RAG means Retrieval-Augmented Generation: a system retrieves information from sources and uses it when generating an answer. Sources can be public web pages or approved internal documents. RAG is therefore not inherently limited to company knowledge.
Context and scope
GEO describes work on discoverability and representation in generated search answers. A public search system may use RAG techniques for that purpose. The terms operate at different levels: RAG describes a technical method, while GEO describes work on visibility.
GlasBox separates products by the customer’s goal. Public visibility analysis examines pages, company information and observed answers. An internal knowledge system additionally needs document access, roles, source approval, answer testing and managed operation. A well-cited website does not repair unclear SharePoint permissions.
The company-information part of SOURCE/01 reviews the publicly supported offer identity and possible follow-up questions. It does not build an internal RAG system. Service names, locations and exclusions may overlap, but data permissions and acceptance criteria remain separate.
Two questions, two different project briefs
Fictional example: a property manager wants accurate descriptions in public AI searches and also wants employees to access contract information. The first task uses public service pages. The second requires permission checks per document and cannot be inferred from successful public citations.
| Review point | Example / subject | Interpretation |
|---|---|---|
| Public question | Which services does the property manager offer in Lucerne? | Review public sources and accurate representation. |
| Internal question | What notice period applies in the approved contract? | Check access, document version and cited passage. |
| GEO acceptance | Documented findings and repeated measurement | No guaranteed mention. |
| RAG acceptance | Permitted answers supported; unauthorised access blocked | Verify with role and failure tests. |
How to apply this
- Identify the user and their specific question first.
- Inventory public and protected sources separately.
- Assign permissions and owners for internal sources.
- Agree visibility measurement and system acceptance as separate work packages.
Sources and further reading
The source supports the rules it describes. This article’s working templates are our methodological examples, not client test results.
nDSG and AI-search tools
Next step
The free short check provides an initial assessment of one domain, three buying questions and three observations. An introductory call establishes whether deeper analysis is useful.
Request an introductory call