Understand the Spam Filter
Learn how ConvoRev evaluates the first inbound message from a new contact, protects the normal AI workflow and helps you recover legitimate conversations caught by mistake.
The filter protects the first-message boundary
The filter reviews a new contact’s first inbound message using the message and available business context. Likely abuse is kept away from the normal AI response path; legitimate messages continue into the expected workflow. It is one protection layer, not a replacement for channel policies or human review.
What the spam filter protects
The filter is designed to reduce obvious abuse before it reaches the normal AI response path. Its purpose is to protect the inbox, the agent workflow and the attention of the people who review conversations.
Inbox quality
Reduces irrelevant, automated or malicious first messages in the normal working queue.
AI workflow
Helps prevent inappropriate automated replies to messages that should not enter the customer journey.
Team attention
Keeps obvious abuse from consuming the same review time as legitimate customer conversations.
How classification works
The filter considers the first message and the available business context. Completing the business name and description helps it distinguish an unusual but legitimate customer question from unrelated spam.
The text a new contact sends when beginning the conversation.
The account’s business name, description and other available context.
The message follows the normal path or is kept aside for review.
The first inbound message is evaluated
This is the boundary where there is little or no prior conversation context.
Prior context changes the situation
Messages are treated differently because the account already has conversation history.
Exact signals can change
Classification may evolve as abuse patterns change, so avoid relying on a fixed keyword theory.
What a spam decision does
A message classified as spam is kept away from the normal AI response path and may be hidden or marked for review. It is not treated as a normal customer conversation.
First inbound message arrives.
Message plus available business context.
The conversation proceeds according to the configured workflow.
The contact may be hidden or marked for review without a normal AI reply.
Correct a false positive
If a real customer is caught, restore the contact carefully and verify the complete path. Correcting the label is only the first step; the contact must also reach the expected queue and receive the expected handling.
- 1
Confirm the contact is legitimate
Review the message, channel and business context. Do not open attachments or follow links that may be unsafe.
- 2
Use the available restore action
Use the available unmark or restore action to return the contact to the appropriate workflow.
- 3
Request a new message when needed
If the original content is unavailable, ask the contact to send a new message through the same approved channel.
- 4
Verify the queue and agent
Confirm that the contact reaches the intended queue and that the AI Agent behaves normally.
- 5
Record safe diagnostics
Save the contact, channel, approximate time and message text when it is safe and appropriate to do so.
- 6
Watch for a repeatable pattern
Repeated false positives around the same legitimate topic may mean the business context needs clarification.
Recovery confirmation
The issue is resolved only after a new controlled message reaches the correct queue and follows the expected agent or human path.
Keep the business context clear
Context helps the filter understand which unusual questions may still be relevant. Use a specific description of what the business does and whom it serves, without turning the profile into a list of every possible phrase.
“We help customers.”
This provides little information for separating a niche customer question from unrelated content.
Describe the actual business
State the core services, products, audience and legitimate types of inquiries in plain language.
- Use the complete business name.
- Describe the main products or services.
- Identify the audience or locations when relevant.
- Include legitimate niche inquiry types.
- Remove outdated or conflicting descriptions.
- Retest a safe example after material changes.
Why the filter may not have an off switch
A minimum protection layer helps prevent automated abuse from consuming team attention or triggering inappropriate replies. Disabling protection for every new contact would remove that boundary for legitimate traffic and abuse alike.
Maintain a consistent minimum layer against obvious automated or malicious messages.
Use safe diagnostics and report repeatable patterns instead of removing protection globally.
Separate inbox filtering from channel restrictions
The ConvoRev inbox filter does not control the sending reputation or policy enforcement of WhatsApp, Meta, a mobile carrier or another channel. These issues have different owners and recovery paths.
Evaluates the first inbound message from a new contact.
Applies the channel provider’s sending, account and usage rules.
A new contact is hidden, marked or kept away from the normal AI response path.
A number or account is limited, restricted or unable to send as expected.
The conversation, contact record and available spam or review area.
The affected channel account, policy notice and provider process.
Restore the legitimate contact and verify the ConvoRev workflow.
Follow the external channel’s recovery or appeal process.
Capture safe diagnostic details
When a suspected false positive cannot be restored, collect only what is needed to locate and reproduce the issue. Avoid forwarding potentially harmful content.
- Account and AI Agent
- Contact identifier
- Affected channel
- Approximate date, time and timezone
- Safe message text or a redacted description
- Expected path and observed result
- Restore action already attempted
- Passwords or access credentials
- Malicious attachments
- Unnecessary private customer records
- Unknown executable files
- Sensitive information unrelated to diagnosis
When filtering does not behave as expected
A legitimate new customer was marked as spam
Confirm the message is safe, use the available unmark or restore action, and verify that a new message reaches the correct queue. Record the channel, approximate time and safe message text if the issue repeats.
The same legitimate topic is repeatedly filtered
Review the business name and description for missing or vague context. Add a clear explanation of the legitimate service, product or inquiry type, then retest a safe example and report the pattern if it continues.
An existing contact’s message behaved differently
Existing contacts already have conversation context and may be treated differently. Confirm the correct contact record, conversation history, assignment and agent workflow before attributing the result to the first-message filter.
I cannot find the original message
Check the available spam, hidden or review area and the contact record. If the original content is unavailable and the contact is known to be legitimate, restore the contact and request a new safe message.
A number or channel account is restricted
Treat this as a channel reputation or policy issue rather than an inbox-filter decision. Review the channel notice and follow its recovery process. ConvoRev support can help identify where the workflow stopped.
The restore action does not return the contact to the expected path
Record the account, contact, channel, approximate time, expected queue and observed result. Confirm the contact’s assignment and configured entry point, then contact ConvoRev support with the safe diagnostic details.
Send one safe example
Include the account, contact, channel, approximate time and timezone, safe or redacted message text, expected result, observed result and restore action attempted. Do not forward malicious attachments or credentials.