Score weekly SEO topic ideas and approve them from Telegram

The poster runs a personal finance site and wanted a weekly answer to what to write next. Every Monday an n8n workflow reads seed keywords from a Google Sheet, collects Google Autocomplete suggestions for each, pulls trending questions from Reddit over RSS, and checks the site's existing post titles to avoid repeats. It sends all of that to Gemini in a single call, which returns ten scored topic ideas as structured JSON. The workflow logs them to the sheet and sends the top five to Telegram, each with an Approve button. A second workflow marks a topic approved in the sheet when the button is tapped.

The workflow runs end to end and the poster published both workflows on GitHub. Most of the effort went into bugs: button data lost its separators when pasted from a table, date-only ids made three test runs in a day write duplicate rows and send three confirmations, and a default Update Row node would have blanked real columns. The author gives no numbers on traffic or posts written yet.

The scores are the model's guesses from free public signals, not real search volumes, so the poster checks numbers before writing. One batched model call keeps it inside free-tier limits. Each Telegram bot can serve only one trigger, so each workflow project needs its own bot.