How Serpio picks topics

Serpio's idea feed is not a keyword tool and it is not regenerated every time Home loads. It is a pipeline that starts from what your site is about, researches public candidates, checks their evidence, and hands a complete set to the app for durable storage. This guide explains the shape of that pipeline. The step-by-step lives in How the idea feed works.

It starts with content focus

During site analysis Serpio detects one or more content focuses: the subjects your site is credibly about, each with clusters and target keywords. You confirm or edit them in the Content strategy section of workspace settings, and nothing is drafted until you do. Each confirmed focus is mapped to a shared idea pool in Serpio's content engine. The pool is shared because public research on, say, cookieless analytics is the same for every site in that space. What Serpio does with the pool is scoped to you.

The content engine supplies candidates

The content engine researches public topic pools for your confirmed focuses and returns evidence-backed candidates. The Serpio app chooses a balanced set of ten, assigns permanent app-owned IDs, and stores an immutable collection. The dashboard then reads only those stored records, so an engine outage cannot make an existing collection disappear.

Knowledge Base documents do not create idea candidates. They join the workflow later, after you choose a topic and start an article, when the writer retrieves useful private facts and possible internal links. This keeps knowledge ingestion, idea discovery, and article generation as three separate lifecycles.

Evidence checks before an idea is shown

A candidate has to clear several checks before it becomes a tile. Every one of them can be traced back to a setting you control.

• It needs at least one piece of evidence that is not stale and has a readable excerpt. No evidence, no idea.
• Its sources must not be on your blocked list, and its topic must not match your excluded topics. Both live in your editorial profile.
• Its title and thesis have to overlap with the focus's keywords, and so does its evidence. A catchy title on unrelated evidence is dropped.
• It must be classifiable as timely or evergreen. Timely needs a recent publication date on the evidence.
• It is deduplicated against the current collection and the rest of the new set, so a repeated provider record or title cannot occupy two slots.

Weekly collection and Refresh

Building a collection uses paid research and model services, so Serpio does it on a controlled schedule. A successful collection remains current for seven days. Opening or reloading the dashboard reads it from the app database without another engine or model call. A weekly worker prepares the replacement in the background.

The Refresh button allows up to three successful replacements per UTC week. The current set remains visible during preparation and through any failure. Only a fully stored set of ten advances the current pointer; every older collection stays under Previous ideas.

You can always bring your own topic

The feed is a suggestion layer, not a gate. Enter your own topic in the workbench, or Use topic on a content idea in the Site Report, starts a draft on anything you choose. Your topic then goes through the same evidence-linked preparation as a feed idea: retrieval from your knowledge, bounded public research, and the quality gate described in No invented links or facts. See Bring your own topic for the how-to.