Business profile and keyword report from a website

Business profile and keyword report from a website

An n8n workflow that turns one web address into a written business profile and two keyword lists. You give a form the
address, a country and a language. A Gemini agent studies the site with seven tools and saves 14 fields about the business
to a sheet. DataForSEO then adds the keywords the page already ranks for and new ones worth a page of their own. Last, the
workflow fills a copy of your Google Doc template, keeps it in Google Drive as a document and a PDF, and emails you both links.

What is inside

• workflow.json: the workflow as n8n exports it, 35 working nodes and seven sticky notes that label its parts. It holds
  no credential.
• README.md: these notes.

What we changed from the original copy

• Copy file and Make Copy of Template pointed at its author's own sheet and document. They now ask for
  YOUR_SHEET_TEMPLATE_ID and YOUR_DOC_TEMPLATE_ID.
• get_internal_links, serp_Search and ai_Mode called a sub-workflow in its author's n8n. They arrive empty, so you
  pick your own (below).
• The Firecrawl header in URL_search reads Bearer <YOUR_FIRECRAWL_KEY>.
• The export had dropped the backslashes from three prompts, and we put them back. In Business Analyst, the line break
  the agent must write between questions and between journey stages had become a bare n, and it reads \n again. In
  online_Search and deep_research, the math markers read \( \) and \[ \] again.

Before you start

• An n8n instance, on n8n Cloud or your own server.
• A Google account for Sheets, Docs, Drive and Gmail.
• A Gemini API key from Google AI Studio.
• An Anthropic API key. Its web search must be on, which it is unless an admin turned it off in the Claude Console.
• A Firecrawl API key.
• A DataForSEO account, for its API login and password.
• A sub-workflow of your own for a site's links and for Google results (below).

Set it up

1. In n8n, create a workflow, open its menu and choose Import from File. Pick workflow.json.
2. Make the sheet template with the three tabs below. Copy its ID, the long part of its address between /d/ and /edit,
   and put it in Copy file in place of YOUR_SHEET_TEMPLATE_ID.
3. Make the document template with the 16 placeholders below, and put its ID in Make Copy of Template in place of
   YOUR_DOC_TEMPLATE_ID.
4. Add Google credentials. Sheets: insert_row_in_data_table, Insert Page Ranked Keywords,
   Insert Suggested Page Keywords and Get Business Data. Drive: Copy file, Make Copy of Template, Generate Doc,
   Add Doc to Drive, Generate PDF and Add PDF To Drive. Docs: Change Variables in Doc. Gmail: Send email.
5. In Send email, replace YOUR_EMAIL_ID in Send to with the inbox that should get the reports.
6. Add a Gemini credential and select it in Business Analyst.
7. Add an Anthropic credential and select it in online_Search and deep_research. Set both to claude-sonnet-5-5
   before 30 November 2026 (Good to know).
8. In URL_search, replace <YOUR_FIRECRAWL_KEY> with your key. The brackets go too.
9. Add an HTTP Basic Auth credential with your DataForSEO login and password. Select it in Get Ranked Keywords,
   Keyword for Site and Search Intent.
10. Build the sub-workflow and pick it in get_internal_links, serp_Search and ai_Mode.
11. Save, send one test through the form's test address, then publish the workflow.

The two templates

The sheet template is copied for every run, and each copy keeps that run's data. Spell the headers exactly as here.

| Tab | Columns |
|---|---|
| Business Data | session_id, client_name, client_country, client_language, client_website, client_description, target_audience_personas, brand_personality_matrix, unique_value_proposition, primary_competitors, people_ask, customer_journey, customer_persona_trait, eeat_signal_integration, GEO_tactic, call_to_action |
| Page Ranked Keywords | keyword, search_volume, cpc, competition_level, estimated_traffic, session_id |
| Suggested Page Keywords | keyword, search_volume, cpc, competition, intent, intent_probability, session_id |

The document template is your report's layout. Write these placeholders where each value should go, braces included:
{{client_name}}, {{client_website}}, {{client_country}}, {{client_language}}, {{date_time}},
{{client_description}}, {{target_audience_personas}}, {{brand_personality_matrix}},
{{unique_value_proposition}}, {{people_ask}}, {{customer_journey}}, {{customer_persona_trait}},
{{eeat_signal_integration}}, {{GEO_tactic}}, {{call_to_action}} and {{ranked_keywords_sheets_link}}. The last one
becomes a link to the run's copy of the sheet, where the keyword lists are.

The sub-workflow

Three of the agent's tools call one sub-workflow, and function tells it which job is wanted. Start yours with a When
Executed by Another Workflow trigger that accepts these five fields:

{
  "function": "internal_links, serp_search or ai_mode",
  "url": "<the form's Page URL>",
  "keyword": "<the agent's search words>",
  "location_code": "<DataForSEO's number for the country>",
  "language_code": "<DataForSEO's code for the language>"
}

get_internal_links sends function and url only. serp_Search and ai_Mode send the other four. Return text or JSON
the agent can read. The file says nothing about what its author's version did, so this is one way to build it, from
DataForSEO's and Firecrawl's API pages read on 2 November 2026:

internal_links  POST https://api.firecrawl.dev/v2/map
serp_search     POST https://api.dataforseo.com/v3/serp/google/organic/live/advanced
ai_mode         POST https://api.dataforseo.com/v3/serp/google/ai_mode/live/advanced

How a run goes

1. Submit Form takes the page's address and picks from two lists. Get Country & Language Code turns the
   country and the language into DataForSEO's codes, and Session ID makes a run ID such as cc48213.
2. Copy file copies your sheet template for this run, named after the address.
3. Business Analyst runs Gemini 2.5 Flash with eight tools. One scrapes the page (URL_search), and the three
   sub-workflow tools list its links and read Google. Two ask Claude to search the web in the chosen language
   (online_Search, deep_research). Think_tool holds its notes, and insert_row_in_data_table saves the 14 fields.
   It may take up to 40 tool rounds.
4. Get Ranked Keywords asks DataForSEO for up to 200 keywords the address ranks for. Filter keywords drops repeats and
   sorts them by search volume, and they land in Page Ranked Keywords.
5. Keyword for Site asks Google Ads data for keyword ideas. Filter Keywords keeps ideas with 1,000 or more monthly
   searches, Search Intent labels each one, and Filter keeps those over 200 searches with an intent probability above
   0.7. They land in Suggested Page Keywords.
6. Get Business Data reads the run's row back by its session ID. Make Copy of Template copies your document, and
   Change Variables in Doc replaces its 16 placeholders.
7. Generate Doc exports the filled document as an OpenDocument file, and Add Doc to Drive uploads it to My Drive.
   Generate PDF and Add PDF To Drive do the same as a PDF.
8. Send email sends the file name, both links and the session ID to the address in Send to.

Change the look

• The report's layout: your document template. Move, style or drop placeholders there.
• What the agent writes: the field table in the Business Analyst prompt. It sets five brand traits, five questions,
  four journey stages and a persona of four to six sentences.
• Which keywords count: 1000 in Filter Keywords, 200 and 0.7 in Filter, and "limit": 200 in the body of
  Get Ranked Keywords.
• The email: the subject and text of Send email.
• Countries and languages: the form's two lists. A new one also needs its code in Get Country & Language Code.
• Models: Business Analyst uses models/gemini-2.5-flash, and both Claude tools use claude-sonnet-4-5-20250929.

Good to know

• Models: Anthropic's deprecations page, read on 2 November 2026, retires claude-sonnet-4-5-20250929 on 30 November 2026
  and names claude-sonnet-5-5 as its replacement. Google's page says the 2.5 models stay on but only for accounts that
  used them before, and points new projects to Gemini 3.8 Flash. If your key is refused, pick gemini-3.8-flash in
  Business Analyst.
• Billing: Google bills the agent's tokens over all its rounds. Anthropic bills tokens plus each web search, up to 5 in an
  online_Search call and 10 in a deep_research call. Firecrawl counts each scrape, and DataForSEO bills every call,
  with Search Intent making one call for each keyword idea. Try one small site first.
• A run can stop without a report. When the page ranks for nothing, or no keyword idea passes the filters, the next node
  gets no items and nothing after it runs, so no email comes. To let it carry on with empty lists, turn on Always Output
  Data in the settings of Filter keywords, Filter Keywords and Filter. The same happens when the agent never saves
  its row, and the run's execution shows which tool it skipped.
• The search floor: a comment in Filter Keywords says 100, but its code keeps 1,000 and up. So the volume test in
  Filter never removes anything, and only the intent probability decides.
• primary_competitors: the save tool has a column for it, but the prompt does not ask for it and the report never
  shows it.
• client_website is saved as typed in the form, though the prompt asks the agent for a bare domain.
• Drive fills up: each run leaves a copy of the sheet, a filled Google Doc, and the uploaded document and PDF, the last two
  in the root of My Drive. The emailed Document link opens the OpenDocument upload.
• Open form: once the workflow is published, anyone with the form's address can start a run on your keys. Turn on the
  form trigger's basic authentication or keep the address to yourself.
• Your data: the address and what the scrape finds go to Firecrawl, Google, Anthropic and DataForSEO.
• Your key: the Firecrawl key sits in a header, so take it out before you share the workflow.