Blog posts written from Google Sheets and sent to WordPress

What it is

This n8n workflow turns one Google Sheets tab into a small blog desk. Each row is a post. The row’s Action names a step, such as a draft, and a model carries it out. Finished posts go to WordPress when their time comes. Twenty-two nodes do the work, beside thirteen sticky notes.

Who it suits

Solo bloggers and small content teams who plan posts in a spreadsheet and want a model to write them in stages. Skip it if you want a visual editor, since every prompt and every draft lives in sheet cells.

You need a Google account with one spreadsheet, a WordPress site, and an OpenAI key or an OpenRouter key.

Where to use it

It runs inside n8n, and once the workflow is active it wakes by itself every hour. You write in Google Sheets and read the result on your WordPress site.

• n8n Cloud or your own n8n: import the file and switch the workflow on
• Google Sheets: plan, write and review in the three tabs
• WordPress: finished posts appear on the site you name in Settings
• A different model for each step: type its name in the Config tab

Setup

Download saves a zip with workflow.json and Mindket’s setup notes in your downloads folder. Build the three tabs first, because every Google Sheets node finds them by name.

1. In n8n, create a workflow, open its menu, choose Import from File and pick workflow.json.
2. Make a spreadsheet with three tabs named Schedule, Config and Log. Give Config the headers Key and Value, and give Log the headers Date, Type and Message.
3. Give Schedule a header row with Topic, Scheduled, Status, Action, Context and Title, plus one column for each text you want written, such as draft and final.
4. Add a Google Sheets credential and select it in the six Google Sheets nodes.
5. In Settings, replace <YOUR_SPREADSHEET_URL> with the spreadsheet’s address and <YOUR_WORDPRESS_DOMAIN> with your site’s domain, such as myblog.com.
6. In Settings, replace <YOUR_WORDPRESS_USERNAME> and <YOUR_WORDPRESS_APPLICATION_PASSWORD>. Make the password under Application Passwords in your WordPress profile.
7. Add an OpenAI credential and select it in AgentLLM. For OpenRouter, put its Base URL on that credential.
8. In Config, add a row prompt_draft holding your prompt and a row prompt_draft_model holding a model name. Write a column name in double braces, such as {{ Topic }}, to fill in that row’s value.
9. Add a row with a Topic and the Action draft, then press Test workflow. Switch the workflow on when it behaves.

How to use it

Add a row per post and set its Action to the step you want next. The workflow runs that step once, stores the text in the column named like the Action, and sets Status to match. Change the Action to move the post on.

To publish, fill Title and final, set Scheduled to a time such as 2026-12-01 09:00:00, and set the Action to publish. The first hourly run after that time sends the post.

Example

You add a row with the topic Cold brew at home and the Action draft. At the next run the workflow fills the draft prompt, asks the model, writes the text under draft and sets Status to draft.

Later you copy the text into final, add a Title, set a time and change the Action to publish. When that time passes, the post appears on your site and Status reads publish.

For newcomers and experts

If you are new. n8n draws a workflow as boxes joined by lines, and each box, a node, does one job. A row is one post in your sheet. The Action is the step you want done on it. XML-RPC is an older way for a program to talk to a WordPress site.

For experts. PreparedData fills each double-brace field from the row first, then from Config, then from Config under the field’s name followed by the row’s Context. Config has no outgoing line, and other nodes read it by name. The chain has no output parser.

Title: Where to look closer
Node | What it does | What to know
`ScheduleTrigger` | Schedule Trigger 1.2, every hour | `ManualTrigger` starts the same path for tests
`Settings` | Set 3.4, nine fields | The WordPress login sits here in plain text
`fetchConfig` and `Config` | Google Sheets 4.5 read, then Code 2 | The Key and Value rows become one object
`Schedule` | Google Sheets 4.5, read | It reads every row on every run
`PreparedData` | Code 2, builds the prompt, the model and a flag | It looks up `prompt_` plus the Action’s name in `Config`
`IfTakeAction` | If 2.2 | Only rows whose Action differs from Status go on
`IfPromptExists` | If 2.2 | An Action with no prompt in `Config` stops here
`AgentLLM` | OpenAI Chat Model 1.1 | The model name comes from the row’s `model` field
`Basic LLM Chain` | Chain 1.5 with the row’s prompt | It has no tools and no output parser
`RecombinedDataRow` | Code 2, tidies the reply | Plain text goes to the column named like the Action
`SaveBackToSheet` | Google Sheets 4.5, update by `row_number` | Fields with no matching column are ignored
`LogStatus` | Google Sheets 4.5, append to `Log` | It runs after the save
`IfScheduledNow` | If 2.2 | It reads `Scheduled` as `yyyy-MM-dd HH:mm:ss`
`PrepareXmlPost` | Code 2, builds the `wp.newPost` call | It escapes the text and sends `post_status`
`CreatePost` | HTTP Request 4.2, raw `text/xml` | It posts to https, your domain and `/xmlrpc.php`
`HandleXMLRPCResponse` | Code 2 | It returns the post id, or the fault code and text
`PostingSuccessful` | If 2.2 | It checks that a post id came back
`SetToPublish` | Google Sheets 4.5, update by `row_number` | It writes the Status `publish`

Good to know

The original copy sent its posts as drafts and logged a line where the save step needed the row.

• Publishing: WordPress’s documentation for wp.newPost, read on 12 November 2026, lists four arguments and says a post is a draft unless post_status says otherwise. The original added a fifth flag that the method does not take, so its posts would have stayed drafts. Mindket’s copy sends post_status from the published flag in PrepareXmlPost. The flag is 1, so posts go live at once, and 0 keeps them as drafts
• Order of steps: the original put the log step between the writing step and the save step. In n8n’s source for the Sheets Append step, an Append that maps its own columns passes on only those columns, so the save step would not find row_number. In this copy the save comes first and the log follows it
• Model replies in JSON: the original wrote [object Object] into the Action’s own column whenever the model answered in JSON. Mindket’s copy keeps those fields as they are and uses the Action’s column only for plain text
• Login: the user name and application password sit in Settings as plain text, so they are saved with the workflow and show in execution data. WordPress’s application passwords guide, read on 12 November 2026, says one can be used in place of the real password for XML-RPC, so never put your own password there
• The sheet: the file does not hold the spreadsheet, so the tab and column names above come from what the code reads. Check them against your own sheet, and match the headers letter for letter
• Dates: Scheduled must read like 2026-12-01 09:00:00, in the time zone n8n runs in, or the row is never published
• Models: the file names no model, and each Action’s model is a cell in Config. The chain has no tools, so gpt-6-luna suits it. OpenRouter lists models as maker and name, such as openai/gpt-6-luna, so copy the exact name from its catalog
• Repeats: every row is read each hour. If WordPress creates a post but its reply cannot be read, the row is not marked and the next run posts it again, so watch the Log tab
• Not run against Google, WordPress or a model. Mindket ran the flow with stand-ins and saw a draft saved, a post sent as publish, an error logged, and a finished or future-dated row left alone. The stand-in has no DateTime.fromFormat, so that one test used DateTime.fromISO
• Your data: your topics, prompts and drafts go to the model’s provider, the posts and your WordPress login go to your site over https, and the sheet stays in your Google account
• Credentials: the JSON holds none, so add the Google Sheets and OpenAI credentials yourself and type the WordPress login into Settings