Citable — give AI the facts, not your web design

ChatGPT, Claude and Perplexity read your site before they answer. A normal page is 400 KB of design, scripts and markup, and under 3% of it is content they can use. Citable hands them the same page as 4.6 KB of plain facts, on your own domain — while people and Google still get your real page.

This response

You are ClaudeBot (Anthropic). Citable recognised you as an AI agent, so this is https://getcitable.in/ with the UI removed — the content only. A browser asking for the same address gets the full designed page.

The core idea: your website without the UI, for AI agents

Citable strips the UI from a website for AI agents and serves them only the content.

When an AI agent — ChatGPT, Claude, Perplexity or any of the 81 that Citable recognises — opens a page, Citable removes the theme, scripts, navigation and layout and serves only the content: the same facts, at the same URL, on the site's own domain. People and Google still get the full designed page.

Every row below is the same address. Citable decides which version to send by checking who is asking, before the response goes out.

What Citable is, and what it is not

Citable is a routing layer. It sits in front of your site's requests, hands recognised AI crawlers a clean, token-light copy of the same page on your own domain, and records every fetch: which engine, which page, when, and what figure was in the response.

Check it yourself

getcitable.in runs on Citable. Fetch any page twice, once as a browser and once with an AI agent's user agent.

curl https://getcitable.in returns the designed page a person sees.

curl -A "GPTBot" https://getcitable.in returns the content-only version an AI agent is served — this document.

The problem

An AI assistant names three sources and links them. Being the fourth is the same as not existing, and there is no second page to be on.

Assistants do not run your JavaScript. They fetch the HTML, take what is in it, and stop. Anything your site draws afterwards was never there as far as the engine is concerned — a price, a fee table, an eligibility rule, a spec loaded from JSON, a rate card behind a tab, a code sample a docs framework hydrates, an availability badge. The failure is a rendering failure, not a retail one: it happens on documentation, on a law firm's practice page and on a university's fee schedule for exactly the same reason.

The second failure is size. Whatever survives is buried in theme, script and navigation an engine has to pay for before it reaches a single fact.

What it runs on — any website, on any stack

Citable is a routing layer, not a plugin for one ecosystem. It needs exactly one thing: to sit in front of the request before the response goes out. It does not care what generates the page behind it, and it never edits that page.

It is not an e-commerce tool and it is not Shopify-only. Shopify is the one-click route in; it is one adapter of four.

Where the facts are read from

The clean copy is rebuilt from wherever your stack already keeps structured content, never scraped from the page it replaces. That is why it cannot state a figure you do not publish.

What we measured across 15 live homepages

Fetched as GPTBot on 2 September 2026, counted with the cl100k_base tokeniser. The sample was Indian direct-to-consumer sites because that is the sample we could take; the numbers describe how heavily modern front-ends render, which is not particular to what a site sells.

What Citable does

It sits in the request path on your own domain. A browser or a search engine gets your site exactly as it is today. A recognised AI assistant gets the same facts as clean, structured HTML it can finish reading.

Because it serves the response, it can also record what happened: which engine read which page, when, and whether a person was waiting on the answer.

The 81 AI agents that get the content-only version

Matched on user agent, and verified against published IP ranges where the operator publishes them. Search engines are never on this list, so rankings are never part of it.

What kind of page it works on

What Citable does not claim

Who it is for

The examples below are categories we have looked at closely; they are not a limit. The test is narrower and has nothing to do with what you sell: does somebody ask an assistant about your category BEFORE they have a shortlist? If they do, being unreadable costs you the shortlist rather than the comparison — and that is as true of a compliance consultancy or a hosting provider as of a shop.

Source: https://getcitable.in/ · Citable · Run the free audit · llms.txt · getcitable@gmail.com