Tesla logo
Tesla+0.00%
Apple logo
Apple+0.00%
BTC logo
Bitcoin+0.00%
MTL logo
Metal+0.00%
Nvidia logo
Nvidia+0.00%
Google logo
Google+0.00%
ETH logo
Ethereum+0.00%
AVAX logo
Avalanche+0.00%
XRP logo
Ripple+0.00%
XAU
Gold+0.00%
XAG
Silver+0.00%
Tesla logo
Tesla+0.00%
Apple logo
Apple+0.00%
BTC logo
Bitcoin+0.00%
MTL logo
Metal+0.00%
Nvidia logo
Nvidia+0.00%
Google logo
Google+0.00%
ETH logo
Ethereum+0.00%
AVAX logo
Avalanche+0.00%
XRP logo
Ripple+0.00%
XAU
Gold+0.00%
XAG
Silver+0.00%
Tesla logo
Tesla+0.00%
Apple logo
Apple+0.00%
BTC logo
Bitcoin+0.00%
MTL logo
Metal+0.00%
Nvidia logo
Nvidia+0.00%
Google logo
Google+0.00%
ETH logo
Ethereum+0.00%
AVAX logo
Avalanche+0.00%
XRP logo
Ripple+0.00%
XAU
Gold+0.00%
XAG
Silver+0.00%

Transparency

How X.News works

Every article is AI-generated from verified X.com sources. Our pipeline, safety rules, prompts, and source list are fully public.

Overview

Curated Sources

700+

verified X.com accounts

Categories

12

topic areas covered

Safety Layers

3

independent defense gates

Art Styles

12

unique image aesthetics

Pipeline

How every article is made

From X.com post to published article — 8 automated steps, fully auditable.

01

Source Discovery

Search X.com every 45 min across 12 categories

02

Quality Filter

700+ curated accounts, verified users, 500+ follower minimum

03

Safety Screen

30 spam/scam keyword patterns rejected at source

04

AI Generation

xAI Grok writes with editorial depth, humanity, and nuance

05

Deduplication

Title similarity + entity overlap + recent article awareness

06

Post-Gen Safety

Second keyword scan on generated title, summary, and body

07

Assessment

Truth score, fact-check, bias detection, dissenting views

08

Image Gen

12 artistic styles, category palettes, controversy detection

Content Safety

3-layer defense system

Three independent gates ensure spam, scams, and low-quality content never reaches readers.

Layer 1 — Source Quality Gate
700+ curated, verified source accounts with priority weighting
Hard minimum of 500 followers for all source accounts
30 spam/scam keyword patterns rejected at source level
No fallback to low-quality sources — category skipped instead
Layer 2 — AI Prompt Rules
Explicit instruction: never write about giveaways, airdrops, or scams
Never promote financial products or investment opportunities
Returns skip=true if all source material is spam
Strict category adherence — off-topic content discarded
Layer 3 — Post-Generation Filter
30 banned keywords scanned in title, summary, and full article body
Blocked articles never reach the database
All blocks logged for audit trail
Periodic database hygiene removes legacy content

Editorial Voice

How we write

Our AI receives a detailed editorial prompt. These are the core principles.

AI Editorial PromptxAI Grok-3-mini · Temp 0.5 · 3K tokens

Intellectual Depth

Every sentence should feel considered, not generated. Genuine human warmth.

Concrete Details

Replace generic phrases with specifics. Say what makes it significant.

Why & So What

Not just what happened — why it matters for lives, industries, futures.

Multiple Perspectives

Acknowledge complexity, present dissent, note uncertainty.

Credit Sources

Integrate @mentions of X.com accounts as the real voices.

Forward-Looking

Genuine insight about what comes next — never 'time will tell.'

Quality Assurance

Built-in verification

Every article ships with two independent assessments, visible to all readers.

Grok Assessment
Truth Score (0–100)
Sentiment analysis & breakdown
Bias indicators
Key takeaways
Broader context note
Confidence level
Dissenting views
Fact-Check
Overall accuracy score (0–100)
Verdict: verified → unverified
Per-claim verification
Red flags & concerns
Source quality rating
Methodology explanation

Visual Identity

12 artistic styles

Each article image uses a deterministically selected style for maximum visual diversity.

Manga
Pop Art
Noir
Propaganda
Cyberpunk
Watercolor
Street Art
Art Deco
Woodcut
Isometric
Collage
Pixel Art

Controversial content — articles with low truth scores or bias indicators are limited to Noir, Propaganda, and Street Art styles.

Curated Sources

700+ verified voices

Hand-picked experts, founders, journalists, and institutions across 12 categories.

CategoryAccounts
AI
100
Technology
74
Blockchain
89
Finance
60
Politics
89
Science
43
Aerospace
29
Robotics
24
Biotechnology
100
Energy
27
Legal
32
Manufacturing
27

Commitments

What we never do

Explicit boundaries we maintain for editorial integrity.

Promote financial products, investment opportunities, or token sales
Publish content from unverified, low-follower accounts
Generate articles about giveaways, airdrops, or scams
Hide our methodology — everything is documented publicly
Accept paid placements or sponsored content
Suppress dissenting views or inconvenient perspectives

Open Source

Public documentation

All editorial rules are published on GitHub. Read, audit, or fork.

FileDescription
TRANSPARENCY.mdPipeline & Methodology
SAFETY.md3-Layer Safety Framework
PROMPT-GUIDELINES.mdEditorial Prompts & Voice
SOURCES.md700+ Curated Source Accounts

Transparency DAO

Community governance for AI journalism. Vote on prompt improvements, propose new sources, flag issues, and shape editorial direction.

Vote on prompts
Propose sources
Flag content
Shape voice
Follow @xnews_app

X.News is an independent project — not affiliated with X Corp or x.com. We use X.com's public API as a data source and maintain complete editorial independence.