---
title: Retrieval-Augmented Generation (RAG) overview
slug: agent-platform/docs/retrieval-augmented-generation-rag-overview
docTags: 
createdAt: 2026-08-12T15:37:06.000Z
---

Retrieval-augmented generation (RAG) gives Agent Platform access to your data, campaigns, content, and tests. Mark uses this data to generate contextual, accurate, and personalized responses.

## Why RAG matters

Organizations are cautious about AI. Without checks, AI can do the following:

- Generate outdated or incorrect content.
- Produce tone-deaf or off-brand messaging.
- Violate compliance or privacy requirements.

RAG changes this dynamic by helping AI act more like your employees: consulting current documentation, campaigns, and customer data before responding. RAG is especially valuable in financial services, healthcare, retail, and enterprise marketing. These industries require accurate, personalized AI content.

## How RAG works

RAG retrieves and processes data through three steps before Mark generates a response.

1. **Retrieval** – The AI searches a vector database to find the most relevant information across structured and unstructured sources.
2. **Processing** – Mark analyzes the retrieved data and adds context for accuracy.
3. **Generation** – Using that context, the AI generates an accurate, brand-consistent response.

## Mark RAG features

:::hint{type="info"}
**Note**

Optimizely automatically provisions Mark RAG for eligible accounts. For provisioning details and the custom request path, see [RAG for Content Marketing Platform (CMP)](docId\:nCmw_zZT-Hd8AMa-ut67Z).
:::

Mark RAG replaces manual searching with a permission-aware retrieval experience embedded in Mark. RAG gives Mark context about your business data and assets.

- **Unified content graph** – Campaigns, tasks, assets, and permissions modeled as nodes or edges.
- **Multi-modal retrieval** – Vector and keyword search across text, PDFs, images, video, and metadata.
- **Natural language queries** – A large language model (LLM) translates user requests into optimized graph queries.
- **Intelligent responses** – Summaries, lists, and recommendations refined through Mark for contextual insights.

## Use cases

The following examples show how different roles use RAG in Mark:

- **Marketers** – *Show me all assets tagged with 'Holiday Promotion' used in live campaigns.*
- **Campaign managers** – *What tasks and approvals are pending for the Black Friday campaign?*
- **Developers** – *List all stale feature flags not updated in 90 days.*

## Data privacy and security

Optimizely designed Mark RAG to meet enterprise security requirements. Mark never exposes your content and campaigns to external systems during retrieval or generation.

- **Internal-only data processing** – All data retrieval, embeddings, and indexing happen within Optimizely's secure infrastructure.
- **Permission-aware access** – Every query respects user roles and permissions, ensuring you only see content you are authorized to access.
- **Managed data sync** – Optimizely operates and monitors RAG pipelines internally.
- **Compliance-ready** – Meets enterprise security and privacy requirements, including General Data Protection Regulation (GDPR).
- **Regional data residency** – RAG is available in the EU region. For European customers, Optimizely processes RAG data within EU-based infrastructure to meet data residency requirements.

## Available RAG source

RAG is available for [Content Marketing Platform (CMP)](docId\:nCmw_zZT-Hd8AMa-ut67Z).

:::hint{type="info"}
**Note**

If you use Opti ID, administrators can turn off generative AI in the Opti ID Admin Center. See [Turn generative AI off across Optimizely applications](docId\:X5U0kKD2azV5cIdNqwGNj).
:::