Understanding how third-party source filtering works for Front AI features

When you connect a third‑party knowledge source (like Notion, Google Drive, etc.), Front evaluates the content found during the sync. Some files are approved and used by Front AI features, while others are rejected (filtered out) to protect quality and prevent irrelevant or unsafe content from influencing AI.

This article explains:

• How sync status works (so you know when content is ready)

• What “approved” vs. “rejected” means

• Why some files are rejected

• What you can do to improve approval rates

When you start a sync, Front runs a pipeline that:

1. Fetches files/pages from the connected source

2. Evaluates each item with a filtering step

3. Processes approved items so they can be used by Front AI

A sync can take time, and content may not be ready immediately when the first sync step finishes. Use the Status column in the Knowledge sources tab to determine when your content is ready.

Front will display the following sync statuses:

• Syncing: Your source is still being processed

• Live: Processing is complete and content is available to Front AI

• Failed: Something prevented the sync from completing

If your source is marked as ready but content is missing, check whether the sync is still in progress.

Filtering is designed to:

• Improve answer quality by focusing on content that’s helpful to customer support

• Reduce noise from low‑signal documents (drafts, link hubs, boilerplate)

• Prevent risky content (for example, internal notes that should not be used for customer responses)

Once your knowledge source sync finishes, you can review content or files that are approved or rejected.

Approved content is used to improve Front AI responses. In general, this is content that is:

• Customer‑facing help documentation

• Product documentation and how‑to guides

• Troubleshooting steps and known‑issue workarounds

• Policies (billing, legal, compliance)

Rejected content is filtered out and is not used by Front AI. In general, this is content that is:

• Internal meeting notes or team discussions

• Internal agent standard operating procedures (SOPs) or playbooks intended only for internal operations

• Marketing or promotional pages with little customer support value

• Navigation / index‑only pages (tables of contents, link hubs)

• Drafts, placeholders, or near‑empty pages

• Highly technical internal engineering documentation

When content is rejected, Front groups the reason into one of the following categories:

• Internal meeting notes

• Internal agent instructions

• Marketing / promotional

• Navigation / index only

• Incomplete / draft / empty

• Technical / engineering internal

• Other

These categories help you quickly understand what kinds of content are being rejected and what to fix.

If a large number of files are rejected, the fastest way to improve outcomes is usually to change what you sync, not to add more content.

1.  Scope the source to a curated folder or space

Connect a location that contains mostly customer‑facing documentation (help center, support documentation, runbooks intended for support teams).

2. Separate internal notes from support documentation

If your source includes meeting notes, brainstorming documents, or internal-only strategy pages, move those out of the synced scope.

3. Publish and complete draft documents

Drafts or placeholders are commonly rejected. Finalize those documents and run a new sync.

4. Expect index / table-of-contents pages to be rejected

Index-only pages are often rejected because they don’t contain enough standalone information. This is usually expected.

Filtering transparency is designed for third‑party application sources. Other knowledge sources (like Front knowledge bases or public websites) may have different syncing and processing behaviors.

No. Rejected items are tracked using lightweight metadata (for example: title, URL, category, and a short reason) rather than storing rejected content.

Not at this time. The best way to influence outcomes is to adjust the scope of your source and improve the quality and structure of the content being synced.

Common reasons include:

• Your source includes a lot of drafts or internal notes

• Many pages are navigation-only

• The connected scope contains mixed content (support documentation + marketing materials + internal documentation)

We recommend focusing on syncing a smaller, curated set of high-signal documentation first.