Draft Google Review Replies with GPT-5, Gmail & Slack (n8n)

Draft Google review replies with GPT-5—then post safely to Google using Gmail & Slack (n8n)

When new Google reviews land on your Google Business Profile, this n8n workflow drafts a reply in your voice with GPT-5, flags anything risky, and only auto-posts when it’s safe. If approval is needed, it routes drafts to Slack—so you stay in control.

What this workflow does

• Checks hourly for new Google reviews on the account and location you select.
• Skips reviews that already have a public reply.
• Reads your configured reply settings (including your sign-off and the star rating threshold that determines when replies should be drafted/posted).
• Uses OpenAI to generate a draft reply in your style, and flags risky content with a reason and a suggested next step.
• Applies guardrails in code to hold drafts whenever they include promises or contact details, exceed the length limit, or come from a failed model call. Reviews that are at or below your configured star threshold are also held for you.
• Logs every draft and status (READY, HOLD, or POST FAILED) to the Review replies tab in your Google Sheet.
• If auto-posting is enabled and the draft is READY, it posts to Google; otherwise it sends the draft to Slack with the reason. If Google refuses a post, a POST FAILED message is also sent to Slack.
• After your job system reports a finished job, a second flow emails the customer a request for an honest Google review, respecting your opt-out flag, sending hours, and timing (days since the last request).

Use cases

• A local contractor responds to every new Google review without writing from scratch.
• A multi-location service business keeps replies consistent with a configured sign-off and star threshold.
• An automation engineer wants approval workflows: auto-post only when safe; otherwise, route to Slack.

Technical details

• n8n nodes/tools: if, set, code, Gmail, Slack, webhook
• Stores draft status in Google Sheets (“Review replies” tab)
• Generates reply drafts with GPT-5 and enforces guardrails for promises, contact details, length limits, and failed model calls