Scrape Website Content and Extract SEO Keywords with GPT-5-mini and Airtable

How It Works

The workflow collects a website URL through a form trigger, fetches the page with an HTTP Request, and strips the HTML down to plain text in a code node. The cleaned text runs through two GPT-5-mini AI Agents in sequence: the first extracts topic-wise information from the page, and the second generates a 90-keyword SEO list. A second code node removes residual markdown artefacts like ### and ** between the two agents. The structured output, cleaned data, keyword list, and workflow status, lands in Airtable through an upsert so repeat runs on the same URL update rather than duplicate.

Setup Steps

1. Prepare Airtable: Create a base with a table that has Website Name (string), Data (string), Keyword (string), and Status (single-select with Todo, In progress, Done).
2. Connect Credentials: Add OpenAI credentials with GPT-5-mini access and Airtable credentials in n8n.
3. Point the Airtable Node: In the Airtable Upsert node, select the base, table, and the field that acts as the upsert key (usually Website Name).
4. Tune the Prompts: Open each AI Agent node and adjust the system prompt if you need keyword counts other than 90 or want topic extraction scoped to a specific industry.
5. Activate the Form: Publish the form trigger so researchers can paste URLs and kick off the pipeline without opening n8n.

Automation Workflow

Form Trigger (URL Input) → HTTP Request (Fetch Page)
↓
Code (HTML to Plain Text) → Split Out → AI Agent (Topic Extractor, GPT-5-mini)
↓
Code (Text Cleanup) → Split Out → AI Agent (Keyword Extractor, GPT-5-mini)
↓
Wait → Merge → Airtable (Upsert Website, Data, Keywords, Status)

Benefits

• Form-Driven Research: Non-technical users submit URLs through a form instead of running the workflow by hand.
• Clean Inputs: The HTML cleaner strips styles, tags, and whitespace before the LLM sees the page, which cuts token spend and improves quality.
• Two-Stage AI: Topic extraction and keyword extraction are separate agents, so each stays focused and you can tune one without breaking the other.
• No Duplicates in Airtable: The upsert pattern keeps the keyword table clean when the same URL is researched more than once.

Advanced Features

• Dual GPT-5-mini AI Agents for topic and keyword stages.
• Two code nodes for HTML cleaning and markdown artefact stripping.
• Split Out and Merge nodes to parallelize work and recombine for the Airtable write.
• Wait node to pace the merge so Airtable receives one complete record per URL.
• Form Trigger that works as a public endpoint for the research team.

Use Cases

• SEO specialists building topic clusters and keyword briefs for new content.
• Content strategists auditing competitor pages for keyword coverage gaps.
• Growth teams enriching a target-account list with site-level SEO fingerprints.
• Agencies producing white-label keyword reports from client site URLs.