Media
3 items
Description
See any page as GPTBot, ClaudeBot and PerplexityBot see it: no JavaScript, no cookies. One click, in a side panel. See any page exactly the way AI-Chatbots see it — in one click. Your page looks perfect in your browser. That is not the page an AI model reads. Almost every AI crawler downloads your raw HTML and never runs your JavaScript, arrives without cookies, and gets whatever your CDN decides to hand it. If your pricing, your specs or your FAQ arrive after hydration, they do not exist for the model — and you have no way of knowing from a normal browser window. LLM Viewer shows you the difference. Open a page, click the icon, pick a bot. The side panel reports what that crawler received, and the tab itself reloads as that crawler — real User-Agent, cookies stripped, JavaScript switched off — so you are looking at the actual crawler view, not a description of it. WHAT YOU GET • THE AI'S VIEW — your page as Markdown and plain text: the exact token stream the model reads. The "so THAT is what it sees" moment. • INVISIBLE CONTENT — text that is on your screen but missing from what the bot downloaded, listed block by block. Anything important in that list is a bug. • WHAT TO FIX — every problem in a plain-English sentence with a concrete fix. No jargon, no score you have to interpret. • ALL BOTS AT ONCE — one row per crawler: HTTP status, robots.txt verdict, bytes, words. A firewall rule that blocks GPTBot but lets Googlebot through becomes obvious in one glance. • ROBOTS.TXT, DECIDED PROPERLY — parsed per RFC 9309 for that bot's own token, showing you the exact matching rule. Longest matching group, longest matching path, Allow wins ties, * and $ wildcards. • STRUCTURE — heading outline, JSON-LD validity, alt text, link text, word and token count. • A SCORE AND A GRADE — a diagnostic you can put in a ticket, not a ranking factor. Copy the whole report, or export it as Markdown and hand it to an AI with "fix this page". THREE THINGS ALMOST EVERYONE GETS WRONG 1. Most AI crawlers do not execute JavaScript. Google and Bing can render, but it is queued, slow and not guaranteed. If your content needs a fetch() to appear, assume the model never sees it. 2. "Live fetch" agents are not training crawlers. Claude-User, ChatGPT-User and Perplexity-User fire the moment a human asks the assistant about your URL. Blocking those does not protect you from training — it just means the person asking about you gets nothing. 3. Google-Extended and Applebot-Extended do not crawl anything. They are opt-out switches for AI training and grounding. Blocking them does not affect your search ranking; blocking Googlebot is the one that makes you invisible. LLM Viewer checks both tokens separately and weighs them differently. THE TWO DIAGNOSES PEOPLE CONFUSE • robots.txt says allowed, but the bot gets HTTP 403 → your CDN or WAF is blocking it, not your robots.txt. Cloudflare, Akamai and AWS WAF all ship AI-crawler blocking that is on by default on some plans. Allowlist the agent, or verify crawlers by their published IP ranges instead of by name. • robots.txt says allowed, HTTP 200, and almost none of your text in the HTML → the bot is let in and receives an empty shell. Server-render or pre-render it (SSR, SSG or ISR). WHO IT IS FOR • Developers checking that a React, Next.js, Vue or Angular page actually ships its content • SEO and GEO teams working on AI search visibility, AI Overviews and citations • Technical marketers who need to know why a competitor gets cited and they do not • Anyone auditing a site before, or after, an AI crawler arrives Works on your own site and on any competitor's, with no setup on either. PRIVACY — NOTHING LEAVES YOUR MACHINE There is no account, no sign-up and no backend. The only network requests are to the page you are inspecting and its robots.txt — the same requests your browser would make anyway. Nothing is collected, nothing is transmitted, nothing is sold. Everything runs locally in your browser, and the per-tab rules disappear when you close the browser. HOW TO USE IT 1. Open any page. 2. Click the LLM Viewer icon, press Alt+Shift+A, or right-click → See this page as an AI. 3. Pick a bot. The panel reports; the tab reloads in that bot's view with a bar showing which crawler you are and its full User-Agent. 4. Click Exit to go back to your normal page. Right-click a link → See this link as an AI checks a page without opening it. Needs Chrome 116 or newer for the side panel. Works in Chrome, Edge, Brave and Arc. GOOD TO KNOW The request goes out from your own IP with the crawler's name on it, so sites that verify bots by reverse DNS or published IP ranges will treat you as a spoofer. The User-Agent-based blocking that most sites actually deploy is detected correctly. And a single-page app with no server rendering will look blank in AI view — that blankness is exactly what the crawler got. Made by snipki.de — the AI enablement company. LLM Viewer is an independent tool and is not affiliated with, endorsed by or sponsored by OpenAI, Anthropic, Perplexity, Google, Microsoft, Apple, Meta, Amazon, ByteDance, DuckDuckGo or Mistral AI. Product and crawler names are used only to describe which publicly documented user agent is being simulated.