Our sources
The same descriptions appear when you tap a source name on a result. Open “More detail” for how we use each one.
How an answer is made
- An AI model reads your question, splits it into its main ideas, and picks search terms for each source, including the medical subject terms PubMed uses.
- All the sources are searched at once. Results that mention more of your question are listed first. FDA device records and drug labels are searched only at Hard effort and above, and only when the question is about a device or a drug.
- A second AI model writes the answer using only those results, and cites each one. At Medium effort and above, it then checks the answer against the sources and rewrites it if it finds a mistake or a gap.
How up to date it is
PubMed, ClinicalTrials.gov and CORE are searched at the moment you ask, so a new article shows up as soon as those sites have it. The HCF library is copied from hcfhub.org every 6 hours. Training courses and jobs are copied into our own search index from TRAIN, CME Wizard and job sites.
Reading the full article
Each research result says how you can read it in full:
- Full text via PHDL: staff of state and local health departments can read it through the Public Health Digital Library. How to get access.
- Free full text: anyone can read it, in PubMed Central or from the repository CORE found it in.
- Abstract only: only the summary is public, so the answer is based on the summary alone.
What happens to your question
SmartSearch isn't private. Your question goes to an outside AI service (the AI model) to plan the search and write the answer, and its search terms go to the sources above. It's saved in our search log with the answer you were shown, and it stays in the page address, so it's in your browser history and in any link you share. Leave out patient names, case details and anything confidential.
Public health is public work. Ask the questions you'd be glad to post on social media under your own name, and share the answers just as freely.