How is it different from Search Console's URL Inspection?

Search Console tells you how Googlebot sees your page. LLM Inspect tells you how AI systems see it: whether GPTBot, ClaudeBot or PerplexityBot can get in, what content reaches them without JavaScript, what text they extract and whether your URL is already cited in the answers you monitor.

Is it exactly what ChatGPT or Claude do?

No. It is a simulation with open tools (httpx, trafilatura, Readability, Chromium and axe-core) that reproduces what a typical crawler does. It is not OpenAI's, Anthropic's or Google's internal pipeline, and blocks by the real bot's verified IP ranges cannot be reproduced. That is why we keep it separate from the real citation data.

How much does it cost?

1 credit per inspection, summary included. Repeating the same URL within 10 minutes is free. Credits come with your LLMFY plan: Starter (50 credits for €49/mo), Pro (150 credits for €99/mo) or Business (400 credits for €199/mo).

Where do the real citations come from?

From your account's LLM Tracker, covering the last 90 days: Citation Tracking (ChatGPT, Gemini, Perplexity and Claude with web search) and the Prompt Tracker. Each figure carries its source and they are never added together. If you have no prompts set up yet, we tell you.

What happens if the inspection fails?

If it cannot be completed (the site does not respond, takes longer than 60 seconds or our service fails), your credit is refunded automatically.

Can I inspect any URL?

Any public page over http or https. For security, private or internal addresses are not accepted, and there is a limit of 10 inspections per minute.

New · URL inspection for AI

See your page the way ChatGPT, Claude and Perplexity read it

LLM Inspect is Search Console's URL Inspection, but for AI. It tells you whether your page is already cited in the answers you monitor and what an AI system receives when it reads it: whether the bots can get in, what they see without JavaScript, what text they extract and how they chunk it.

1 credit per inspection9 AI readersResults in about 15 secondsPDF export included

Inspect my URL See pricing

LLM Inspect

yourdomain.com/garage-flooring-guide

Warning

72/100

GPTBotClaudeBotPerplexityBotAgents

Without JavaScript64%

Without JavaScript only 64% of the content is visible

main#contenido is rendered with JavaScript: serve that text in the server HTML.

Cited 7 times · ChatGPT and Perplexity · last 90 days

Being indexed by Google does not mean AI can read you.

OpenAI, Anthropic and Perplexity crawlers do not run JavaScript, run into anti-bot challenges and keep only an extract of the text. If that breaks, your page does not make it into the answer, however good it is.

✗

What usually goes wrong unnoticed

•The content is rendered with JavaScript and GPTBot, ClaudeBot and PerplexityBot receive an almost empty page.
•A WAF or Cloudflare serves a challenge to AI bots and nobody notices.
•The extractor picks up the menu, the cookie banner or an "[object Object]" instead of the content.

What LLM Inspect does

Requests the page with each bot's user agent and renders it with Chromium to compare what arrives with and without JavaScript.
Shows you the exact text the model receives, with boilerplate and artifacts highlighted, and how it gets chunked.
Prioritizes fixes with the element's CSS selector and cross-checks the URL with your real citations in ChatGPT, Gemini, Perplexity and Claude.

How it works

A simulation with open tools, explainable from start to finish.

Your URL

Any public page. Private or internal addresses are rejected before anything is requested.

Access and server HTML

robots.txt and a request with each bot's user agent: HTTP status, redirects, challenges and meta robots and X-Robots-Tag directives.

Render and extraction

Chromium renders the page; trafilatura and Readability extract the text with and without JavaScript; axe-core reviews what an agent would use.

Traffic lights and summary

Status per block, overall status with gates, an indicative 0–100 score and a summary with the top three priorities.

Six blocks, each with its own traffic light

Good, warning or critical, with actionable findings and the element's selector when there is one.

Access by bot

robots.txt for GPTBot, OAI-SearchBot, ChatGPT-User, ClaudeBot, Claude-User, Claude-SearchBot, PerplexityBot, Perplexity-User and Google-Extended
HTTP status, redirects and challenge pages per user agent
meta robots and X-Robots-Tag (noindex, nosnippet…)

Without JavaScript

Coverage: main-content words in the server HTML versus the rendered page
≥ 90% good, 60–90% warning, < 60% critical
Headings, links and tables that only exist with JavaScript

What the model receives

The Markdown from trafilatura and from Readability
Whether the H1, headings and tables survive
Leaked boilerplate and artifacts such as [object Object], undefined or "Loading…": any artifact is critical

Chunks

Split by H2 and H3 with a cap of about 800 tokens, like a RAG pipeline
Warnings for chunks over 1,000 tokens, under 40 or without a heading
Chunks that start by referring to earlier text ("as we saw", "this"…)

HTML structure

Number of H1s and whether it sits inside main or article; heading jumps
main, duplicate IDs, tables without th, alt text, lang and time[datetime]
Invalid nesting and navigation built with buttons instead of links

Agent readiness

Nine axe-core rules: button-name, link-name, label, landmark-unique, image-alt…
div and span elements with click listeners that are not real controls
Accessibility tree snapshot: regions and controls

No magic score: gates that cannot be skipped

A page bots cannot read is not fixed by a good H1. That is why the overall status follows two rules before any average.

Access gate

Critical

If a search bot cannot get in (robots.txt, HTTP error, challenge or noindex), the overall status is critical.

No-JavaScript gate

Critical

If less than 60% of the content arrives without JavaScript, the overall status is critical even if everything else is fine.

The indicative 0–100 score weights access (25), no JavaScript (25), extraction (25), chunks and structure (15) and agents (10). It always comes after the gates: with one triggered it cannot go above 40.

It is a simulation with open tools (httpx, trafilatura, Readability, Chromium and axe-core), not OpenAI's, Anthropic's or Google's internal pipeline. Each bot's user agent is reproduced; blocks by the real bot's verified IP ranges cannot be simulated.

Everything you get in each inspection

A diagnosis and, above all, what to fix first.

URL in AI (real data)

Citations of the URL in the answers your LLM Tracker already monitors: engines, prompts, last date and the source of each figure.

Overall status and reader chips

An overall traffic light plus one for GPTBot, ClaudeBot, PerplexityBot and agents.

AI view

The Markdown the model receives, with boilerplate highlighted. Switch between trafilatura and Readability.

Coverage without JavaScript

How much of the content arrives in the server HTML and what only appears with JavaScript.

Chunks

Each chunk with its title, its tokens and its first lines, and the ones worth rewriting.

Fixes with selectors

Each finding with the element's CSS selector and the suggested fix.

Summary with priorities

A plain-language summary with the three highest-impact actions.

History and comparison

Each inspection is saved per URL and compared with the previous one: what got fixed and what is new.

PDF export and server HTML

A PDF report to share and the HTML the server returned, downloadable for 30 days.

Frequently asked questions

Is your page ready to be cited by AI?

Get started with the Starter plan (50 credits for €49/mo). Each LLM Inspect inspection costs 1 credit.

Inspect my URL See all plans

Want to know how you are cited against your competitors too? Ampliaciones GEO checks your page against the sources ChatGPT, Gemini, Perplexity and Claude cite. Discover Ampliaciones GEO