N8N Workflow: Auto-Mark Unread Gmail Junk with GPT-5

Auto-Mark Unread Gmail Junk with GPT-5 (plus Undo)

Keep your Gmail Inbox clean automatically: this n8n workflow scans unread Gmail messages, uses OpenAI GPT-5 to spot high-confidence junk from unknown senders, moves them to Gmail Spam, and sends you a Telegram report with a one-tap Undo.

What this workflow does

• Runs on a schedule or on demand: starts daily via a cron schedule or triggers from a Telegram message/callback button.
• Builds a “known” sender list: pulls unread messages from your Gmail Inbox and also uses recent Gmail Sent mail to gather contacts/domains you’ve already interacted with.
• Pre-filters to reduce false positives: excludes emails from people you have written to, and only sends remaining candidates to OpenAI GPT-5 for junk classification.
• Uses confidence scoring: keeps borderline/low-confidence items in the Inbox, while preparing an undo batch for only the high-confidence junk.
• Moves selected emails to Spam: moves each chosen message to Gmail Spam and removes Inbox/Unread labels.
• Logs every run: appends a run entry to Google Sheets.
• Telegram reporting + one-tap Undo: sends a report listing what moved and why, including an inline Undo button. If tapped before expiry, the workflow restores the batch from Spam back to Inbox and confirms the result in Telegram.

Use cases

• Reduce inbox clutter from newsletters, spam, and suspicious senders without manually checking Gmail.
• SaaS operators automate low-risk cleanup of unread messages while keeping uncertain items untouched.
• Automation engineers need auditability via Telegram notifications and Google Sheets run logs.

Technical details

• n8n workflow nodes: if, set, code, gmail, switch, telegram
• Integrations: Gmail (OAuth2 with permission to read/modify labels), OpenAI (model set to gpt-5), Telegram bot, Google Sheets run logging