Chapter 05 · Knowledge Base

Usage settings: brand knowledge, evidence, and permissions

Every source carries two settings that decide how Serpio may use it: what kind of knowledge it is, and whether its facts may appear in article copy.

Updated Sep 20, 2026 · 5 min read

On this page

Two settings on every source decide how Serpio may use it. Use this as says what kind of knowledge it is. Permission for writing says whether facts from it may appear in article copy. A third, Content focuses, limits which topics it applies to. You set them when you add a source and can change them at any time in the drawer under Source settings.

Use this as

Brand knowledge
Material about you: what the product does, who it is for, how you talk about it, what you charge. Pages found by the website crawler default to this. It gives the writer accurate brand context after an article starts.
Reference evidence
Material that backs up claims: research, standards, third-party data, a partner's documentation. Documents and added URLs default to this. Passages from it are offered to the writer as evidence for specific statements.

The choice appears in the list as the first part of the usage label, for example Brand knowledge · Facts allowed. Both kinds are retrieved the same way; the label tells the writer how to treat the passage.

Permission for writing

This is the setting that protects you. The form shows one of two explanations depending on your choice.

OptionCopy shown in the formEffect at write time
Facts allowed in articlesRelevant facts may appear in article copy. Uploaded files stay private; public citations require a public source URL.Passages are retrieved as evidence and the writer may state facts from them. This is the default for every new source.
Background context onlyUse for understanding and context. Its private facts and quotations are not approved for article copy.Passages are given to the writer as background only. They are excluded from the evidence list, so their facts cannot be quoted, and they are not used for internal-link targets.

Why private facts need explicit permission

An uploaded file is private. If the writer states a figure from it, the article carries a fact that readers cannot verify anywhere public, and that you may not have intended to disclose. Board decks, unreleased roadmaps, and internal metrics are the usual culprits. Mark them Background context only: Serpio still learns your positioning from them, but nothing in them becomes copy.

Content focuses

If your brand has content focuses, the form lists them as checkboxes under Content focuses, with the note none selected means all. Leave them all unticked for material that applies everywhere, such as your brand brief. Tick one or more to scope a source: a compliance document ticked for a regulatory focus will not be retrieved when Serpio writes for a lifestyle focus. Up to 20 focuses can be selected per source. See Reviewing content focus for how focuses are set.

Usage settings pending

Sources indexed before these settings existed can show Usage settings pending in place of the usage label. Open the source, expand Source settings, and you will see Usage settings have not been verified. The selected settings take effect when you save. The form is pre-filled with the defaults. Review them, adjust if needed, and press Save settings. Do this before relying on such a source for facts.

Changing settings later

  1. 1

    Open the source

    Click its row on the Knowledge base page.

  2. 2

    Expand Source settings

    The section is collapsed by default, under the Preview. It holds Title, How should Serpio use this?, and the same fields as the add modal.

  3. 3

    Save settings

    Changes apply immediately to future article retrieval and bump the brand's knowledge revision. Idea collections and articles already written are unchanged.

Changing settings never re-fetches or re-processes a source. If you also want fresh content, press Check URL now or Reprocess original.

How retrieval respects these at write time

When Serpio writes, it first lists the sources that are currently eligible: not paused, not deleted, with a ready version, marked Facts allowed in articles, inside their effective and review window, and either unscoped or scoped to the article's focus. It then searches only those sources, combining meaning-based matching with keyword matching, and keeps at most two passages per source so one long document cannot crowd out the rest.

Every passage that survives ranking is checked once more against the live source record: same active version, same permission, still inside the window, still in scope. A setting you changed a second ago wins over anything cached. Context-only sources are fetched in a separate, smaller set as background and are never in the evidence list. This is why retired knowledge does not leak into a new draft, and why the Knowledge used panel on an article can name the exact version each passage came from.

Common questions

  • What are the defaults?

    Documents and added URLs default to Reference evidence and Facts allowed in articles. Crawled website pages default to Brand knowledge and Facts allowed in articles. No focuses are selected, so the source applies to all.

  • If I switch a source to Background context only, do existing articles change?

    No. The change affects retrieval from now on. Articles already generated keep their text and their quality report.

  • Can a context-only source influence ideas?

    No Knowledge Base source generates ideas. A context-only source is consulted after you start an article as background for the writer; its facts and quotations are kept out of the copy.

Your next article starts here.

Turn a topic that is trending right now into your first article.

Write my first article

3 free articles a month. No credit card.