In partnership with
The National Institutes of Health launched Linked Discoveries on September 24 as an experimental resource inside PubMed.
The tool starts with a PubMed record and builds a neighborhood of related publications. NIH says more than 29 million PubMed publications were represented at launch.
200+ Proven Ways to Make Money With AI in 2026
The next wave of millionaires will be people who figured out how to make AI work for them.
The window to get ahead is still open. But not for long.
Here are 200+ proven ways to make money with AI in 2026.
Sign up for Superhuman AI, the free daily newsletter read by 1M+ professionals, and get instant access to all 200+ ways to profit from AI this year.
Linked Discoveries is designed for a problem that researchers face after finding an individual study. The relevant evidence may be spread across reviews, related experiments, replication studies, citation records, and later corrections.
A keyword search can retrieve articles without showing how their findings connect. The new tool presents those relationships through graph and timeline views. It also exposes context that can change how a paper is read.
The visible context can include citation links, reviews, retractions, expressions of concern, and NIH funded publications. Users can narrow a neighborhood using conditions, genes, and chemicals linked through NLM databases.
The technical method is narrower than a general purpose chatbot. The NLM user guide says Linked Discoveries uses a version of BiomedBERT, a language model trained on biomedical and clinical text.
The model compares the title, abstract, and keywords of a seed article with other PubMed records. It uses semantic similarity to build a related set. The tool does not use full article text for that calculation.
That design has a practical consequence. A relationship shown by the system reflects available metadata and abstracts, not a full reading of every paper. A related label can help find evidence, but it does not establish that two studies agree.
The tool also does not judge research quality. NIH states that it cannot determine whether a finding has been successfully replicated. It organizes material so researchers can perform that assessment themselves.
The inclusion rules are specific. Linked Discoveries supports PubMed records for original research, reviews, and related evidence when an abstract is available. Biographies, interviews, and preprint records are excluded as seed articles.
Built for Product Teams moving at AI Speed.
Your teams are moving fast, burning tokens, and shipping more than ever.
But more output doesn’t mean more impact.
Jira Product Discovery brings your ideas, customer insights, and business context together so product teams can weigh the evidence, make the tradeoffs, and decide what’s actually worth building. Then connect those decisions directly to delivery in Jira, so everyone knows what you’re building and why.
Jira Product Discovery. For better product decisions in the AI era.
New PubMed records enter the system daily. The guide says condition, gene, and chemical tags can appear later because separate indexing processes supply them. It notes that indexing can take up to a month in some cases.
Citation links also have limits. The guide says cited by and cites relationships may not be complete when PubMed does not have the needed data. A missing link is therefore not proof that no later work exists.
Those limits matter for replication research. A neighborhood can make less prominent or contradictory findings easier to discover, but visibility is not the same as methodological agreement.
The launch is part of an NIH initiative focused on replication and reproducibility. Its purpose is to improve evidence discovery, not to replace peer review, statistical analysis, or expert reading.
For biomedical researchers, the main change is workflow. A paper can become the starting point for a structured search through nearby studies, topic labels, citation relationships, and publication updates.
The tool can also reveal why a simple citation count is an incomplete guide. A highly cited paper may sit beside reviews, corrections, or dissenting results that alter its context.
The model introduces another question. Semantic similarity can identify related language and concepts, but it cannot decide whether a study’s population, method, outcome, or bias matches another study.
Linked Discoveries is therefore best understood as an evidence-discovery layer. It reduces the work of locating possible connections while leaving interpretation and replication claims with the researcher.
NLM describes the September release as its first public version and says it will continue testing and refining the resource. The most important evidence will come from how users check its neighborhoods against the underlying PubMed records.
Learn AI in 5 minutes a day
You don't have to scroll every AI thread, track every new tool, or watch every demo.
The Rundown AI breaks it all down for you — the latest AI news, tools, and tutorials in one free 5-minute email every morning.
Trusted by 2M+ professionals at Apple, Google, and NASA.
