AI-powered academic literature assistant

Submit a topic.
Get a review you can verify.

ScoutLit retrieves papers from arXiv and Semantic Scholar, synthesizes a structured literature review with inline citations, and shows you exactly what the AI based each claim on.

Join research study
ScoutLit is currently in development. We're interviewing researchers to make sure we're building the right thing.

From topic to traceable review

No manual database searching. No copying abstracts into a chatbot. ScoutLit runs the full retrieval-to-synthesis pipeline and hands you something you can actually work with.

  1. 01
    Search
    Any research question or keyword. ScoutLit handles the search strategy.
  2. 02
    Retrieve
    Concurrent arXiv + Semantic Scholar fetch, deduplicated by DOI and title.
  3. 03
    Classify
    Each paper labeled Seminal, Foundational, Recent, or General — no LLM needed.
  4. 04
    Synthesize
    Structured review with inline [Author et al., Year] citations and research gaps.
  5. 05
    Evaluate
    Groundedness scored 0–100. Unsupported sentences flagged explicitly.
  6. 06
    Export
    Download the full review as Markdown. Refine with follow-up questions.
Literature Review
Research Gaps
Sources

The transformer architecture, introduced by [Vaswani et al., 2017], replaced recurrent networks with self-attention, enabling parallelization across sequence positions. Scaled dot-product attention computes compatibility between query and key vectors [Vaswani et al., 2017], producing weighted sums of value representations [Brown et al., 2020].

⚠ Not directly supported by retrieved sources Transformers have become the dominant paradigm across nearly all domains of modern machine learning.
Example groundedness score 87%
Illustrative example — not a published benchmark.

Know what the AI based it on

Most AI writing tools give you output without attribution. ScoutLit is built around the opposite: every claim in the review links back to a specific paper, a groundedness score tells you what fraction of sentences are supported, and flagged sentences show exactly where the model went beyond the source material.

This isn't just a transparency feature. It's how you decide whether to trust the review.

  • Every claim cites a specific retrieved paper
  • Unsupported sentences flagged, not hidden
  • 0–100 groundedness score per review

Built for people who work with research literature

Academic researchers

Graduate students & faculty

Conducting literature reviews for dissertations, grant applications, or research area exploration. Needs traceable citations, not plausible-sounding summaries.

PhD students Postdoctoral researchers Faculty Systematic reviewers
Research support

Librarians & research teams

Research librarians, coordinators, and evidence synthesis teams who help researchers navigate and synthesize academic literature efficiently.

Research librarians Research coordinators Evidence synthesis teams
R&D professionals

Industry researchers

Technical teams who need to survey the academic literature before building — without a dedicated research librarian or weeks of manual searching.

R&D teams Applied researchers Technical leads
Policy & analysis

Analysts & strategists

Policy analysts, think tanks, and strategy teams who need evidence-grounded literature coverage to inform decisions and public-facing documents.

Policy analysts Think tanks Strategy teams

Research with confidence

Help us build this right.

We're interviewing researchers and research support professionals to understand how you approach literature review — what tools you use, where you get stuck, and what would actually help. Interviews are 30–45 minutes via video call. Your input directly shapes what ScoutLit becomes.

Sign up for a research interview
  • No product access required
  • Participation is voluntary and unpaid
  • 30–45 minute research interview