Why we built a news engine for the age of AI search
Search is changing. People increasingly get their answers from AI assistants — ChatGPT, Gemini, Perplexity and Google's AI Overviews — rather than from a list of blue links. For publishers this raises a hard question: will an AI model find your content, understand it, and quote it accurately?
That question is the reason we built our news aggregation engine: a turnkey platform for launching and running vertical news networks whose content is engineered, from the ground up, to be indexed by search engines and quoted by AI systems.
What the engine does
In short, it turns a curated set of trusted sources into a fast, well-structured, SEO-first news site. It ingests articles from sources you select, rewrites them with AI into clean, original, structured pieces, and publishes them with machine-readable blocks that both Google and large language models can parse.
How it works
Content moves through a transparent, human-supervised pipeline:
- Crawl. A scheduled job fetches your enabled sources, extracts the main text with a Readability parser, and deduplicates by content hash so nothing is published twice.
- Editorial queue. Raw items land in an admin queue where an editor approves or discards them — nothing goes live automatically.
- AI rewrite. Approved items pass through a Gemini-powered rewrite pipeline that produces an original article in a consistent, structured format.
- Publish. An editor reviews, edits if needed, and publishes. Pages are server-rendered for speed and crawlability.
Built to be indexed by AI bots
The core idea is "AI-quotable" content. Every published article is assembled from structured blocks designed to be lifted cleanly by a model:
- Quick Answer — a 2–3 sentence summary an assistant can quote directly.
- Key Specs and Pros / Cons where relevant.
- What changed and why it matters.
- FAQ — 5–8 questions and answers, emitted with FAQPage structured data.
- Sources — transparent citations linking back to the originals.
On top of that, each page ships schema.org structured data (NewsArticle and FAQPage JSON-LD), a dynamic sitemap, tuned robots rules, semantic HTML and server-side rendering. The result is content that follows current SEO best practices and is easy for both classic crawlers and AI systems to read, trust and cite.
Editorial control and fair sourcing
Automation does not mean a free-for-all. A human approves everything before it is published. A configurable, per-domain media policy governs how images and video are handled — OEM press kits are attributed, Creative Commons licenses are preserved, YouTube and Vimeo are embedded rather than copied, and unknown publishers link back to the source. Editorial personas represent aggregated domain expertise, and every article cites where its information came from.
The technology
The engine is built on a modern, maintainable stack: Next.js (App Router) and TypeScript, Postgres via an ORM, the Google Gemini API for rewriting, and Vercel for hosting and scheduled jobs, with a shadcn/ui + Tailwind interface and a full admin panel for sources, queue, articles and media policies.
See it in production
One live network running on the engine is atvupgrade.com — a news, reviews and guides hub for the ATV, UTV and side-by-side community. It publishes fresh, structured articles every day across News, Reviews, Guides, Recalls and Gear. The same engine can power a vertical news network in virtually any niche.
Launch your own news network
If you want to launch an SEO-first, AI-ready news site in your industry — or turn an existing audience into a steady stream of structured, citable content — this is exactly what the engine is for. Get in touch and we'll help you get started.
FAQ
Is the content original? Yes. Source material is rewritten by AI into new, structured articles, and every piece links back to its sources.
Is it fully automated? No. Crawling and rewriting are automated, but a human editor approves and publishes every article.
What niches does it support? Any. The category structure, sources and editorial personas are fully configurable.
