Connect Google Drive, Notion or WordPress to a RAG: sources and permissions

Published on 2026-11-16 by ArduBot

What to check when integrating Drive, Notion or WordPress with a RAG index: synchronization, private files, webhooks, deletions and access control.

A company's information is almost never in a single PDF. It may live in a Drive folder, Notion pages, and WordPress articles. For an assistant to consult it via RAG, the challenge is not just extracting text: you must maintain its origin, its permissions, and its updates.

Three sources with different rules

Google Drive: it may contain private documents, shared folders, and files with individual permissions. An integration must work with an authorized identity and respect the granted scope. It should not index the entire Drive for convenience.

Notion: information is organized into pages and databases with hierarchies and access levels. The extractor needs to preserve titles and relationships, in addition to detecting which blocks or properties are queried securely according to the integration.

WordPress: public articles can usually be retrieved via their pages or API, but drafts, private areas, and plugin content may require additional permissions. A public crawler should never invent content from pages it cannot read.

How I would approach synchronization

First, select the folders, pages, or content types expressly allowed. Then perform an initial load and save for each piece: source, URL or ID, owner, version, date, permissions, and last read result.

In subsequent synchronizations, use available change, event, or periodic polling mechanisms. Avoid assuming that all providers offer the same webhook or that every update arrives only once. A version or hash comparison helps avoid duplicates.

Permissions are not resolved with a prompt

If a Notion page can only be read by the human resources team, the sales assistant must not access its fragments, even if the question is semantically relevant. There are two patterns: index only content authorized for the target audience or implement document-level permissions verified during retrieval.

The second option offers more flexibility, but requires synchronizing access changes as well. A person who loses authorization should not retain it through cached results.

What happens when a source is deleted?

If a page is deleted, access is revoked, or a file no longer exists, the index must react according to its policy: deactivate fragments, review caches, and record the result. Distinguish a temporary 403 or an expired credential from a confirmed deletion.

Do not use a failed integration as an excuse to keep private information accessible indefinitely.

A concrete support case

A company publishes FAQs on WordPress, procedures on Notion, and PDF manuals on Drive. RAG can query authorized sources and return links or reference pages. A question about a specific customer's balance still requires an authenticated transactional service; it should not be resolved by searching through all documents.

How to decide if you need an automatic connector

Measure the number of sources, frequency of changes, content structure, and cost of reviewing errors. In a library of ten PDFs that changes once a year, manual loading may be sufficient. With hundreds of documents daily, synchronization justifies more investment.

ArduMaker is evaluating a RAG node to connect sources and query via API; this article describes integration requirements, not already announced connectors. If you want an integration, specify which source you would use and whether your files are public or private on ArduMaker.

References: Retrieval concepts in Microsoft · RAG security according to OWASP.

  • RAG
  • Google Drive
  • Notion
  • WordPress
  • conectores

On ArduMaker

  • WordPress and WooCommerce: Sites, WooCommerce stores and Elementor on your VPS, with one-click backups and moving in from your current host.

Keep reading

All articles · VPS plans