DEV Community

Cover image for Business and website Analysis AI Agent with N8N and Bright Data
Nestor Campos
Nestor Campos

Posted on

Business and website Analysis AI Agent with N8N and Bright Data

#devchallenge #ai #webdev #n8nbrightdatachallenge

n8n and Bright Challenge: Unstoppable Workflow

This is a submission for the AI Agents Challenge powered by n8n and Bright Data

What I Built

Our AI agent, developed with N8N and Bright Data, is designed to intelligently and automatically analyze business websites. Its purpose is to provide a complete snapshot of the business and its digital presence, identifying both strengths and opportunities for improvement.

🔍 What problem does it solve? Companies often struggle to understand how their value proposition is perceived online, whether their website effectively communicates their services, and whether the user experience is optimized. Our agent addresses this challenge by:

  • Evaluating the business's content and positioning, detecting whether it conveys clarity, differentiation, and added value.
  • Analyzing the website's structure, design, and accessibility, including navigation, formatting, loading times, and mobile compatibility.
  • Generating actionable feedback to improve both the digital strategy and the user experience.

🚀 The result: a comprehensive view that allows companies to make informed decisions to strengthen their online presence and better connect with their customers.

Demo

Here is the video with an explanation and demo: https://www.youtube.com/watch?v=A9qcpINAonQ

n8n Workflow

Image demo

Github gist of the n8n agent: https://gist.github.com/nescampos/9a3c4e80c11273c5c3a98c4fc2a5010b

Technical Implementation

My AI agent is made up of:

  • N8N Form: Used as a trigger, where the user enters the website they want to analyze.
  • Bright Data Verified Node: I use this part to extract all the content from the specified website, in HTML format.
  • Extract HTML: I extract only the necessary HTML (body).
  • AI Agent Node: Connected to OpenAI (using a compatible service and the gpt-oss-120b model), I instruct the agent via a specific prompt to perform a thorough analysis.
  • N8N Form: To display the results of the agent's analysis.

AI Agent details:

Prompt

Act as a web development and digital communication expert. Analyze the following HTML code along with its textual and visual content (if applicable). Your objective is to evaluate both the technical structure and the communication effectiveness of the site. Submit a report divided into the following sections:

  1. ✅ Positive Aspects Good HTML practices (structure, semantics, accessibility, performance)

Clarity of the message and the product or service offered

Elements that generate trust or engagement

  1. ⚠️ Negative Aspects Technical errors or bad practices in the HTML

Ambiguity in the content or lack of clarity about the product/service

Usability, accessibility, or user experience issues

  1. 🛠️ Recommendations for Improvement Specific suggestions for optimizing the HTML code

Improvements to writing, information hierarchy, and communication of the product/service's value

Proposals to improve conversion, SEO, and the visitor experience

Be clear, precise, and structured. Use technical language accessible to intermediate developers and writers. If you find the site isn't communicating its value proposition well, explain why and how it could be better.

HTML:
{{ $json.body }}

Send the results in HTML format to display in a website

Tools

No tools in this workflow

Memory

No memory

Model

For this demo, I used gpt-oss-120b, but any other model can be used.

Bright Data Verified Node

I decided to use the Web Unlocker resource to extract all the information provided by the user from the website, allowing me to access the entire site's HTML for subsequent analysis by the agent. In this case, the trigger is a form with the URL to be scraped and subsequently analyzed.

Bright Data Verified Node

Journey

  1. First, I started by wanting to solve a problem, and I realized I had one (understanding my product websites to see how to improve or if they were understandable to end users).
  2. Next, I analyzed the use of N8N and each component.
  3. Agent AI was key, but it wasn't the initial step, as I had to keep it very simple, so I decided to use a form as a trigger to reduce errors and focus on what was important.
  4. I made sure the prompt was clear and objective.
  5. I ran extensive tests to ensure the flow was well-designed so anyone could use it.

Team

Top comments (0)