You can now check your AI visibility from inside ChatGPT and Claude

The short version: you can now audit any site's AI visibility without leaving ChatGPT, Claude or Cursor. Connect the free Alice MCP once, ask your assistant to check a URL, and it reads back the score, the blocked crawlers, and the fixes. The same scan that runs on our website, one message away from wherever you already work.

Why this is a real gap, not a nice-to-have

A site can rank perfectly on Google and be invisible to the answer engines. Googlebot and the AI crawlers obey different lines in the same robots.txt, and none of the AI crawlers execute JavaScript. So a page that looks healthy in Search Console can arrive empty at GPTBot, or be blocked outright by a robots.txt rule written before those crawlers existed. Nothing in your dashboards shows the symptom, because from every tool you already own, nothing happened.

The problem is easy to check and easy to miss. What was missing was a way to check it in the place people now do this kind of thinking: inside the assistant, mid-conversation, without opening another tab.

Ask your assistant, get a real scan

The Model Context Protocol is the open standard ChatGPT, Claude and Cursor use to reach outside tools. The Alice MCP exposes two read-only audit tools over it. Once connected, you ask in plain language and the assistant runs the actual scan: it resolves your robots.txt the way each crawler does, reads the raw HTML with no browser, and checks your structured data, then hands the model a scored result to reason about.

The one that travels is the comparison. Give it your site and a couple of rivals:

Seeing your own number sitting below two competitors is the moment the abstract problem becomes concrete. And because it renders inside the chat, the assistant narrates it for you without being asked.

Connecting it takes about two minutes

There is one server URL for every client: https://tranx.io/api/mcp. In Claude or ChatGPT you add it as a custom connector with no authentication; in Cursor it is a few lines in .cursor/mcp.json; in Claude Code it is one command:

claude mcp add --transport http alice https://tranx.io/api/mcp

The step-by-step for each client, with copy buttons, is on the Alice MCP page. It is free and needs no account.

What you can ask once it is connected

Two tools, both read-only. Your assistant picks the right one from what you say:

• “Check the AI visibility of my site.” A 0 to 100 score, which AI crawlers are allowed or blocked, whether your content survives without JavaScript, schema coverage, and the highest-impact fixes, with a shareable report link.
• “Compare my site against these two competitors.” The same, across several sites, ranked by score.

Both tools fetch and grade public pages and never change anything. The MCP has no write access to your site, your repository, or your assistant.

A scan is capability. Citation is the product.

The audit tells you whether the answer engines can read a site right now. That is the capability question, and it is the right place to start, because an engine that cannot read you will never cite you. What a one-off scan cannot do from inside a chat window is watch that answer change over time: notice the morning a deploy blocks a crawler, remember what it told you last week, or track whether you actually get cited.

That part is what Alice does once you connect Search Console, GA4 and your ad accounts. It checks a set of buyer questions against the AI engines on a schedule and ties discovery through to signups, orders and revenue. The audit you just ran, in your assistant or on our site, is the free front door to it.