Medifact User Manual

1. What is Medifact?

Medifact cross-checks medical claims and health information from blogs, news, video, and social media against multiple evidence sources, including PubMed, Cochrane, KCI, NECA, openFDA, MFDS, and Google Fact Check.

It extracts verifiable claims, removes duplicate evidence, and classifies each claim as Verified, Contradicted, or Inconclusive. Results are for reference only and do not replace professional diagnosis or medical care.

2. How to Use the Engine

3. Result Reuse and Evidence Refresh

• Reuse of identical text: If the same user submits the same text with the same analysis options and a valid completed result exists, Medifact opens that report immediately. No new record or job is created, and no additional credit is charged.
• In-progress requests: If an identical analysis is already running, Medifact opens the existing in-progress report instead of starting a duplicate.
• Evidence validity: Standard results are considered current for 24 hours after evidence is checked. Results containing high-risk claims use a shorter 6-hour window by default.
• Manual refresh: On a text report you own, select "Check for latest evidence" to search again. This may create a new analysis version and use a credit. If the evidence is unchanged, the report shows "No new evidence · Previous assessment retained."
• Failures and cancellation: If a new analysis or refresh fails, or you cancel an in-progress analysis, the deducted credit is refunded to registered users.

4. Trust Scores and Grade Levels

The final score is not taken from a generative AI response. The server applies the same deterministic rules to the verdict, evidence level, source, and publication year. Cochrane and higher-level evidence receive greater weight.

When all claims are inconclusive or no usable evidence is available, the report may show N/A instead of a score. Review the insufficient evidence rate, claim verdicts, and evidence check date together with the score.

5. History and Public or Private Results

• My history: Registered users can review their fact-checks and refreshed versions in newest-first order. Guest history is available only in the current browser session and cannot be recovered if its cookie or session is lost.
• Version details: The report header shows the evidence check date, analysis version, and next evidence refresh time.
• Change visibility: You can use "Make public" or "Make private" on your completed report at any time.
• Protection before publishing: Medifact anonymizes personal and private expressions in the title and source text before publication. Content that cannot be safely anonymized or is unsuitable for publication remains private.

6. Credit Billing Policy

A new fact-check analysis uses 1 credit. Opening a valid reused result does not deduct another credit.

• Guest Users: Up to 3 free checks per 24 hours, subject to browser-session and IP limits. History access is limited to the current browser session.
• Registered Members: Get 10 initial welcome credits free, with permanent log storage.
• Billing: Add credits from the Pricing page by selecting Premium Plan or Basic Booster.

7. Ask Medifact (Medical Q&A) Guide

At the bottom of each verified report, you can ask follow-up medical questions directly to the AI regarding the specific context.

• Evidence-based Answers: Follow-up answers use the extracted medical claims and the literature and guidelines matched in the report.
• Differential Credit Policy:
  • Subscribed members and special users (e.g., Doctor, Admin) can ask unlimited questions for free.
  • Standard members get 3 free questions per report, after which 1 credit is deducted per question.
  • Guest users are restricted from sending questions; register to unlock free Q&A sessions.
• Layout Toggle Option: When chat threads grow long, toggle the 'Expand' view option to display the thread vertically without scroll boundaries. Your view preference is kept in local storage.

8. Semantic Relevance & AI Review Details

Review each claim verdict and the source, evidence level, publication year, original link, and relevance details shown on each evidence card.

• Relevance Match Level (%): Evaluates and displays the semantic proximity score between the medical claim statement and the referenced study abstract.
• AI Literature Relevance Analysis & Review: Shows real-time cross-referencing logs, verifying details like age demographic alignment, control groups, and key medical terminology matches to prevent hallucination.
• Insufficient Evidence Rate: The percentage of claims marked inconclusive because relevant evidence is lacking. If this rate is high, do not rely on a high overall score alone.

9. High-Quality AI Text-to-Speech (TTS) Guide

Listen comfortably to report context and summaries using integrated stream playback features.

• High-Quality AI Voice (Brian): Synthesizes and narrates the "AI Fact-checking Summary Report" and "Academic Literature Comparison Results" in a natural, pleasant multilingual voice.
• Playback & Pause Controls: Click the Listen (Speaker icon) button to immediately initiate audio streaming and toggle play/pause states.
• Resilient Fallback Engine: If the external API limit is reached or networks fail, the engine automatically falls back to local built-in browser voices (e.g., Samantha) without interruption.
• Navigation Safeguard: Navigating away from the report details instantly cancels background audio cues and queue threads, ensuring no overlapping audio occurs.