How to optimize your website for generative chatbots like ChatGPT, Claude, and Perplexity

HTML structure 

Make sure your site’s code is organized into clear, semantic sections such as:  <header>, <nav>, <main>, <section>, <aside>, <footer>. 

This helps both search engines and language models grasp the hierarchy and logical flow of your content. 

Use Schema.org markup 

Schema.org isn’t new. Back in the mid-2010s, Google and other search engines used its formats to structure search results. While Google has scaled back its use of certain schema types since 2022, many generative models — including ChatGPT and Claude — explicitly mention that schema markup helps them understand web content. 

Here are some useful schema types: 

WebPage General page information

Article , BlogPosting , NewsArticle Articles and blog posts

FAQPage , GAPage Frequently asked questions

Product Product details

Organization , LocalBusiness Company or organization info

Event Event listings, time, and location

Person Biographies and contact information

BreadcrumbList Navigation and site hierarchy

Review , AggregateRating Reviews and ratings

HowTo , Recipe Step-by-step guides and recipes

You add schema markup as a JSON snippet inside your HTML to describe the content. 

Server-rendered content 

ChatGPT can read the raw HTML of a site but cannot interpret JavaScript. 

If your website is built with frameworks like React or Vue, where content is rendered in the browser, ChatGPT won’t be able to access it. The solution is to render pages server-side — either with traditional HTML or frameworks like Next.js or Nuxt.js that perform server-side rendering. 

As an editor, you can also take several steps to improve your content. Many of these resemble traditional SEO principles, but with an extra focus on clarity and structure. 

Headings and hierarchy 

Use proper heading levels (H1, H2, H3, etc.) to make your pages semantically structured. This helps users, search engines, and AI models understand your content. 

Write clearly and consistently 

Use clear, consistent language. Avoid technical jargon unless necessary. 

It often helps to frame headings as questions and then answer them — mirroring how people interact with chatbots. 

Create an llm.txt file 

llm.txt is a new format designed to guide large language models in understanding your site’s content. It’s written in Markdown, using headings and sections. While there’s no fixed structure, best practice is to describe: 

• Who your company is and what you do 

• Your key pages (e.g., About, History, Values) 

• Your main services or products 

Example: 

# About Us 

We are a digital agency helping businesses with web development and SEO.    # Key Pages 

- /about 

- /contact 

- /services    # Services 

- Web development in .NET and Next.js 

- SEO and AI optimization 

Action-oriented AI Agents 

So far, generative chatbots have mainly provided information — users have had to act on it themselves. 

But over the next 1–2 years, AI agents will be able to take action directly in interfaces like ChatGPT. They’ll be able to register you for an event, schedule a meeting, or even help with your tax filing. 

OpenAI’s upcoming Model Customization & Personalization (MCP) and Agents features will enable developers to: 

• Build GPTs that maintain persistent state 

• Let agents act on behalf of users (via browser control or API integration) 

• Create advanced AI apps with custom logic and data connections 

There’s no public launch date yet, but full agent functionality is expected between 2025 and 2026. 

When it arrives, companies will be able to define what actions ChatGPT is allowed to perform within their systems — reclaiming part of the user experience. 

The way users interact with websites is about to change. 

By structuring your code, using schema.org markup, writing clearly, and serving readable HTML, you ensure that both search engines and AI agents can discover, interpret, and present your content correctly. 

Website optimization is no longer just about Google — it’s about being understood by the conversational assistants of the future.