Landing page
AI-based automated landing page generation for a B2B SaaS
#AI #automation #ContentOps #B2B #SaaS
Situation
Latenode is a B2B SaaS platform that enables businesses to automate workflows by connecting various applications, similar to tools such as n8n and Zapier.

The platform supports over 5,000 integrations, each with its own SEO landing page designed to attract organic traffic and convert visitors into users.

Since dozens of new integrations were added every week, creating and maintaining these pages manually wasn't feasible. Instead, every landing page was generated automatically from a single AI prompt and a shared generation workflow inside Latenode.

The existing prompt had been written quite a long time ago. While the generated pages ranked well in search engines, they converted poorly.
Task
To increase registrations from organic traffic while preserving existing search rankings.
Action
Step 1: Analysis

I started by analyzing the existing landing pages and reviewing user behavior.

Web analytics showed that visitors rarely engaged with the content beyond the first few sections of the page. The reasons were quite apparent:
  • the hero section contained little more than the integration name and a small CTA;
  • value propositions were generic, AI-generated phrases like "Unlock a new realm of possibilities", "Leverage the incredible capabilities", or "Transform the way you [do something]";
  • compatible applications were presented as a wall of logos with little context.
This was followed by a long wall of poorly generated text that users rarely scrolled to, leaving most of the page unread.
The good thing was that users consistently interacted with the application logos and trigger/action examples.

Based on these findings, I suggested a different approach: instead of describing the product, we should demonstrate what users could accomplish with it. The landing pages should communicate the product's value more clearly and reduce friction before users could experience it themselves.

Step 2: Reorganizing the information hierarchy

I redesigned the landing page structure around value and interaction.
Instead of showing only the integration name, the hero section now immediately communicated the key benefits of the integration through concise value proposition labels. One of these highlighted Latenode’s cost efficiency, which repeatedly emerged during user interviews as a key differentiator. Another promoted the platform’s AI Agent, which at the time was still a relatively unique feature.

The hero copy was rewritten to focus on business outcomes rather than product descriptions, explaining how a particular integration could help users solve real problems.

I also introduced a secondary CTA alongside the primary "Try it now" button, allowing users to explore how the integration worked before committing.
Throughout the page, I shifted the focus from generic marketing copy to interactive product demonstrations.

The most important interactive elements were moved higher on the page, including:

  • compatible applications;
  • trigger and action examples;
  • integration templates.

Where available, I surfaced ready-made automation templates that users could import directly into their account with a single click, significantly reducing the effort required to experience the product.

Instead of long descriptions, users could immediately explore practical automation scenarios.

To create more touchpoints, I added navigation blocks leading to similar integrations and broader application categories.

Even if visitors were not ready to connect the featured application, these recommendations encouraged them to continue exploring related solutions instead of leaving the site.
I also added a customized setup guide that walked users through connecting the two applications using actual Latenode interface screenshots. Although the tutorial contained around ten steps, it allowed visitors to understand exactly how the integration worked before signing up.
Step 3: Creating prompts and scaling the solution

After designing the new structure in Figma, I wrote model content illustrating how each section should communicate value.

Together with another prompt engineer, I redesigned the AI prompts responsible for generating the landing pages. We iteratively tested and refined the prompts until they produced consistent, high-quality output across different integrations.
Then, working together with a Frontend Engineer, we first generated several landing pages for newly released integrations as an experiment before rolling out the updated template across the existing library.
Result
Although the experiment was conducted on a relatively small sample, the results were encouraging.

The redesigned pages:

  • maintained their SEO rankings despite containing significantly less text;
  • increased registrations from organic traffic;
  • improved user engagement, including time on page and interactions with page elements.

After I left the project, the approach continued to evolve.

Today’s Latenode landing pages still follow the same principles introduced during this redesign:

  • product interactions instead of long-form explanations;
  • value propositions are communicated immediately;
  • friction is minimized before users experience the product.
Since then, the platform’s AI capabilities have become significantly more advanced, allowing visitors to build automation flows directly from the landing page. While the implementation has evolved, the underlying UX principles remained the same, which validated the direction of the redesign.
Made on
Tilda