SEO article from keyword research to WordPress draft

SEO article from keyword research to WordPress draft

An n8n workflow that turns one topic from a form into a researched article and leaves it in WordPress as a draft. DataForSEO
supplies the keyword and search research, Google Gemini plans, writes and edits the article, and Claude's web search finds the
sources it links to. Two Google Sheets tabs keep a record of each run.

What is inside

• workflow.json: the workflow as n8n exports it, 47 working nodes and four sticky notes that label its stages (Input,
  Research, Analysis, Content Creation). It holds no credential.
• README.md: these notes.

What we changed from the original copy

• The Tavily key in the Authorization header of url_Search is now <YOUR_TAVILY_KEY>.
• The author's sheet is gone. Append row in sheet and Append row ask for YOUR_SHEET_ID and find their tab by name:
  Articles and Research Data.
• The original copy had lost every backslash, so two Code nodes would not run at all. Filter Google stopped at
  /^https?:///, and HTML Format stopped at its first closing tag in a regular expression. Mindket put the backslashes back in
  both, and in the smaller spots that still ran but did the wrong thing: /[^\w\s]/g and \n in Filter Google,
  \w in Brainstorm Ideas, \s and \b\w in Filter Keywords, /^\n+/ in Insert Links, the slug in Create a post
  and join('\n') in Append row in sheet. Nothing else in the code changed.

Before you start

• An n8n instance, on n8n Cloud or your own server.
• A Gemini API key from Google AI Studio.
• A DataForSEO account: its API login and API password.
• An Anthropic API key.
• A Tavily API key.
• A WordPress site, and a user there that can write posts, with an application password (Users, Profile, Application
  Passwords).
• A Google account, and a Google Sheet with two tabs (below).

Set it up

1. In n8n, create a workflow, open its menu and choose Import from File. Pick workflow.json.
2. Open Client Data and replace its four values, <<YOUR COMPANY NAME>>, <<YOUR COMPANY WEBSITE>>,
   <<YOUR COMPANY DESCRIPTION>> and <<FEW STRENGTHS ABOUT YOUR COMPANY>>, with your own. Most prompts read them.
3. Add a Google Gemini (PaLM) API credential and select it in 2.5 pro, 2.5 PRO and 2.5 Pro.
4. Add an HTTP Basic Auth credential, using your DataForSEO login as the user and the API password. Select it in
   Keyword Ideas, Search Intent, SERP Google, SERP Bing and Related Keywords.
5. In url_Search, replace <YOUR_TAVILY_KEY> in the Authorization header with your key. The brackets go too, so the
   header reads Bearer, a space and the key.
6. Add an Anthropic credential and select it in Simple_search.
7. Add a WordPress credential with your site's address, the user name and the application password, and select it in
   Create a post.
8. Make the sheet with the two tabs below and copy its ID: the long part of its address between /d/ and /edit. Add a
   Google Sheets OAuth2 credential, select it in Append row in sheet and Append row, and replace YOUR_SHEET_ID in both.
9. Save. Open On Form Submission, use its test address and send one topic. Publish once a draft reaches WordPress.

The sheet

| Tab | Columns |
|---|---|
| Articles | Date, Target Keywords, Primary Goal, Internal Link Reference, Content Type, Title, Meta Description, Article Draft Link, Post ID, Image Gen |
| Research Data | Date, Main Topic \| Keyword, Primary Intent, Internal Link, Primary Goal |

Spell the headers as above: the two nodes map their values to them by name. Image Gen always gets "Not Generated", because
the workflow makes no images.

How a run goes

1. On Form Submission takes the topic, the search intent, a country, a language, the audience, a tone and one page of your
   site. Client Data and Prepare data add your company to it, and gl & hl code turns the country and language into
   DataForSEO's codes.
2. Internal Link Data Prep, an agent, reads your page through url_Search (Tavily's extract endpoint) and returns its
   title, H1, main subheadings and how it relates to the topic.
3. SEO Research Analyst writes five inputs for DataForSEO: a Google query, a Bing query, two short seed keywords and a list
   of intent keywords.
4. Five HTTP nodes call DataForSEO: keyword ideas, search intent, related keywords, and the organic results on Google and
   Bing. Five Code nodes boil each response down to a report, and five Gemini steps summarize each report in Markdown.
5. Merge waits for all five summaries, and SEO Content Silo Strategist turns them into a plan for one article: its focus,
   title, type, goal and target keywords.
6. AI Generated First Article Draft writes it in Markdown, with a placeholder link to your page. EEAT Fortification Agent
   then searches with Simple_search (Claude with web search) and returns three or four sources, each with a suggested
   anchor text.
7. Technical Editor polishes the draft and returns its title, its H2 headings, the body and a meta description. Next,
   Insert Links links each source at the first place its anchor text appears, and HTML Format turns the Markdown into
   HTML with a table of contents and a folding FAQ.
8. Final Checker reads the HTML once more, and Create a post saves the HTML as a WordPress draft. Append row in sheet logs
   the article in Articles, and Append row logs the research in Research Data.

Change the look

• Your company: the four values in Client Data, which most prompts quote.
• Research size: the JSON bodies of the five DataForSEO nodes. Keyword Ideas asks for 44 ideas with a monthly volume over
  50 and a competition of 0.8 or less, SERP Google reads 44 results and SERP Bing 22.
• Voice and structure: the system message of AI Generated First Article Draft. The form's tone and audience reach it only
  through the plan.
• Sources: the system message of EEAT Fortification Agent, which asks for neutral sources such as public agencies and
  industry bodies, and never a competitor.
• Post status: Create a post saves a draft. Change its status to publish only once you trust the output.

Good to know

• Billing: each run makes twelve Gemini calls (agents may make more), five DataForSEO calls, one or more Tavily extracts and
  as many Anthropic calls as the source agent decides. Anthropic bills the web searches inside each call as well. Try one
  topic first.
• The language field does not reach DataForSEO. gl & hl code reads a field named What language should the content be in?,
  which the form does not send, so the research is always in English (en). To pass your pick on, change that name in the
  code to Target Language (DataforSEO). The plan's title and keywords still follow the language you picked.
• The three Gemini nodes carry Pro in their names, but the file sets no model in them, so n8n uses the node's default,
  models/gemini-2.5-flash. Pick Gemini 2.5 Pro in each if you want what the names say. Google's deprecations page, read on 2
  November 2026, gives neither a shutdown date but limits both to accounts that have used them before. A new key
  may need a newer Gemini model picked in all three.
• Simple_search uses claude-sonnet-4-5-20250929. Anthropic's deprecations page, read on 2 November 2026, retires it on 30
  November 2026 and names claude-sonnet-5-5 as the replacement. Pick it in the node before then.
• A source is linked only where its suggested anchor text appears word for word in the edited article, so some sources may
  not show. The debug_info that Insert Links returns lists which ones did.
• The code is written for local service businesses: it looks for "near me" searches, phone numbers written with the area code in
  brackets, and review ratings. Other topics still run, with less to say in those reports.
• Structured Output11 connects to nothing and does no work.
• Who can start it: once published, anyone with the form's address can start a run on your keys. Turn on the form's basic
  authentication or keep the address to yourself.
• Where the data goes: the topic and your company details go to Google, the research terms to DataForSEO, your page's address
  to Tavily and the search words to Anthropic. The article goes to your WordPress site and the log to your sheet.