What is happening beneath the feed?
By the time a paper reaches your feed, a great deal of sorting has already happened. Here is how papers are found, judged, routed, summarized, and turned into visuals.
Featured answerHow are MedLit infographics made, and how do you keep them accurate?
Early versions of MedLit tested GPT Image 2 and Gemini as final-image renderers. They could make beautiful cards, but occasionally introduced subtle micro-hallucination errors: a number drifted, a label changed, text became corrupted, or a chart was redrawn imprecisely. In medicine, "occasionally subtly wrong" is disqualifying.
The current pipeline separates interpretation from drawing. AI may extract a structured visual specification, but an adaptive deterministic HTML/SVG renderer computes the chart. The layout responds to the study, while the same validated inputs produce the same graphic every time.
That gives us a repeatable artifact whose values, labels, and geometry can be checked against the structured source data before publication.
- 01Ground every valueNumbers, units, confidence intervals, and relative claims must match the available source.
- 02Check the claimHeadline entailment and causal-language guards reject unsupported framing.
- 03Read the render backA no-naked-numbers rule fails any unapproved numeric text in the final document.
- 04Harden the contextP2 checks distinguish absolute values, change from baseline, and endpoint timing.
- 05Repair or omitOverflow, collision, grounding, or context failures trigger repair, fallback, or no image.
What MedLit is doing for you
What is MedLit Discovery?
MedLit Discovery is a ranked medical-literature feed. It helps you find the papers worth opening first, including important work that appears outside the journals and specialties you routinely follow.
The larger ambition is to answer the questions you did not know you had.
How is MedLit different from PubMed?
PubMed is an essential search index: you ask it a question and it returns matching records. MedLit is a discovery and prioritization layer.
It retrieves broadly, classifies papers, evaluates them for different clinical audiences, and builds selective feeds for scanning.
Why not rely on PubMed alerts and social media?
PubMed is indispensable, but it is too large to browse as a daily feed. Alerts are useful for questions you already know to ask, and there are only so many searches a person can maintain.
Social and professional feeds often surface a paper after it has attracted attention, along with someone else's interpretation. MedLit watches across journals and disciplines, then brings forward papers that may matter before you know the exact search terms to use.
How does a paper enter MedLit?
Candidate papers are retrieved on a recurring schedule, classified by subject and study characteristics, scored in specialty context, and routed into the feeds where they belong. Summaries and visual abstracts are added only after the paper clears the relevant rules.
What determines whether a paper appears in a feed?
Relevance is the price of admission. MedLit then considers evidence strength, novelty, interestingness, confidence, freshness, and topic coverage.
A feed is composed, not merely sorted, so one repetitive cluster cannot crowd out everything else worth seeing.
Why can the same paper rank differently across specialties?
Importance is contextual. A paper can be central to one specialty, useful to another, and peripheral to a third.
MedLit scores the paper through each audience's lens instead of pretending one universal rank fits everyone.
Can one paper appear in several feeds?
Yes. Medicine is cross-disciplinary.
A spinal cord stimulation paper may matter to pain, neurology, rehabilitation, and neurosurgery for different reasons. Multiple memberships are a feature, not contamination.
Why does MedLit look beyond specialty journals?
Medicine is increasingly cross-disciplinary. A paper that matters to a pain clinician might appear in Nature Medicine, neurology, rehabilitation, or engineering rather than a pain journal.
No one can monitor every neighboring field. MedLit follows the clinical idea across journal boundaries so your usual reading list does not define your awareness.
What is the difference between relevance and evidence strength?
Relevance asks whether a paper matters to a particular audience. Evidence strength asks how much confidence its design and available data support.
A highly relevant early study can have limited evidence, and a rigorous study can still be marginal to your practice.
Why are some feeds smaller than others?
Selectivity is deliberate. Some fields produce more strong, relevant work than others, and a feed should be allowed to stay short.
An honest small feed is better than one padded with weak matches.
What the system knows, and what it refuses to fake
Are MedLit summaries generated by AI?
MedLit uses AI-assisted classification, scoring, and summarization alongside deterministic rules and validation. Unless explicitly stated otherwise, summaries are grounded in the abstract and bibliographic metadata.
The source paper remains the authority.
Does MedLit read the full text of every paper?
No. Full text is not available for every publication. MedLit does not claim access to methods, appendices, or results that were absent from the available source material, and it distinguishes abstract-grounded work from any full-text workflow.
Why doesn't every paper have a visual abstract?
Not every paper benefits from a chart, and not every abstract contains enough grounded detail to support one safely. MedLit would rather omit an image than invent a number, overstate a conclusion, or publish a polished but misleading graphic.
Is MedLit exhaustive?
No. MedLit is a discovery and prioritization product, not a systematic review, clinical guideline, or exhaustive evidence search. Use primary papers and formal review methods when completeness is required.
Is MedLit medical advice?
No. MedLit helps users discover and assess literature. It does not provide patient-specific recommendations or replace clinical judgment, local policy, specialist consultation, or direct review of the evidence.
What to expect while we learn
Who is the pilot for?
The pilot is designed primarily for physicians, advanced practice clinicians, students, and researchers. Other curious readers may explore it as an informational discovery tool.
How often is MedLit updated?
New literature is ingested on a near-daily schedule. Classification, scoring, summaries, membership, and visual enrichment happen in stages, so a new paper may appear before every optional enrichment is complete.
What should I report during the pilot?
Tell us when a paper feels irrelevant, an important paper is missing, a summary or visual seems misleading, the ranking feels wrong, or the workflow creates friction. Specific examples reveal which rule or product decision needs attention.
See how the system works.
Follow the path from the literature firehose to a ranked, evidence-aware feed.