users can choose whether to allow the tool to train on their queries. It is also possible to register for a personal account with Google credentials to save searches.

Usage costs: Free

Link: https://asta.allen.ai/chat

Link to further information

Last update: 22.04.26

Semantic Scholar is a multidisciplinary database where searches are based on the semantic analysis of the texts of the articles it covers. It was created by the Allen Institute for AI.

Information source: It contains over 200 million articles across all fields of science, including those from PubMed.

Special features: Semantic Scholar displays citation information for articles, provides a summary sentence (TLDR) for each article, allows users to browse information by topics, and enables the creation of a personal account to receive tailored recommendations for articles based on those saved. It also serves as a source of information for many other tools.

Usage costs: Free

Link:https://www.semanticscholar.org/