LinkedIn job search with a cover letter for each role

LinkedIn job search with a cover letter for each role

An n8n workflow for job hunting. Send a form with a place, a job title and your CV as a PDF. An Apify Actor collects
matching LinkedIn listings, gpt-4o-mini rates each one against your CV and writes a cover letter, and every job lands
in a Google Sheet.

What is inside

• workflow.json: the workflow as n8n exports it, nine nodes, one of them a sticky note. It holds no credential.
• README.md: these notes.

What we changed from the original copy

• The Authorization header of Linkedin Job Scrapper held a bare placeholder. It now reads <YOUR_APIFY_TOKEN>.
• The author's spreadsheet is gone. Google Sheets1 asks for YOUR_SHEET_ID and finds its tab by the name Sheet1.

Before you start

• An n8n instance, on n8n Cloud or your own server.
• An Apify account and its API token, with credit for the LinkedIn jobs Actor the file calls by its ID,
  BHzefUZlZRKWxkTck. Read its page in Apify Store first: some Actors there charge a monthly rental on top of usage.
• An OpenAI API key.
• A Google account, and a Google Sheet with a tab named Sheet1 and these headers in row 1: Link, Title,
  Job Description, Date, Company Name, Cover Letter.

Set it up

1. In n8n, create a workflow, open its menu and choose Import from File. Pick workflow.json.
2. Make the sheet, then copy its ID: the long part of its address between /d/ and /edit.
3. Add a Google Sheets OAuth2 credential, select it in Google Sheets1 and replace YOUR_SHEET_ID with the ID.
4. Add an OpenAI credential and select it in OpenAI4 and OpenAI3.
5. In Linkedin Job Scrapper, replace <YOUR_APIFY_TOKEN> in the Authorization header with your token. The
   brackets go too, so the header reads Bearer, a space and the token.
6. Save, open the form's test address and fill in the three fields.
7. Press Publish when the first rows look right.

How a run goes

1. On form submission takes Location, Title and Your Resume. All three are required.
2. Extract from File reads the PDF's text.
3. Linkedin Job Scrapper posts the place and the title to the Apify Actor and waits for its listings.
4. Loop Over Items sends one job at a time on. Edit Fields1 keeps its link, title, description, date posted,
   company and benefits.
5. OpenAI4 rates the job against the CV's text and answers with score.
6. OpenAI3 writes the letter from the job and the CV and answers with cover_letter.
7. Google Sheets1 adds the row, or updates the one with the same Link. Then the loop takes the next job.

Keep the rating

OpenAI4 rates every job, but nothing reads its answer: the sheet has no column for it, and a letter is written
whatever the score. To keep it, add a Score header and map this to it in Google Sheets1:

{{ $('OpenAI4').item.json.message.content.score }}

To skip weak matches, put an If node between OpenAI4 and OpenAI3 and send the low scores back to
Loop Over Items. The prompt asks for a score out of 5, while its points add up to 10, so settle on one scale in the
prompt before you filter on it.

Change the result

• What counts as a fit: the system message of OpenAI4. It hands out points for matching skills, the right level, a
  remote role, past roles and meeting the qualifications.
• The letter: the system message of OpenAI3, a single line asking for a letter shaped to the job.
• The search: the JSON body of Linkedin Job Scrapper, which sends the place and the title. The Actor's page lists
  the other fields it takes.
• The columns: the mapping in Google Sheets1. Benefits is kept in Edit Fields1 and never saved.
• The model: both OpenAI nodes use gpt-4o-mini.

Good to know

• Each job costs two gpt-4o-mini calls, and Apify bills the Actor's run.
• The body sets no count, so the Actor's default decides how many jobs come back. The call asks for Apify's
  residential proxies.
• Apify's API reference gives a synchronous call like this one 300 seconds. A search that takes longer ends without
  its listings, so keep the first searches narrow.
• Extract from File reads a PDF's text layer. A scanned CV gives the models no text.
• A place or a title with a double quote in it breaks the JSON body.
• OpenAI's deprecations page, read on 2 November 2026, lists GPT-4o mini (gpt-4o-mini-2024-07-18) for removal from
  the API on 14 January 2027 and names gpt-6-luna as its replacement. Change the model in both OpenAI nodes before
  then.
• Once published, anyone with the form's address can start a search on your accounts. Keep the address to yourself,
  or turn on the form trigger's basic authentication.
• Your CV's text and each job description go to OpenAI. The place and the title go to Apify.
• The Apify token sits in a header, so take it out before you share the workflow.