Frequently Asked Questions

SatyaX is an AI-powered fact verification and misinformation analysis platform. It combines BERT-based linguistic analysis, real-time web evidence retrieval, source reliability scoring, and Google Gemini AI reasoning to evaluate the credibility of articles, headlines, URLs, and online claims.

No. SatyaX is an AI-assisted tool that provides a rapid first-pass assessment. AI systems can make errors, especially for recent events, niche topics, or content that requires domain expertise. Always cross-reference critical claims with multiple trusted primary sources and professional fact-checkers.

SatyaX can analyze:
• News article text (pasted directly)
• Headlines and short claims
• Social media posts
• News article URLs (auto-scrapes the full text)
• Generated satirical content (Comedy Generator output)

The Truth Score (0–100) represents the AI's overall credibility assessment:
• 90–100: Highly accurate, strongly supported by evidence
• 70–89: Mostly accurate with minor gaps
• 50–69: Mixed — some claims verified, others not
• 30–49: Low accuracy — significant issues found
• 0–29: Very low accuracy — content appears largely false

BERT is a writing style detector — not a fact-checker. It was trained on 40,000+ articles and learned to recognize how fake news is written (sensational, clickbait) vs real news (formal, structured). It has no world knowledge. Gemini AI, on the other hand, reasons about factual content using its training data and the web search results. When they disagree, Gemini's verdict takes priority. BERT's output is supplementary context about writing style only.

MISLEADING: The content is technically true but intentionally framed to distort reality, omit crucial facts, or create a false impression in the reader's mind.

CONTEXT NEEDED: The content is accurate but incomplete — a reader without the necessary background context would likely misunderstand its significance. There is no deceptive intent implied.

High-tier publishers (Reuters, NYT, WSJ, Bloomberg) use advanced bot-protection systems (Imperva, Cloudflare, TLS fingerprinting) that block all automated scrapers. Even our 5-layer fallback pipeline cannot bypass these protections reliably.

Workaround: When extraction fails, SatyaX shows you direct links to Google Cache and Archive.org versions of the article. Copy the text from there and paste it into the text input to analyze it.

SatyaX tries each layer in sequence, stopping when one succeeds:
1. Trafilatura — High-precision content extraction (primary)
2. newspaper3k — News-site optimised extractor
3. BeautifulSoup — Raw HTML paragraph extraction
4. Google Cache — Cached version for bot-protected sites
5. Archive.org — Wayback Machine snapshot for paywalled content

When a fallback layer is used, an amber notice is shown in the results.

After any analysis, click the Share Result button in the action bar. SatyaX saves the result to the server database and generates a unique shareable URL (e.g. /report/SX-A82KQ1). Copy the link and share it with anyone — no account needed to view it.

Analysis history is stored in your browser's localStorage — it is private to your browser and not sent to any server unless you explicitly click "Share Result." History is available at /history. It persists across sessions but is cleared if you clear your browser data. Up to 20 analyses are kept; older ones are auto-removed.

Click Download PDF in the post-analysis action bar. SatyaX generates a professional 8-section PDF report including the verdict, truth score, writing style analysis, extracted claims, evidence sources, and AI reasoning. The file is named SatyaX_Report_SX-XXXXXX.pdf.

Submitted content is not stored unless you click "Share Result." If you share, the analysis result (including your input text) is saved to the SatyaX database with a unique report ID. Shared reports are publicly accessible via their URL. Your analysis history is kept only in your browser's localStorage.

SatyaX uses:
• BERT (fine-tuned on ISOT/WELFake datasets) — Writing style analysis
• Google Gemini Flash (via REST API) — Deep fact verification and reasoning
• GPT-2 — Comedy generator fallback when Gemini is unavailable
• DuckDuckGo News — Real-time web evidence retrieval (no API key needed)