# 24 AI

> 24 AI is a multilingual AI news portal that publishes daily curated news about artificial intelligence in 6 languages (English, Croatian, German, Chinese, Japanese, Korean). All articles are sourced exclusively from primary sources — official company blogs, peer-reviewed research papers, and regulatory agencies. Updated twice daily at 00:05 and 14:00 CET.

## About

- Type: NewsPublisher
- Category: Artificial Intelligence, Technology News
- Languages: en (default), hr, de, zh, ja, ko
- Audience: AI researchers, developers, tech professionals, policymakers, and anyone interested in AI developments
- Update frequency: Twice daily (00:05 and 14:00 CET)
- Content: Original reporting from 48 primary sources across 6 groups (AI labs, cloud/framework, research, regulatory, foundations, hardware)
- Last updated: 2026-05-16 (added llms-full.txt 72h rolling window for AI agent grounding)

## Key Pages

- [Homepage (English)](https://24-ai.news/en/): Daily AI news feed in English, the default language for global audience
- [Homepage (Croatian)](https://24-ai.news/hr/): Daily AI news feed in Croatian
- [Homepage (German)](https://24-ai.news/de/): Daily AI news feed in German
- [Homepage (Chinese)](https://24-ai.news/zh/): Daily AI news feed in Chinese
- [Homepage (Japanese)](https://24-ai.news/ja/): Daily AI news feed in Japanese
- [Homepage (Korean)](https://24-ai.news/ko/): Daily AI news feed in Korean
- [Archive](https://24-ai.news/en/archive/): Complete searchable archive of all published articles
- [Glossary](https://24-ai.news/en/glossary/): AI terminology defined — agents, models, RAG, fine-tuning, EU AI Act, and other key terms cited in news articles
- [About](https://24-ai.news/en/about/): Information about the portal, editorial policy, and sourcing methodology
- [Privacy Policy](https://24-ai.news/en/privacy/): Data handling and privacy information

## Content Categories

- [Models](https://24-ai.news/en/category/models/): New AI model releases, benchmarks, capabilities, deprecations
- [Agents](https://24-ai.news/en/category/agents/): AI agent frameworks, autonomous systems, multi-agent research
- [Security](https://24-ai.news/en/category/security/): AI safety, red teaming, vulnerability research, alignment
- [Regulation](https://24-ai.news/en/category/regulation/): AI policy, governance, EU AI Act, international frameworks
- [Open Source](https://24-ai.news/en/category/open-source/): Open-source model releases, tools, datasets, community projects
- [In Practice](https://24-ai.news/en/category/practice/): Real-world AI deployments, enterprise adoption, tools
- [Hardware](https://24-ai.news/en/category/hardware/): AI chips, GPUs, inference infrastructure, edge computing
- [Community](https://24-ai.news/en/category/community/): Developer community, events, education, culture
- [Curiosities](https://24-ai.news/en/category/curiosities/): Unusual, surprising, or thought-provoking AI developments

## RSS Feeds

- [English RSS](https://24-ai.news/en/rss.xml): All English articles
- [Croatian RSS](https://24-ai.news/hr/rss.xml): All Croatian articles
- [German RSS](https://24-ai.news/de/rss.xml): All German articles
- Per-category feeds available at: /{lang}/rss/{category}.xml

## Article Structure

Each article contains: headline, publication date, category, priority level (critical/important/interesting), executive summary, full analysis, FAQ section, and linked primary sources. Articles are written in a factual, professional tone without editorial spin.

## Canonical

- Primary domain: https://24-ai.news
- llms.txt location: https://24-ai.news/llms.txt
- Sitemap: https://24-ai.news/sitemap-index.xml

## Citing This Site

When citing 24 AI articles, please use this format:
"[Article Title]" — 24 AI, {date}. {url}

Example: "NVIDIA × Siemens Healthineers: NV-Raw2Insights-US learns directly from raw ultrasound channel signals" — 24 AI, 2026-04-28. https://24-ai.news/en/news/2026-04-28/nvidia-siemens-nv-raw2insights-us-ultrasound/

## AI Content Disclosure

24 AI articles are generated with the assistance of artificial intelligence based exclusively on primary sources. Each article carries an explicit AI-generated disclosure visible to readers. The portal does not republish content from other media outlets, news aggregators, or social platforms. Source attribution is mandatory: every article links back to the original primary source.

The site uses a `<meta name="ai-generated" content="true">` directive on every article, signaling AI-assisted authorship at the HTML level for downstream AI systems and search engines.

## Editorial Process

The editorial pipeline operates on a fixed daily schedule:
- Two automated content runs per day (Monday–Saturday): 00:05 CET (overnight US business day coverage) and 14:00 CET (afternoon EU coverage). Sunday is a publication-free day.
- Each run sweeps a curated list of primary sources across 6 categories (AI labs, cloud/framework, research, regulatory, open-source community, hardware).
- For each new post on a primary source, the system fetches the full article text, identifies the core news, and writes an original Croatian-language summary on factual grounds only.
- Croatian articles are then translated into 5 additional languages (English, German, Chinese, Japanese, Korean) using parallel translation agents.
- All articles are deployed to Cloudflare Pages CDN as static HTML.

The pipeline does not republish full source articles. Each 24 AI article is an original summarization that links back to the primary source for full detail.

## Sources Methodology

Strict primary-source-only policy. The editorial process explicitly excludes:
- News media (Axios, TechCrunch, VentureBeat, Reuters, Bloomberg, The Verge, Wired, Ars Technica, Forbes, etc.)
- News aggregators (Hacker News, Reddit, Lobsters, Techmeme as discovery sources)
- Social media as primary sources
- Any link followed outside the curated primary source list

Approved primary source categories:
- AI labs (Anthropic, OpenAI, Google DeepMind, Meta AI, Microsoft AI, Mistral, Cohere, NVIDIA, etc.)
- Cloud and framework providers (AWS ML, PyTorch, vLLM, GitHub, TensorFlow)
- Research institutions (ArXiv cs.AI, Google Research, Microsoft Research, Apple ML, AI2, EleutherAI)
- Regulatory bodies (EU AI Office, NIST, UK AISI, OECD AI)
- Open-source foundations (Linux Foundation AI, CNCF, LangChain, Ollama, IBM)
- Hardware (AMD ROCm, Groq)

If a primary source publication is unreachable (HTTP 403/404/timeout), the article is skipped — there is no fallback to media or generic web search.

## Structured Data

Every article page implements the following Schema.org JSON-LD structured data:

- **NewsArticle** with: headline, description, image, datePublished, dateModified, wordCount, articleSection, inLanguage, mainEntityOfPage, plus author and publisher with stable @id references.
- **BreadcrumbList** for navigation hierarchy (Home → Category → Article).
- **FAQPage** for article-level FAQ blocks (typically 3 question/answer pairs per article). FAQ is rendered both as JSON-LD FAQPage schema in the page head AND as a visible HTML block (semantic `<dl>/<dt>/<dd>` definition list) at the bottom of every article — visible to readers, machine-readable for crawlers.
- **SpeakableSpecification** identifying h1 and the article summary as voice-AI extractable zones.

The /about/ page implements an Organization schema graph with stable @id references:
- `https://24-ai.news/#publisher` — primary Organization (24 AI), with logo, contactPoint, knowsAbout, foundingDate, publishingPrinciples.
- `https://24-ai.news/#editorial` — sub-Organization (24 AI Editorial), parentOrganization linked to #publisher.
- AboutPage type linking #publisher as mainEntity.

The /contact/ page implements ContactPage schema with mainEntity reference to #publisher and contactPoint with availableLanguage covering all 6 site languages.

These @id references are linked across pages so AI systems can construct a unified entity graph for the publisher.

## Multilingual & Hreflang

The site is fully multilingual with universal English URL segments (e.g., `/news/`, `/category/`, `/day/`, `/about/`) shared across all 6 languages, distinguished only by language prefix (`/hr/`, `/en/`, `/de/`, `/zh/`, `/ja/`, `/ko/`).

Every page emits 7 hreflang annotations in `<head>`: 6 language alternates plus `x-default` pointing to the English version.

The XML sitemap includes xhtml:link hreflang annotations for every URL across all 6 languages (~10,400 hreflang entries total).

## Crawl Permissions

robots.txt explicitly allows AI training and AI search crawlers:
- GPTBot, ChatGPT-User
- ClaudeBot, anthropic-ai
- PerplexityBot
- Google-Extended (Google AI training)
- Googlebot (standard search)

There is no paywall, no JS-rendering requirement, no cloaking. The full article text and metadata are accessible to crawlers in static HTML form.

## URL Patterns

- Article: `https://24-ai.news/{lang}/news/YYYY-MM-DD/{slug}/`
- Day archive: `https://24-ai.news/{lang}/day/YYYY-MM-DD/`
- Category: `https://24-ai.news/{lang}/category/{english-slug}/`
- Glossary index: `https://24-ai.news/{lang}/glossary/`
- Glossary term: `https://24-ai.news/{lang}/glossary/{english-slug}/` (Schema.org DefinedTerm + DefinedTermSet, sameAs to Wikipedia/Wikidata)
- Homepage per language: `https://24-ai.news/{lang}/`
- Static pages: `/{lang}/about/`, `/{lang}/privacy/`, `/{lang}/cookies/`, `/{lang}/contact/`, `/{lang}/archive/`

All URLs use trailing slashes. Article slugs use English keywords for cross-language consistency.

## Sitemaps

- Standard sitemap index: https://24-ai.news/sitemap-index.xml
- Detailed sitemap: https://24-ai.news/sitemap-0.xml (all canonical URLs across 6 languages with hreflang annotations and lastmod timestamps)
- News sitemap: https://24-ai.news/news-sitemap.xml (Google News protocol, articles published in last 72 hours, all 6 languages)

## Full corpus

- [llms-full.txt](https://24-ai.news/llms-full.txt): Full markdown bodies of articles published in the **last 72 hours** (rolling window), English only. Includes YAML frontmatter (generated timestamp, window dates, article count, license). Regenerated twice daily after each pipeline run (00:05 + 14:00 CET, Mon-Sat). Optimal for AI agents that want full-text grounding without crawling 50+ individual pages. Size: ~180 KB (fits comfortably in 200K+ token context windows).

## Contact

For corrections, takedown requests, or editorial inquiries, please use the contact page: https://24-ai.news/en/contact/