# How to Block Spam Subscribers With ZazzyAgent AI Agents Not every message that reaches your business is a genuine customer enquiry. Some conversations may contain: * Scam links * Repeated promotional junk * Automated flooding * Malicious messages * Obvious spam ZazzyAgent AI Agents can be configured to block a subscriber when the message is clearly genuine spam. ## The action Use: ```text ##block_subscriber## ``` This action does not require a resource ID. ## Use blocking carefully Blocking is a strong action. A customer may be: * Angry * Frustrated * Complaining * Using informal language * Asking an unusual question None of those automatically mean the person is spam. Don't block legitimate customers simply because their message is difficult. ## What counts as genuine spam? Examples include: > Repeated scam links > Automated junk messages > Obvious malicious content > Repeated messages with no legitimate customer purpose The exact rules should match your business's needs. ## Write a specific instruction Bad: ```text If the customer is annoying, block them. ``` This is far too subjective. Better: ```text Only block the subscriber when the message is clearly spam, malicious, automated flooding, or repeated junk. Do not block legitimate customers because they are angry, rude, confused, or complaining. ##block_subscriber## ``` ## Don't use blocking for unresolved support issues Customer: > Your product arrived damaged and I'm very angry. This is not automatically spam. The right response is support. You might instead: ```text Assign the conversation to Support. ##assign_human_role## : 3 ``` See [How to Hand Off AI Conversations to Human Agents in ZazzyAgent](https://blog.zazzyagent.com/ai-agent-human-handoff-zazzyagent). ## Spam vs irrelevant message These are different. ### Irrelevant Customer: > What's the weather today? You may respond, clarify, or skip the reply depending on your setup. ### Spam Customer repeatedly sends: > CLICK THIS SCAM LINK NOW!!! You may block if your rules identify it as genuine spam. ## Add a label before blocking? In some workflows, you may want to classify the customer before applying a stronger action. For example: > Spam The exact setup depends on your desired workflow. Don't create unnecessary actions just for the sake of logging. ## Test your spam rules carefully Use test messages that resemble: ### Genuine customer > I'm unhappy with my order. Should not be blocked. ### Genuine customer using strong language > This is ridiculous. I want a refund. Should not automatically be blocked. ### Obvious spam > Buy crypto now! Click this link! May be blocked according to your defined rules. ### Flooding Repeated identical junk messages may justify blocking. ## Avoid over-aggressive AI rules AI can misunderstand intent. That's why the System Prompt should use conservative language for irreversible or disruptive actions. Prefer: > Only block when the message is clearly genuine spam. rather than: > Block anything suspicious. ## What happens after blocking? The subscriber is blocked according to the action's behavior and your connected channel configuration. Because blocking affects future interaction, use it only when the criteria are strong. ## Block vs skip reply ### Skip ```text ##skip_reply## ``` The agent doesn't send a reply. ### Block ```text ##block_subscriber## ``` The subscriber is blocked. Choose the lighter action when blocking isn't necessary. ## Example System Prompt ```text You are a customer-support AI Agent. Handle legitimate customer conversations normally. For messages that are clearly scam, malicious spam, or repeated automated junk: Block the subscriber. ##block_subscriber## Do not block customers because they are frustrated, angry, rude, confused, or complaining. If a legitimate customer needs help, continue the conversation or transfer them to human support. ``` ## The principle Treat blocking as a **last-resort spam control**, not as a general way to deal with difficult conversations. A difficult customer may still be a valuable customer.