Create n8n template

Development of Workflow "AI-Editing 2025"

**Goal:** Create an n8n workflow for generating expert reviews of casinos with multi-level fact-checking and cascading data collection logic.

### 1. Data Management and Database

The primary data (country, language, keyword) is filled in by the content manager in a Google spreadsheet.

After that, a template (webhook) is launched in the document.

For data about the author and page matching by language, a database (Supabase) is used. The template already has node data.

### 2. Data Collection (SERP & Scraping) — Fallback Logic

It is necessary to implement fault-tolerant parsing:

1. The logic of DataForSeo needs to be implemented. The task of this node is to collect related keywords ONLY for the main keyword (which is specified in the Google spreadsheet) and place them in the "directory" tab. Also, in this tab, it is necessary to add title, description, autocomplete keywords, and data from Bing (if available).

2. **Node FireCrawl API:** Executes `v2/search` to obtain the Top-10 results in Markdown format. This node is launched for additional pages, such as (mobile app, login, etc.)

3. **Logic If/Error:** If FireCrawl returns an error or `< 3` results:

    - **Node 2 (Serper.dev Search):** Obtaining a list of URLs.

    - **Node 3 (Serper.dev Scraper):** Parsing each URL separately.

4. **Aggregation:** Collecting texts from all competitors into one variable for analyzing the average volume (`average_words`).

### 3. Intelligent Cycle (LangChain Agents)

- **Agent "LSI-Architect":** Accepts the export of keys from Ahrefs (manually exported to a table). Task: to highlight entities and LSI phrases, excluding garbage. The data from the export will be in the LSI tab.

- **Agent "Blueprint":** Creates the structure H1-H4. It must insert technical tags `<MULTIMEDIA>` and `<E-E-A-T>` in places where tables or quotes are needed.

- **Agent "Live Fact-Checker":** Performs a search through Perplexity (Sonar Deep Research) to update data for **December 2025** (needs to be added as a variable to keep the data current).

- **Agent "Writer":** Generates text, considering the variable `Exclude_words` (this data is in the Author tab) and Tone of Voice. Each H2 should start with a direct answer (20 words) for AI output.

  - **Agent "HTML Structure Architect":** An agent based on existing templates (which will be provided in Google Docs) creates a complete HTML structure of the site with styles, JS styles, etc.

    It is important to set up and specify re-checking for content removal, compression, and more. The main task is to ensure that AI does not remove content.

- **Agent "Microdata":** Extracts FAQs from the finished text and forms JSON-LD (`FAQPage` and `Review`, Author, etc.). I will add an example.

### 4. Finalization (Google Doc API)

Exporting the result to Google Doc, divided into Tabs:

There will be 8 tabs:

Main

Bonuses

Mobile app, etc.

It is necessary that in each tab

- **Section 1:** Meta Title/Description (5 options).

- **Section 2:** Main text (HTML markup: `<strong>`, `<h2>`, interlinking links).

- **Section 3:** Technical block (JSON-LD microdata, prompts for Midjourney).