Build an AI Agent in 20 mins that Roasts your Landing Page (No Code Needed)

Learn how to build a no-code AI agent that scrapes any landing page, grades its copy, and emails you a detailed audit—using n8n, Firecrawl, and GPT-4o. Perfect for founders, marketers, and agencies.

Aug 08, 2025

Friends,

This week I’m coming to you with a brand new AI marketing use-case: Your very own AI landing Page Roaster. I’ve worked with a ton of SaaS companies optimizing their landing pages and often times, it starts with just an audit of their existing landing page.

And AI can help with that.

The best workflow would look something like this - you enter any landing page URL as input and you get back an audit report of your landing page with action items on how you can increase your conversions.

And in this issue, I’m going to share with you on how you can do just that.

The best part is that you can set it up as a standalone SaaS. Since we will be using a webhook, you can hook it up to any frontend the user can interact with.

This means that you can set it up as a free tool to collect traffic and users or spin it up as a paid SaaS where users submit their landing page and get back an audit report. Or, you can just use this internally if you have many landing pages and are running ads to it - that way you can AB test and keep experimenting with what works and what doesn’t.

In this guide, you’ll learn how to build a no-code AI “agent” that scrapes any landing page, runs it through an LLM for a detailed copy critique, and emails you a structured audit… all on autopilot.

This setup takes 15–20 minutes to replicate, requires zero coding, and runs entirely inside n8n using APIs from Firecrawl, OpenAI, and Gmail.

Let’s dive in!

More of a video person? Check this out 👇

Want to read along? Here you go: 👇

What You’ll Build

By the end of this tutorial, you’ll have an automation that:

Accepts a URL from you (or your site visitors).
Scrapes every piece of text from that page—headlines, body copy, features, pricing.
Feeds that text to an LLM, acting as a “world-class SaaS landing-page copywriter.”
Returns:
- Scores (0–10) for headline, clarity, messaging, value proposition, persuasiveness, and overall.
- Specific improvement suggestions.
- An estimated “potential score” after fixes.
Emails the full roast straight to your inbox.

Step 1 – Prerequisites

Before building, make sure you have:

n8n (self-hosted or cloud account – free tier works).
Firecrawl.dev account + API key.
OpenAI API key with GPT-4o access.
Gmail account with OAuth credentials set up in n8n.

Step 2 – Building the Workflow

1. Create a New Workflow in n8n

We’ll be using a simple, linear chain of nodes.

Nodes Overview:
Webhook → HTTP Request (Firecrawl) → Edit Fields → OpenAI Chat → Gmail

2. Webhook / Chat Trigger Node

Method: POST
Expected input: JSON with one property: "url".
Extend the timeout to handle slower scrapes.

You can later embed this webhook in a site chat widget, so visitors can request their own roasts. Or hook it up to Lovable and build a custom frontend so users can interact with it. (Lemme know if you wanna know how to do that, and I’ll help you out - reply to this email and I’ll email you back).

3. HTTP Request Node – Firecrawl.dev

Endpoint: https://api.firecrawl.dev/v1/scrape
Method: POST
Body (JSON):

{
  "url": "{{ $json.chatInput }}",
  "formats": ["json"],
  "onlyMainContent": true,
  "removeBase64Images": true,
  "blockAds": true,
  "proxy": "auto",
  "jsonOptions": {
    "prompt": "Extract the landing page's copy like the headline, the body copy, the features section headline, subheadline, value proposition section, pricing page. Scrape every text that's on the page and leave nothing out."
  }
}

Auth: API key from your Firecrawl credentials.
Timeout: 120000 ms (large pages take time).

4. Edit Fields Node

Keep only the main body text.
Clean up whitespace:

5. OpenAI Chat Node

Model: GPT-4o.
System prompt: Position the AI as a world-class SaaS landing-page copywriter.
Instructions:
- Grade six areas: headline, clarity, messaging, value prop, persuasiveness, overall.
- Provide actionable recommendations for each.
- Include both current score and potential score after improvements.
User message: Pass the cleaned Firecrawl text.

Here’s the full prompt:

You are an elite and world class SaaS landing page copywriter. You know how to write persuasive copy that engages users and gets them to convert. You are able to understand and audit any landing page and identify weak points and areas for improvement to make it more converstion friendly. 

Your role - Your job is to analyse the copy of this landing page : {{ $json.data.json.allText }} & create a comprehensive and professional SaaS Landing page audit report. 


1. Check for headline effectiveness - is it clear, engaging and catchy as to what the product does? give a headline score + action items to improve the headlines. 

2. Check for sentence clarity and conciseness - is each sentence clear, concise and adds value to the user? Suggest areas for improvement with actionable steps the user can execute right away. 

3. Check for product messaging clarity - How clear is the hero messaging - does it make people want to click the CTA - is it clear who it is for - is the promise clear as to what the product does? Suggest areas for improvement with actionable steps the user can execute right away. 

4. Do a Value Prop Clarity Audit - how clear is the main value proposition of the product in the copy? is the benefit of the product clear to the user? Suggest areas for improvement with actionable steps the user can execute right away. 

5. How persuasive is the copy? Does it use powerful storytelling? Does the copy flow naturally like a conversation or does it read more like an essay? Suggest areas for improvement with actionable steps the user can execute right away. 

6. Does it use social proof? If so, how can it be enhanced and optimized even more? 

7. What other shortcomings are there in the landing page that if fixed will lead to better outcomes? What is the percentage of conversions that will go up if the user makes these changes? 

---

Create a comprehensive audit report based on this. Give it an overall effectiveness score out of 10 and then based on your suggested improvements, give a potenial Overall effectiveness score that will happen if the user implements those suggestions.

6. Gmail Node

From: Your Gmail account (via OAuth in n8n).
To: Your email (or dynamic recipient).
Subject: “Landing Page Roast.”
Body: Directly insert OpenAI’s response.

Step 3 – Test the Workflow

Send this payload to your webhook:

{ "url": "https://example.com" }

Watch n8n:
- Firecrawl scrapes the page.
- LLM scores each section and suggests improvements.
- Gmail delivers the formatted audit to your inbox.

When tested on cluely.com, the AI suggested “add quantifiable benefits” and “bring differentiators to the top,” bumping the potential headline score from 6/10 to 9/10.

This is the report I got in my inbox after I ran the workflow:

Optional Enhancements

Embed in Chat Widgets – Let site visitors request their own audits or set up a Webhook and connect it to Lovable to create a stand alone SaaS (micro-SaaS) out of it.
Save Results – Store audits in Google Sheets or Airtable for tracking.
Change Delivery Channel – Send results to Slack, Discord, or Notion instead of email.

Landing-page feedback often costs hundreds of dollars and takes days to arrive.
This workflow gives you instant, AI-powered, structured audits—perfect for founders, marketers, and agencies.

Liked this issue? Here’s how you can support me:

If you liked this guide, consider sharing this with your friends.

And you can subscribe to learn how to automate more of your marketing stack without writing a single line of code.

And if you want to hire me to build out a custom AI workflow for you, book a call with me and we’ll jam!

Hire me

Or you can just say hi to me on LinkedIn.

See you soon!

Best,
Shounak.