AI Workflow Automation

n8n Website Data Extraction to JSON

An n8n and OpenAI workflow discovers website datasets, processes Excel, CSV and HTML tables, and prepares structured JSON with schemas and tags.

Website datasets branch into Excel, CSV and HTML processors, then become structured JSON grouped by starting URL.

Illustrative overview of website dataset discovery, format-specific processing and the JSON handoff.

Project data processor canvas, showing the separate Excel, CSV and HTML table paths.Project Excel-to-JSON subworkflow for turning spreadsheet content into structured data.Project table-to-JSON transformer, including schema generation and tag extraction.Project error-handler subworkflow for recording processing failures in a structured log.Project fatal-error notification workflow, kept separate from routine error logging.Project main orchestration canvas connecting source discovery, subworkflows and JSON output. These editor captures show the workflow structure, rather than a measured completeness or accuracy result.

The problem

Website datasets were spread across downloadable spreadsheets, CSV files and HTML tables. The task was to start from supplied URLs, discover relevant data sources and organize their contents for downstream analysis instead of handling each format separately.

What I built

n8n orchestration with an OpenAI discovery agent to find candidate dataset links from starting URLs.
Separate processing paths for Excel, CSV and HTML tables, with reusable table-to-JSON subworkflows.
Schema generation and classification, including geographic and subcategory tags.
JSON output organized by starting URL, containing dataset entries and a processing log.
Configurable crawl settings and URL-pattern filtering for the requested language scope. The language filter checks URL patterns rather than translating or verifying every page's language.
Structured error logging and a separate notification path for fatal failures.

The outcome

The workflow connects discovery and format-specific processing to a common JSON handoff. Schemas, tags and logs provide context for downstream tools and for investigating failures. New website structures and unusual file formats can require processor changes.

Who this approach suits

Teams collecting recurring website datasets for analysis or internal tools. To scope similar work, bring starting URLs, example source files, required fields and the intended JSON format.

Discuss website data extraction.

Working on something similar?

Discuss a similar project