diff --git a/communicate/alerts/overview.mdx b/communicate/alerts/overview.mdx
index da5646ac..b3292a3a 100644
--- a/communicate/alerts/overview.mdx
+++ b/communicate/alerts/overview.mdx
@@ -106,6 +106,32 @@ To test the whole flow, from a failing check to a notification:
3. If your threshold is more than one failure, for example "alert after 2 failures", schedule more runs until it's reached. With a time-based threshold, keep the check failing until that time has passed. Alerts are sent when a check changes state and the threshold is met, not for every failed run.
4. Undo your change and run the check again to get a recovery notification.
+## Troubleshooting missing alerts
+
+If a check failed and you didn't get a notification, open the check's overview page. The **Alerts** table lists every alert Checkly generated for the check, such as **Failed** or **Recovered**. The **Notifications** column shows how many notifications went out for each alert. Select it to open them in the [alert notification log](/communicate/alerts/notification-log).
+
+### The failure isn't in the Alerts table
+
+Checkly didn't generate an alert. Check that:
+
+- **The check changed state.** Checkly alerts when a check starts failing, degrades or recovers, not for every failed run. If the check was already failing, more failed runs send nothing new.
+- **The threshold was reached.** With "alert after 2 failures", the first failed run sends nothing. A run and its retries count as one failed run. See [escalation](/communicate/alerts/configuration#escalation-configuration), and check [which level's settings apply](/communicate/alerts/configuration#configuration-inheritance).
+
+### The alert has no notifications
+
+If the **Notifications** column shows `-`, Checkly generated the alert but sent it to no channel. Check that:
+
+- **The check is subscribed to the channel.** You'll find the subscribed channels under **Alert channels** in the check's **Retries & Alerting** tab. If the check's group has **Group override for alert settings** turned on, the group's alert channels are used and the check's own are ignored.
+- **The channel sends this kind of alert.** A channel only sends the **Notification events** selected on it: *a check fails*, *a check degrades* and *a check recovers*. A recovery from a degraded state only goes to channels that have both *a check degrades* and *a check recovers* selected.
+- **Your plan sends to this channel type.** On the Hobby plan, alerts only go to email, Slack, Microsoft Teams, Discord, Telegram and generic webhook channels. Other channels the check is subscribed to, such as SMS, phone call, PagerDuty, Opsgenie or FireHydrant, are skipped and don't appear in the notification log.
+- **The check isn't muted.** Look at **Muted** in the editor of the check and of its group. Muting a group mutes every check in it.
+- **No [maintenance window](/communicate/maintenance-windows/overview#silence-alerts) silenced the alert.** If the check started failing during the window and was still failing when it ended, no failure alert is sent.
+- **The channel was subscribed before the check started failing.** A channel you subscribe while the check is failing gets nothing for that failure, only for the check's next state change.
+
+### Notifications were sent but didn't arrive
+
+The **Notifications** column also counts failed deliveries, for example "1 sent, 1 failed". Select it, or browse the [alert notification log](/communicate/alerts/notification-log), to see each notification's **Delivery status** and **Notification result**: the response Checkly got from the receiving service, or the error. If the delivery status is **Successful**, the receiving service accepted the notification, so look for it there.
+
Use Checkly's alert notification log to analyze your alerting patterns and identify opportunities to reduce noise while maintaining coverage of critical issues.
\ No newline at end of file
diff --git a/platform/groups.mdx b/platform/groups.mdx
index 01edaf9a..b6d853f9 100644
--- a/platform/groups.mdx
+++ b/platform/groups.mdx
@@ -125,6 +125,10 @@ that variable to be overridden at the group level or check level.
- **Retries:** Configure retry strategy for failed checks (fixed, single, linear, exponential)
- **Alert settings:** Set up alert channels for the group - email, SMS, Slack, webhooks, etc. When you enable the group override, these settings replace the alert settings of every check in the group.
+
+Turning on **Group override for alert settings** selects **Use the global account notification settings**. That covers when to alert, not where: the group's **Alert channels** still replace the alert channels of every check in the group. If you select no alert channels for the group, its checks alert no one.
+
+
### Testing & Runtimes
- **Testing:** Run checks in this group as E2E tests locally or from CI/CD pipelines