What is BioSkepsis

BioSkepsis is an agentic AI research assistant for the life sciences. You give it a single research question; it plans the work, searches and reads the literature, weighs the evidence, and returns a cited brief you can verify line by line.

#What BioSkepsis is

BioSkepsis is not a chatbot that answers from memory and not a search box that hands you a list of links. It is an agent — an autonomous system that carries out an entire research workflow on your behalf, the way a careful researcher would. From one question, the agent decides what to look for, retrieves and reads the primary literature, judges how strong each piece of evidence is, screens sources for reliability, and writes up a grounded answer.

Under the hood, a research run moves through the same disciplined steps a scientist follows:

• Plans the question. It breaks your question into focused sub-questions and works through each one, rather than treating a complex query as a single keyword search.
• Searches the literature. It searches broad life-science literature — 40M+ papers — and weights top-tier, authoritative sources over weaker ones.
• Reads full text. It reads entire papers, not just abstracts, so it can capture methods, sample sizes, caveats, and the context that determines what a result actually means.
• Weighs the evidence. It types each finding by the kind of evidence behind it and weighs supporting studies against contradictory ones.
• Screens for reliability. It blocks retracted papers and hijacked-journal sources and flags corrections before they can reach your conclusions.
• Synthesizes a cited brief. It verifies every claim against the source full text, then writes a brief with a Trust Index score.

Because the agent works from the primary literature and shows its sources, you are never asked to take its word for anything. Every statement traces back to a specific paper and, in most cases, to the exact passage that supports it.

#What a research run produces

A single run does not return a paragraph of prose — it returns a small, structured dossier. Each part is designed to be inspected and challenged, not just read.

#A cited brief

The centerpiece is a synthesized answer to your question, written in plain scientific language. Every claim carries a numbered citation that links to the source it came from, so you can follow any sentence back to the paper — and the passage — behind it.

#Findings typed by evidence

The agent does not treat all evidence as equal. It labels each finding by the kind of study that produced it — for example causal, associational, synthesis, or preclinical — so you can see at a glance whether a claim rests on a randomized trial, a correlation, a review, or a cell-line experiment. Strong and weak evidence never blur together.

#A Trust Index (0–100)

Every brief receives a Trust Index — a single score from 0 to 100 that summarizes how well the answer is grounded in reliable evidence. It reflects checks across facets such as source provenance, how tightly claims are grounded in their citations, the strength of the underlying evidence, and reproducibility. The score is a starting point for your own judgement, not a verdict that replaces it.

#Contradictions & gaps

When studies disagree, the agent surfaces the conflict rather than quietly picking a side. And where the literature is thin or silent, it names the gap — the questions the evidence does not yet answer. These are often the most useful parts of a run, because they point you toward what still needs to be established.

#A research notebook

Alongside the brief, the run keeps a notebook: a transparent record of what the agent searched for, which papers it read, and how it reached its conclusions. The notebook makes a run auditable — you can retrace the reasoning instead of trusting a black box.

#Who it is for

BioSkepsis is built for people who work with biomedical evidence and are held accountable for getting it right:

• Researchers and graduate students scoping a new field, writing a thesis or manuscript, and needing to find and cite the primary literature quickly.
• Scientists in industry tracking targets, mechanisms, and methods across a fast-moving literature they cannot read exhaustively by hand.
• Clinicians and health professionals who want an evidence-typed, cited synthesis rather than an unsourced summary.
• Anyone reading biomedical papers who wants to understand a mechanism and trace claims back to their original sources.

What unites them is a need for answers that are traceable. BioSkepsis is designed for work where the citation matters as much as the conclusion.

#What it is not

Being clear about the boundaries matters as much as describing the capabilities.

• It is not a source of medical advice. BioSkepsis synthesizes published research; it does not diagnose, treat, or make clinical decisions for any individual.
• It is not a replacement for reading the papers. The brief points you to the evidence and helps you weigh it — it does not absolve you of checking the sources that matter to your conclusions.
• It is not infallible. Like any AI system, it can misread a study, over-generalize, or miss relevant work. The Trust Index, evidence types, and notebook exist precisely so you can catch this.
• It is not a general-purpose assistant. It is tuned for the life-science literature, and it is at its best on questions the published evidence can actually address.