Skip to content

Notifications

Overwatch pushes results out as notifications: emails and chat webhooks, configured per organization with optional per-site overrides. All notification-settings endpoints are session-only (owner/admin) — receiver config is dashboard administration, not something a CMS key can change.

Each kind has its own recipient list, so on-call, editorial, and stakeholder audiences stay separate:

Kind Fired when Receivers
Uptime incidents A monitor goes down or recovers (multi-region consensus) uptimeRecipients, uptimeWebhooks
Crawl summaries A crawl finishes (broken links + accessibility summary) crawlReportRecipients
Performance regressions A site’s Core Web Vitals verdict flips from pass to fail perfAlertRecipients, perfAlertWebhooks
Search traffic drops A site’s Google Search clicks fall at least 35% week over week (Search Console) searchAlertRecipients, searchAlertWebhooks
AI access lost A crawl finds the site blocked from AI search after a run that wasn’t (AI readiness) aiAlertRecipients, aiAlertWebhooks
Pages dropped out of search A crawl finds pages newly blocked from search (seo.indexability_lost) searchAlertRecipients, searchAlertWebhooks
Old pages broke after launch After a relaunch, an old page with search traffic that worked now 404s (launch.pages_broken) searchAlertRecipients, searchAlertWebhooks
Stakeholder reports The scheduled branded PDF report goes out reportRecipients

Recipients are either org members ({ "kind": "user", "userId": … } — resolved to the member’s current email at delivery time) or external addresses ({ "kind": "email", "email": … }). Webhooks support slack, discord, and teams (formatted for the chat app) plus generic http (a structured JSON POST to any endpoint).

Generic http webhooks for pages dropped out of search receive:

{
"event": "seo.indexability_lost",
"site": { "id": "…", "hostname": "example.com" },
"runId": "…",
"total": 2,
"issues": [
{
"id": "…",
"rule": "meta-noindex",
"pageUrl": "https://example.com/pricing",
"detail": "This page is in your sitemap but blocked from search"
}
],
"reportUrl": "https://…/sites/…/scanner/seo",
"message": "…"
}

issues is capped at 50, and total counts every page affected. Each id is the stable issue id served by GET /v1/sites/{id}/seo-issues. See SEO checks for when this alert fires.

Generic http webhooks for old pages broke after launch receive:

{
"event": "launch.pages_broken",
"site": { "id": "…", "hostname": "www.example.com" },
"pages": [{ "path": "/visit/opening-hours", "clicks": 43, "status": 404 }],
"reportUrl": "https://…/sites/…/scanner/launch",
"message": "…"
}

Each page alerts once, when it goes from working to broken; it alerts again only if it’s fixed and then breaks again. See Launch readiness.

Settings live at two scopes:

  • Org defaults — GET/PATCH /v1/notifications: the baseline receivers for every site.
  • Per-site overrides — GET/PATCH/DELETE /v1/sites/{id}/notifications: client isolation, so one site’s alerts stay out of another client’s channels. An override starts from empty receivers with includeOrgDefaults: true — the org defaults still apply until you set includeOrgDefaults: false, so a partial save can only narrow delivery deliberately, never by accident. Deleting the override reverts the site to the org defaults.

Uptime monitors additionally carry their own alertEmails/alertWebhooks, delivered on top of whatever the settings resolve to.

Changes take effect on the next delivery — settings are read at send time.

POST /v1/notifications/test delivers a one-off test message to a single email or webhook — synchronously, so a 200 means it arrived and a 502 carries the delivery error. Targets don’t need to be saved first, so you can verify a webhook URL before committing it.