Knocklog

How to make a Lovable site visible to ChatGPT and other AI agents

A site that looks great in a browser can reach an AI agent almost empty. Here is how to check your Lovable site in a minute, have Lovable itself make the fixes, and measure what AI assistants send to it.

Checked on 13 November 2026

Why it happens

Lovable projects created before May 2026 use React with Vite: the page reaches the visitor nearly empty and the content is built in the browser by JavaScript. Many AI agents do not run JavaScript; they read the HTML they received and move on. For these projects Lovable serves a pre-rendered copy of the pages to the search and AI crawlers it recognises. New projects use TanStack Start, which builds pages on the server, and Lovable offers the upgrade to older projects that can be migrated.

In practice: a new or upgraded project delivers its content to any agent. An older project depends on the agent being recognised. The way to know is to check the published site.

Check your site

Knocklog's check reads the site as an unknown visitor and shows a 0–100 score, what robots.txt allows or blocks for each AI agent, whether the content appears without JavaScript, whether there is an llms.txt and whether the site offers markdown to agents. If the result says the content depends on JavaScript, upgrading the project to TanStack Start is the fix that works for every agent.

Connect Knocklog to Lovable's chat

Lovable accepts MCP servers in chat on every plan. With Knocklog connected, Lovable itself can check the site and fix what lives in the project.

Using the Knocklog connector, check how my published site
(example.com) treats AI agents.
Then fix what you can in this project:
- robots.txt (allow AI search and assistants;
  AI training as I prefer);
- an llms.txt describing the site;
- each page's title and description.
Finally, tell me what cannot be fixed inside the project.

The connector is read-only: it checks public websites and looks up Knocklog's directory of AI agents. It needs no key and does not read your project's data.

Measure who arrives through AI

Fixing shows what agents may read. To see what actually happens, create a free Knocklog project with the site's address. There are two paths, depending on the domain:

Add this line to the <head> of every page of the site,
and change nothing else:
<script src="https://api.knocklog.ai/leads.js" data-project="YOUR_PROJECT" async></script>

Replace YOUR_PROJECT with the code shown in the Your plan tab of the Knocklog report.

See how your Lovable site treats AI agents

The check is free and needs no sign-up.

Sources

Read next