Daily Reddit trends read by AI and logged to Google Sheets

Daily Reddit trends read by AI and logged to Google Sheets

An n8n workflow that watches one subreddit. Every morning it fetches the 20 hottest posts, keeps the ones with a score above 3, and sends the top 15 to an OpenAI model. The model names three trending topics, suggests two content ideas and rates the mood. Each run adds one row to a Google Sheet, so you can follow how the topics change.

What is inside

• workflow.json: the workflow as n8n exports it, 6 nodes beside 1 sticky note. It holds no credential.
• README.md: these notes.

What we changed from the original copy

Mindket repaired the copy so it runs today:

• It fetched Reddit's public .json address with no sign-in. A plain request to it got HTTP 403 and a "Blocked" page when Mindket tried it on 12 November 2026. Reddit's Data API terms, last revised 20 July 2026, tell developers to use the access Reddit gives them, such as an OAuth token, and not to mask the user agent. Mindket's copy calls oauth.reddit.com with n8n's Reddit OAuth2 credential, and its user agent names your own Reddit user: <YOUR_REDDIT_USERNAME>.
• The subreddit came from one environment variable and the sheet id from another. Mindket's copy puts <YOUR_SUBREDDIT> in the address and <YOUR_GOOGLE_SHEET_ID> in the sheet node, so nothing depends on server settings.
• The AI step was written as n8n's older OpenAI node at version 1.8, a version that node does not have. Mindket rebuilt it as the current OpenAI node, set to send a message and return JSON.
• The model was gpt-4o-mini. OpenAI's deprecations page, read on 12 November 2026, says it shuts down on 14 January 2027 and names gpt-6-luna as the replacement. Mindket's copy uses gpt-6-luna.
• The temperature of 0.5 is gone, since OpenAI's page for the new model does not mention it. The token limit went from 800 to 2000, because OpenAI's guide says a reasoning model's thinking counts against that limit and a run can end before it writes any answer.
• The code that reads the answer now accepts JSON that n8n has already parsed.
• The file's three tags are gone, and its sticky note is reworded.
• No credential is inside.

Before you start

• An n8n instance, on n8n Cloud or your own server.
• A Reddit account, and an app created in it, so you have a client id and a client secret for the Reddit OAuth2 credential.
• An OpenAI API key.
• A Google account, and a Google Sheet with a tab named Lite Log. Its first row holds the headers Date, Posts, Trends, Content Ideas and Sentiment.

Set it up

1. In n8n, create a workflow, open its menu and choose Import from File. Pick workflow.json.
2. Add a Reddit OAuth2 API credential with your client id and secret, and select it in Reddit: Fetch Posts.
3. In that node, replace <YOUR_SUBREDDIT> in the address with the name of your subreddit, without r/, and replace <YOUR_REDDIT_USERNAME> in the User-Agent value with your Reddit username.
4. Add an OpenAI credential and select it in AI: Quick Analysis.
5. Add a Google Sheets credential and select it in Log to Sheets. Replace <YOUR_GOOGLE_SHEET_ID> with the id from your sheet's address.
6. Run the workflow once by hand and check the new row in the Lite Log tab. Then publish it.

How a run goes

1. Run Daily (9 AM) starts a run every day at 9:00 in the time zone of your n8n.
2. Reddit: Fetch Posts fetches the 20 hottest posts of the subreddit. It tries up to 2 times, 5 seconds apart.
3. Parse & Rank Posts keeps each post's title, score, comment count, address and subreddit. It drops posts with a score of 3 or less, sorts the rest by score and keeps the top 15.
4. AI: Quick Analysis sends the numbered titles and scores to OpenAI. The system message asks for three trends, two content ideas, each with a platform, and the sentiment, as JSON. It tries up to 3 times, 2 seconds apart.
5. Merge Results joins the parsed posts with the model's answer. If the answer cannot be read, it fills in empty trends and ideas and the sentiment unknown.
6. Log to Sheets appends a row with Date, Posts, Trends, Content Ideas and Sentiment to the Lite Log tab. It tries up to 2 times.

Change the look

• The subreddit and the number of posts: the address in Reddit: Fetch Posts. Replace hot in the address with new or top for another listing.
• How many posts count: the score limit and the post limit in the code of Parse & Rank Posts.
• What the model returns: the system message in AI: Quick Analysis. It names Twitter as the platform for the ideas, so change that line for another network.
• The time: the cron expression in Run Daily (9 AM).
• The columns: the five values in Log to Sheets.

Good to know

• Reddit's terms keep commercial use for a separate agreement with Reddit. This workflow suits your own monitoring, and you should read the terms before you use it for anything else.
• Only post titles and scores go to OpenAI. The rest of each post stays in n8n.
• A quiet subreddit may leave only a few posts, or none. The run still calls the model with an empty list, so the row can hold empty ideas.
• Each run makes one model call, and up to three if the first fails.
• The file has Save manual executions turned on in its settings, so test runs keep their data.
• It runs only when the workflow is published. Until then, run it by hand.
• The JSON holds no credential.