pubmed2blog

CLI tool that transforms PubMed papers into SEO-optimized healthcare blog articles.

Why pubmed2blog?

Healthcare marketers need to produce evidence-based content, but:

• Manually searching PubMed takes hours
• Evaluating study quality requires medical literacy
• Writing accessible articles from dense papers is time-consuming

pubmed2blog automates the entire pipeline — from research discovery to publishable blog posts.

Features

• PubMed Search — Search with MeSH terms, journal filters, and date ranges
• AI-Powered Scoring — Ranks papers by blog suitability (journal tier, study type, sample size, novelty)
• Multi-Source Full Text — Fetches full text from PMC, Unpaywall, Europe PMC, and OpenAlex
• Brand Voice — Inject your own brand voice into generated articles
• 4 Article Types — Research Explainer, Patient-Facing, Differentiation, Service Connection
• Multi-Language — Generate in English, German, or both
• Open Source — MIT licensed, npm packageable

Installation

npm install -g pubmed2blog
pubmed2blog init

Quick Start

# Configure your API keys (interactive)
pubmed2blog init

# Discover relevant papers (ranked by blog suitability)
pubmed2blog discover "cardiovascular prevention"

# Get paper details
pubmed2blog extract 39847521

# Generate blog article
pubmed2blog generate 39847521 --type research-explainer

# Full pipeline (discover + extract + generate)
pubmed2blog pipeline "sleep quality" --top 3 --save

Commands

| Command | Description | |---------|-------------| | init | Interactive setup (API keys, brand voice) | | discover <keyword> | Search and rank papers by blog suitability | | extract <pmid> | Fetch full paper details | | generate <pmid> | Generate blog article | | pipeline <keyword> | Full pipeline: discover → extract → generate | | config | Show current configuration | | keywords | List configured keyword categories | | types | List article types with descriptions |

Options

discover

pubmed2blog discover "keyword" [options]

• --days <n> — Days to search back (default: 90)
• --tier <tiers> — Journal tiers: tier1,tier2,tier3 (default: tier1,tier2)
• --limit <n> — Max results (default: 20)
• --study-type <type> — Filter: individual, meta, any (default: any)
• --json — Output as JSON

extract

pubmed2blog extract <pmid> [options]

• --json — Output as JSON
• --save — Save to file

generate

pubmed2blog generate <pmid> [options]

• --type <type> — Article type (default: research-explainer)
• --lang <langs> — Languages: en,de (comma-separated)
• --model <model> — LLM model override
• --save — Save article to ./output/

pipeline

pubmed2blog pipeline <keyword> [options]

• --top <n> — Number of papers to process (default: 1)
• --type <type> — Article type
• --save — Save all outputs

Article Types

research-explainer

Study findings for an educated lay audience. Lead with surprising finding, include numbers with explanations, mandatory limitations section, actionable takeaways.

patient-facing

Accessible explanation without jargon. Short paragraphs, analogies, clear "Bottom Line" with 3 bullets.

differentiation

"Why we don't offer X" — Acknowledge hype fairly, present pro + contra arguments, explain evidence standards.

service-connection

80% science, 20% relevance. Connect study findings to preventive health services.

Supported Providers

| Provider | Environment Variable | Default Model | |----------|---------------------|--------------| | Anthropic | ANTHROPIC_API_KEY | claude-sonnet-4-20250514 | | OpenAI | OPENAI_API_KEY | gpt-4o | | Z.AI | ZAI_API_KEY | glm-4-flash |

Model auto-detection: claude* → Anthropic, gpt*/o1*/o3* → OpenAI, glm* → Z.AI

Configuration

Config is stored in ~/.pubmed2blog/config.json:

{
  "provider": "anthropic",
  "apiKey": "sk-ant-...",
  "model": "claude-sonnet-4-20250514",
  "languages": ["de", "en"],
  "brandVoice": "Your brand voice description...",
  "journalTiers": ["tier1", "tier2"],
  "defaultArticleType": "research-explainer",
  "outputDir": "./output"
}

Scoring Algorithm

Papers are scored by blog suitability:

• Journal Tier — Tier 1 (IF>50): +30, Tier 2 (IF 15-50): +20, Tier 3 (IF 5-15): +10
• Study Type — RCT: +25, Clinical Trial: +20, Cohort: +15, Meta-Analysis: +5
• Sample Size — ≥10,000: +20, ≥1,000: +15, ≥100: +10
• Novelty — +5 for signals like "first", "novel", "breakthrough"
• Actionability — +5 for signals like "recommendation", "guideline"

Full Text Sources

Tried in order:

1. PMC — PubMed Central (free, official)
2. Unpaywall — Legal OA versions
3. Europe PMC — European mirror
4. OpenAlex — Research graph API

If no full text available, article is generated from abstract only.

Requirements

• Node.js 18+
• API key for at least one LLM provider

Use Cases

• Healthcare clinics publishing evidence-based blog posts
• Healthtech companies creating content from research
• Medical marketers automating content production
• Preventive medicine practices sharing latest findings

Roadmap

• [ ] Tests
• [ ] CI/CD
• [ ] More article types
• [ ] Custom prompt templates

License

MIT

Links

• npm: https://www.npmjs.com/package/pubmed2blog
• GitHub: https://github.com/Holic101/pubmed2blog