Noah AI automates full medical literature review across 25 papers

Blogger @socialwithaayan posted a thread on October 27 testing Noah AI's free literature-review tool, Academic Workshop, on the topic of GLP-1 weight-loss drugs (semaglutide, etc.) in heart failure, walking through the complete workflow from topic selection to finished manuscript.

Confirmed

• Given a research topic, the tool first has users define scope via four questions (study population and scope, evidence time span, length and style, focus areas), offering options and trade-offs among narrative/systematic/scoping review types, plus 4 selectable research scopes
• The search phase covers PubMed and clinical trial databases, compiling 25 records each with PMID, PMCID, and DOI; source journals include Lancet, JACC, NEJM, and JAMA; the key design is that the user reviews and approves the reference list before the AI proceeds
• The output Word file includes a structured abstract and a safety section, with each reference annotated as full-text or abstract-sourced; the manuscript has 9 chapters, every claim cited, plus 25 references numbered in Vancouver format
• The tool auto-generates a key-study comparison table listing each paper's PMID, study population, main findings, and whether data came from full text or abstract—tables the author notes usually take researchers hours to build by hand
• The synthesis step classifies evidence by study design, distills consensus and disagreements across studies, and proactively flags open questions: whether GLP-1 drugs' cardiovascular benefits come from weight loss itself or direct effects on the heart

Real-world constraints

• 19 of the 25 papers lack open-access full text, so without library login credentials any tool can only read abstracts; the tool proactively flags this limitation and offers two options: upload your own PDFs for full-text reading, or proceed with abstracts only
• The author also demonstrated having the AI self-review the manuscript draft, which passed in one go

Why it matters

The author found the synthesis step and the comparison table the biggest time-savers—work that previously took hours of manual effort is done automatically by AI; meanwhile, the design retains human review of the reference list and per-item source and limitation annotations, offering useful reference for medical researchers assessing such tools' usability and boundaries.

2026-10-27 ~ 2026-10-27 · 10 related posts

• [source] Hands-on: Noah AI runs a full 25-paper literature review on GLP-1 drugs in heart failure — socialwithaayan · 2026-10-27
• Noah AI builds a 25-record PubMed evidence map for human approval before writing — socialwithaayan · 2026-10-27
• Using LLM to scope a medical review: model trades off 4 research scopes like a reviewer — socialwithaayan · 2026-10-27
• AI review tool proposes 4 scoping options and flags overinterpretation risk — socialwithaayan · 2026-10-27
• [source] 19 of 25 papers lacked open-access full text — a hard limit for AI review tools — socialwithaayan · 2026-10-27
• The biggest time-saver: sorting evidence by design and mapping consensus vs disputes — socialwithaayan · 2026-10-27
• Auto-generated key-studies table with PMIDs, populations and findings — socialwithaayan · 2026-10-27
• AI builds a side-by-side study comparison table with PMIDs, a task that takes hours by hand — socialwithaayan · 2026-10-27
• AI drafts a 9-section manuscript with 25 Vancouver citations, then self-reviews like a journal referee — socialwithaayan · 2026-10-27
• [source] Noah AI's free tool turns a research topic into a cited, publication-ready literature review — socialwithaayan · 2026-10-27