Auto-generate & schedule LinkedIn posts from Google Sheets with GPT-4o

Target audience

Solo creators, PMs, and content teams who queue LinkedIn ideas in Google Sheets and want them posted on a fixed schedule with AI-generated copy.

How it works

The workflow runs on a schedule (Mon/Wed/Fri at 09:30). It pulls the first Google Sheet row with Status = Pending, generates a LinkedIn-ready post from Post title using an OpenAI prompt, publishes to your LinkedIn profile, then updates the same row to Done and writes the final post back to the sheet.

Prerequisites (use your own credentials)

• Google Sheets (OAuth2) with access to the target sheet
• LinkedIn OAuth2 tied to the account that should post — set the person field to your profile’s URN in the LinkedIn node
• OpenAI API key for the Chat Model node
• Store secrets in n8n Credentials. Never hard-code keys in nodes.

Google Sheet structure (exact columns)

Minimum required columns

• id — unique integer/string used to update the same row later
• Status — allowed values: Pending or Done
• Post title — short prompt/topic for the AI model

Recommended columns

• Output post — where the workflow writes the final text (use this header, or keep your existing Column 5)
• Hashtags (optional) — comma-separated list (the prompt can append these)
• Image URL (optional) — public URL; add an extra LinkedIn “Create Post” input if you post with media later
• Notes (optional) — extra hints for tone, audience, or CTA

Example header row

id | Status | Post title | Hashtags | Image URL | Output post | Notes

Example rows (inputs → outputs)

1  | Pending | Why I moved from Zapier to n8n | #automation,#nocode |  |  | Focus on cost + flexibility
2  | Done    | 5 lessons from building a rules engine | #product,#backend |  | This is the final posted text... |

Resulting Output post (for row 1 after publish)

I switched from Zapier to n8n for three reasons: control, flexibility, and cost.
Here’s what changed in my stack and what I’d repeat if I had to do it again.

#automation #nocode

> If your sheet already has a column named Column 5, either rename it to Output post and update the mapping in the final Google Sheets Update node, or keep Column 5 as is and leave the node mapping untouched.

Step-by-step

1. Schedule Trigger Runs on Mon/Wed/Fri at 09:30.
2. Fetch pending rows (Google Sheets → Get Rows) Reads the sheet and filters rows where Status = Pending.
3. Limit Keeps only the first pending row so one post goes out per run.
4. Writing the post (Agent + OpenAI Chat Model + Structured Output Parser) Uses Post title (and optional Notes/Hashtags) as input. The agent returns JSON with a post field. Model set to gpt-4o-mini by default.
5. Create a post (LinkedIn) Publishes {{$json.output.post}} to the configured person (your profile URN).
6. Update the sheet (Google Sheets → Update) Matches by id, sets Status = Done, and writes the generated text into Output post (or your existing output column).

Customization

• Schedule — change days/time in the Schedule node. Consider your n8n server timezone.
• Posts per run — remove or raise the Limit to post more than one item.
• Style and tone — edit the Agent’s system prompt. Add rules for line breaks, hashtags, or a closing CTA.
• Hashtags handling — parse the Hashtags column in the prompt so the model appends them cleanly.
• Media posts — add a branch that attaches Image URL (requires LinkedIn media upload endpoints).
• Company Page — switch the person field to an organization URN tied to your LinkedIn app scope.

Troubleshooting

• No post created

  • Check the If/Limit path: is there any row with Status = Pending?
  • Confirm the sheet ID and tab name in the Google Sheets nodes.

• Sheet not updating

  • The Update node must receive the original id. If you changed field names, remap them.
  • Make sure id values are unique.

• LinkedIn errors (403/401/404)

  • Refresh LinkedIn OAuth2 in Credentials.
  • The person/organization URN is wrong or missing. Copy the exact URN from the LinkedIn node helper.
  • App lacks required permissions for posting.

• Rate limit (429) or model errors

  • Add a short Wait before retries.
  • Switch to a lighter model or simplify the prompt.

• Post too long or broken formatting

  • LinkedIn hard limit is ~3,000 characters. Add a truncation step in Code or instruct the prompt to cap length.
  • Replace double line breaks in the LinkedIn node if you see odd spacing.

• Timezone mismatch

  • The Schedule node uses the n8n instance timezone. Adjust or move to a Cron with explicit TZ if needed.

Need to post at a different cadence, or push two posts per day? Tweak the Schedule and Limit nodes and you’re set.