Overview

Featurebase filters unwanted incoming emails, like promotions and spam, before it reaches your team's active Inbox or receives a reply from Fibi AI Agent.

You can review spam in two places:

• Inbox → Spam contains emails stopped by the built-in email spam filter before a Conversation was created

• AI Agent → Spam contains email Conversations that Fibi or a teammate marked as spam

How spam filtering works

Featurebase applies built-in spam filtering to incoming email. When Fibi handles email, Fibi AI spam filtering adds a second layer.

Built-in email spam filtering

The built-in email spam filter checks incoming email for signals such as:

• Unusually high volume – too many new Conversations from the same sender in a short period

• Auto-responders – out-of-office messages and automatic replies, which go to Spam by default

• Sender authentication – authentication failures can contribute to a spam decision, but a failure alone does not automatically send an email to Spam

Emails stopped by this filter appear in Inbox → Spam. They do not create Conversations, appear in reporting, or receive a reply from Fibi.

Trusted email addresses and domains bypass these checks, including the per-sender rate limit.

Fibi AI spam filtering

When Fibi handles email, it also checks new email Conversations for clear:

• Phishing and credential-harvesting attempts

• Scams, including prize, investment, and cryptocurrency scams

• Unsolicited cold sales, SEO, backlink, and link-building outreach

• Unsolicited bulk promotions and marketing newsletters

• Other automated content-spam patterns

If Fibi identifies spam, it does not respond. The Conversation is closed and appears under AI Agent → Spam with a note explaining why it was flagged.

Fibi filtering runs when a live email Workflow uses Let Fibi answer. It applies to new email Conversations and does not retroactively classify existing Conversations.

Configure spam filtering

Go to Settings → Emails → Spam filter to control auto-responders, add custom guidance, test sample email, and manage Trusted senders.

1. Route auto-responders to the Inbox

By default, out-of-office messages and other automatic replies go to Inbox → Spam.

Turn on ‘Route auto-responders to the inbox’ to create Conversations for these messages and deliver them to the normal Inbox instead.

2. Add Spam guidance

Spam guidance lets you describe what your Workspace considers spam in plain language. Use it for recurring unwanted email that the default Fibi classifier does not consistently identify.

To add guidance:

1. Open Spam guidance

2. Add a short title and a specific rule

3. Click Add live to apply it immediately, or Save as draft to keep it inactive

Good guidance describes a specific, observable pattern, for example:

• Treat unsolicited SEO and backlink offers as spam

• Treat cryptocurrency airdrop promotions as spam

• Treat cold sales outreach containing urgent calls to action and several external links as spam

Keep each rule narrow enough that genuine customer questions are unlikely to match it.

3. Add Trusted senders

Trusted senders are legitimate addresses or domains that should always reach your Inbox, including automated systems that send many new emails.

To add a Trusted sender:

1. Open Trusted senders

2. Enter an exact email address such as partner@example.com, or a bare domain such as example.com

3. Click Add

Adding a domain trusts everyone who sends from that domain. Trusted senders bypass spam filtering and the per-sender rate limit, so only add addresses and domains you recognize.

4. Test an email with Preview tester

Use ‘Preview tester’ to check how the AI classifier would handle a sample email before adding or changing guidance:

1. Open Preview tester

2. Enter the sender in From, then add a Subject and Body

3. Click Run preview

4. Review whether the email ‘Would be flagged’ or ‘Would be delivered’, together with the reason

Review Inbox Spam

Open Inbox → Spam to review emails stopped before they created Conversations.

The table shows each email's sender, email address, subject, date, and reason. You can search by sender email, filter by start and end date, and sort by date.

Select an email to open a read-only preview and see ‘Why this was flagged’. Remote images are removed from the preview to prevent tracking, so some email designs may appear without their images.

From there, you can either:

• Restore it by clicking ‘Not spam’ if the email was filtered incorrectly

• Delete it by clicking ‘Delete’ to permanently remove it.

Featurebase processes the original email again, creates the Conversation, and sends it to its intended Inbox. Any follow-up emails held with the same thread are restored with it. Emails left in Inbox → Spam are automatically deleted after 60 days.

Review AI Agent Spam

Open AI Agent → Spam to review email Conversations marked by Fibi or a teammate.

Fibi adds a note explaining why it identified the Conversation as spam. The Conversation remains closed, and Fibi does not respond while it is marked.

Click ‘Mark as not spam’ to clear the flag, reopen the Conversation, and let Fibi resume handling it.

Mark an email Conversation as spam

You can manually mark an unwanted email Conversation after it reaches the Inbox:

1. Open the email Conversation

2. Click the three dots in the header

3. Select ‘Mark as spam’

Marking a Conversation as spam:

• Closes the Conversation and locks its composer

• Excludes it from reporting while it remains marked

• Moves it to AI Agent → Spam

• Does not block the sender

Click ‘Mark as not spam’ to clear the flag and reopen the Conversation for normal handling.

Handle spam from other channels with Workflows

The ‘Mark as spam’ is also available as a workflow step for email Conversations. For unwanted Messenger or other non-email Conversations, use Workflows to identify message content or customer attributes, apply a tag, and close the Conversation automatically.

Workflow-closed Conversations remain regular closed Conversations and do not appear in either Spam view.

FAQs