EvidenceScope

Share your feedback

EvidenceScope is a transparent literature search, screening, and review tool for clinicians, researchers, and academics — and its core works entirely without AI, so you stay in control of the evidence and avoid AI hallucinations. Search literature (an AI query builder or a manual Boolean/MeSH search), refine with MeSH terms, filter by journal impact (OpenAlex, SCImago) and citations (NIH iCite), and screen studies yourself — every result traceable to a real article. AI is an optional assist: it can draft a narrative review with a PRISMA 2020 flow diagram and Cochrane risk of bias, but it only ever cites the papers you selected — no invented references. Save any search to re-run later or feed Your digest — a personalized email of new papers. Two more tools sit alongside the review pipeline: Ask literature answers a clinical question with grounded, cited responses (every claim linked to a real PubMed article), and the Journal Finder suggests where to publish, ranked by fit and enriched with impact, Open Access, and APC-deal data.

Data Sources

• PubMed/MEDLINE — article search, metadata, MeSH terms, abstracts (NCBI E-utilities)
• Europe PMC — preprint search (bioRxiv, medRxiv)
• OpenAlex — primary journal impact metrics (250K+ sources, live API, updated weekly)
• SCImago Journal Rank 2024 — fallback journal impact scores, specialty areas
• NIH iCite — citation counts and Relative Citation Ratio
• PubMed Central — full-text figures and tables for open-access articles
• PubMed eLink — reference extraction from article bibliographies
• Claude (Anthropic) — AI query building, PICO enrichment, risk of bias, reviews

© 2024-2026 EvidenceScope ·info@evidencescope.com
EvidenceScope is not affiliated with PubMed, NCBI, SCImago, or NIH.

The six-stage pipeline

Beyond the review pipeline

Three more tools sit in the top tabs, alongside Search literature:

All features

How to use EvidenceScope

1. Search — Enter a research question (AI builds the query) or type terms manually. Choose a specialty quick-start. Set date period. Toggle study type filters (include/exclude). Enable preprints if needed.
2. Fetch metadata — Click "Search" then "Fetch metadata". MeSH terms and title keywords are extracted automatically.
3. Refine by content — Use the boolean group builder: drag terms into ALL/ANY/NONE zones. Create groups with AND/OR/NOT. Click "AI refine" for automated suggestions.
4. Filter by impact — Choose a journal-impact cutoff (Top 10%, Top 25%, or All). Set minimum citations. Select specialty areas. Toggle "Include unranked".
5. Screen articles — Include or exclude each article. Use "Include all" for bulk selection. Enrich with AI for PICO data.
6. Assess quality — Run "Risk of Bias" for Cochrane-style D1-D7 assessment on all articles. View results inline and as overview diagram.
7. Generate review — Click "AI Review" (uses included articles only). Review includes PRISMA flow diagram, RoB overview, and PRISMA checklist.
8. Export — Download as Word document. Export Excel with all data (PICO, RoB, screening, citations). Save session for later.

Subscription plans

Free plan: test all features with limited uses (lifetime, not monthly). Institutional pricing available — contact us. Payments via Stripe. Cancel anytime.

Technical details

APIs and data sources

• NCBI E-utilities — esearch (search + count), esummary (metadata), efetch (full XML with MeSH and abstracts), elink (PMC linkage)
• SCImago Journal Rank 2024 — bundled locally as JavaScript dataset. Matched by journal abbreviation and ISSN
• NIH iCite — REST API for citation counts and Relative Citation Ratio per PMID
• PubMed Central — full-text XML for figure/table extraction from open-access articles
• Claude API (Anthropic) — Sonnet model via Google Apps Script proxy for all AI features
• Ask literature — grounded Q&A with retrieval, generation, and usage metering handled server-side (backend at /evidencescope-ask/); answers stream to the browser with inline PMID citations
• Your digest — built by a Google Apps Script engine on a schedule (Firestore for preferences and per-user caching, Resend for email); the shared general digest and each signed-in user's personal digest are pre-built twice a week in the background and served from cache, so opening the tab never waits on a build

Privacy

• Search queries are sent directly from your browser to PubMed. They are not stored on EvidenceScope servers.
• Article data is sent to Claude only when you explicitly click an AI feature button.
• User identity (Google Sign-In) is used only for subscription management and usage tracking.
• Search history is stored in your browser's local storage only.
• Payment processing is handled entirely by Stripe. EvidenceScope never sees your card details.

Rate limiting

NCBI requests are batched with 350ms delays between calls to comply with E-utilities guidelines. Up to 10,000 article IDs can be fetched per search. Full articles are loaded for up to 150 matched results.

Browser support

Any modern browser (Chrome, Firefox, Safari, Edge). No installation required. Works on desktop and mobile.

Limitations & disclaimers

• Search coverage: Results depend on PubMed indexing. Articles not yet indexed or in databases not covered (Embase, Web of Science, Cochrane) will be missed.
• Journal matching: Uses SCImago 2024 dataset for initial ranking. Unranked journals are automatically looked up via OpenAlex (live API, 250K+ sources, updated weekly) using their 2-year mean citedness metric.
• MeSH sampling: MeSH terms are fetched for the first 200 results only. Content filtering by MeSH is approximate for large result sets.
• AI accuracy: AI summaries, reviews, PICO extractions, and figure data may contain errors, omissions, or hallucinations. Always verify against sources.
• Figure extraction: Experimental feature. Only works for PMC open-access articles. Numerical data extracted from graphs by AI should be treated as approximate.
• Citation data: iCite citation counts may lag behind current counts. Recently published articles may show zero citations.
• Ask literature: Answers are grounded in the PubMed abstracts retrieved for your question. Relevant work that isn't retrieved can be missed, and every answer should be verified against its cited papers.
• Not a systematic review: Does not replace PRISMA-compliant methodology. Use as a starting point, not a final product.