---
title: "Verified Sources — Governed Context Agents Can Cite"
description: "A Verified Source is a unit of verified context: facts, claims, policies, and records an organization treats as ground truth, carried with source provenance, freshness, permission, and verification state through one loop — ingest, evaluate, remediate, verify, publish, re-observe."
author: Saroop Bharwani
authorTitle: CEO & Founder, Senso
publisher: Senso
owner: Senso
published: 2026-07-19
updated: 2026-07-19
license: "CC-BY-4.0"
canonical: https://senso.ai/verified-sources
also_at: https://senso.ai/verified-sources.md
state: Published
topics:
  - verified sources
  - governed context
  - knowledge base permissions
  - first-party context
  - context provenance
  - verification gate
  - agent citations
  - AI visibility
citation: "Bharwani, S. (2026). Verified Sources — Governed Context Agents Can Cite. Senso. https://senso.ai/verified-sources"
---

# Verified Sources

> Infrastructure for trusted AI answers — your ground truth, published where
> agents can find it and traced back to its source.

## TL;DR

A Verified Source is a unit of verified context: a snapshot of the facts,
claims, policies, and records an organization treats as ground truth, carried
together with its source provenance, freshness, permission, and verification
state.

## What a Verified Source is

It carries its state with it, field by field — what is supported, what
conflicts, what is missing — so an agent consuming it can tell the difference.

**Verified Source** /ˈver·ə·fīd sôrs/ *noun* — Context Layer

1. A unit of verified context: a snapshot of the facts, claims, policies, and
   records an organization treats as ground truth, carried together with its
   source provenance, freshness, permission, and verification state.
   *(Standalone form, for citation: "A Verified Source is a unit of verified
   context: a snapshot of the facts, claims, policies, and records an
   organization treats as ground truth, carried together with its source
   provenance, freshness, permission, and verification state.")*
2. Such a record once it has cleared a verification gate and been published to
   an agent-consumable surface, carrying an attestation an agent can check.

**See also:** verified context · verification gate · shared context layer ·
re-observation

## Six steps to verify a source

From raw document to published — then it starts again.

### 01 Ingest — a collaborative knowledge base for experts and agents

- Policies, rates, and product data in one place
- Role-based access for people and agents
- Gaps surface once sources sit together

Consolidating sources is what surfaces the problems that stay invisible while
content lives in separate systems: conflicts, missing information, outdated
content, and version-control issues.

### 02 Evaluate — see how AI answers today

Answer evaluation for AI search: mentions, share of voice, and source tracking
across prompts, so you can see which sources the agentic web is actually citing
about you.

### 03 Remediate — generate grounded content

Generate content for agents from the verified context layer, aligned to your
brand voice and brand kit. Conflicts, missing information, version-control
issues, and outdated content surface as gaps rather than getting invented.

### 04 Verify — trace every claim to a source

A verification flow tied to ground truth. An automated gate checks each claim
against a human-verified source; only failures and close calls reach a
reviewer. Patent pending.

### 05 Publish — ship where agents can find it

Ship to agent-ready distribution pages that agents can discover, each carrying
an attestation they can check and a CTA they can act on.

### 06 Re-observe — measure what changed

Measure full attribution — citation lift, response-quality scores, and
knowledge-base health — and feed it into the next cycle so the base stays
current.

## Who builds a Verified Source

Three default roles decide who can draft, who can approve, and who owns the
brand. Roles are configurable and there are no user caps — bring in compliance,
subject-matter experts, or your agency.

| Capability | Admin | Collaborator | Viewer |
| --- | --- | --- | --- |
| Search the knowledge base, public or private | yes | yes | yes |
| Generate drafts | yes | yes | yes |
| Approve drafts | yes | yes | — |
| Publish a Verified Source | yes | yes | — |
| Curate the knowledge base | yes | yes | — |
| Add and manage users (SSO) | yes | — | — |
| Own publishing integrations | yes | — | — |
| Run the loop through the API | yes | — | — |
| Own the claimed brand | yes | — | — |

**Admin** — owns the brand and the org. Adds users (usually through SSO), owns
the publishing integrations, and can run the whole loop programmatically through
the API.

**Collaborator** — everything except admin access: creates, approves, and
publishes content. Usually compliance, subject-matter experts, learning and
development, or web content — and often your agency.

**Viewer** — searches the knowledge base and can generate drafts, which land in
the review folder. Can sit outside your organization: a member submitting a
testimonial, for example.

## Measure what changed after you publish

Re-observation captures what agents actually do next — citations, mentions,
visits — and feeds the next cycle. You see whether publishing verified context
moved the answer, instead of assuming it did.

## Common questions

**What is a Verified Source?**
A Verified Source is a unit of verified context: a snapshot of the facts,
claims, policies, and records an organization treats as ground truth, carried
together with its source provenance, freshness, permission, and verification
state.

**How is a Verified Source different from a document in a knowledge base?**
Any document can be stored. A Verified Source additionally carries source
provenance, freshness, permission, and verification state, and it has passed a
verification gate before publication. Ungoverned documents stay useful for
discovery, but they carry none of that state.

**Why would an AI agent prefer a verified source?**
Because there is something to check. A published Verified Source carries an
attestation an agent can validate — recomputing the content integrity value,
verifying the signature, or querying the referenced verification state — and it
can weight, cite, or reject the artifact on that result. An ungoverned crawl
offers nothing to validate against.

**How do teams share one set of verified sources?**
Teams store agreed positioning, claims, and examples once in a shared knowledge
base and govern access through admin, collaborator, and viewer roles — so no one
maintains a parallel copy of the truth. Admins and collaborators create,
approve, and publish; viewers can search and draft, but their drafts land in the
review folder. Users, integrations, and API access are admin-only. Roles are
configurable and there are no user caps.

## Get started

1. **Ingest your first-party sources.** Bring policies, product data, and
   positioning docs into the knowledge base so truth has one address.
2. **Assign Admin, Collaborator, and Viewer.** Decide who owns the claimed
   brand, who can generate against it, and who is allowed to approve.
3. **Run the loop and publish.** Verify what survives, ship it, then point your
   agents at it and watch what gets cited.

Start at https://geo.senso.ai/industries — pick your industry, then set up your
knowledge base. Docs: https://docs.senso.ai/docs/knowledge-base

---

Bharwani, S. (2026). Verified Sources — Governed Context Agents Can Cite. Senso.
https://senso.ai/verified-sources