A Content Engine That Takes a Site From Raw News to Published Article, Hands Free
An automated editorial pipeline that pulls trending news, selects the strongest story, writes a full SEO-optimised article in a controlled editorial voice, generates a matching image, and publishes it live, on a schedule, across a network of sites.
Where they started.
Running a network of news and blog properties means an unending demand for fresh, relevant, well written content. Producing that by hand does not scale past a handful of sites, and the quality drifts the moment volume goes up. The client needed a pipeline that could take a publication from raw, real time news all the way to a finished, published, search optimised article, with no person in the loop, and do it on a repeatable schedule across many sites at once.
What we did.
The pipeline was designed as a newsroom compressed into a workflow. On a schedule, it gathers what is happening, decides what is worth writing about, writes it properly, illustrates it, and publishes it. The two hard problems were selection and quality. Selection was solved by pulling from several sources at once, merging and de duplicating them, and ranking to surface the strongest current story rather than the first one seen. Quality was solved with a controlled editorial voice: the writing stage runs against a strict style system that enforces reporting density and bans the generic, hollow connective language that makes automated content obvious. A duplicate checker prevents the same story being covered twice, pulling from an alternate source when a headline has already been used.
A scheduled trigger fires the run. The pipeline pulls headlines from news feeds and a news data API in parallel, merges them into one pool, and strips duplicates by normalising URLs. It ranks the pool, selects the top stories, then classifies the topic and picks an editorial angle and a set of trusted sources to cite based on that classification. A style layer injects the tone and the rules, and a prompt is assembled instructing the model to write a long form, properly structured article using the selected stories as source material and the trusted domains for citations. A matching header image is generated, and the finished article is pushed straight to the site's publishing platform over its content API.
The whole workflow is a template. Adapting it to a new property is a controlled process of swapping the routing, the author identities, the target URLs, and the credentials, so one proven pipeline propagates across the network rather than being rebuilt each time. On a related build for a financial vertical, the same shape runs continuously off aggregated feeds, generating articles under consistent author bylines.
What changed.
A site can publish fresh, relevant, search ready articles on a daily schedule with zero human involvement per article, and the same engine can be cloned across an entire network of properties. Because selection ranks the strongest current story and the writing stage enforces a real editorial voice, the output reads like reporting rather than filler. The client gets newsroom output at software cost, repeated across many sites at once.
Built with n8n, News RSS + news data APIs, LLM article generation, AI image generation, WordPress REST publishing.
AI Automation
This is AI Automation applied to publishing: a pipeline that turns raw input into finished, on-brand articles at network scale. We can build the same content engine around your topics, your voice, and your sites.