Cartexel.AI — Page Auditor

About

Cartexel.AI — Page Auditor
See any web page the way search engines, LLMs, and AI agents actually see it — then fix what's holding it back.

Your pages weren't just written for people anymore. They're read by Google's crawler, by ChatGPT, Claude, Gemini and Perplexity when they answer questions, and by autonomous AI agents that navigate and act on sites. Each of these "readers" sees something different from what a human sees in a browser — and most pages quietly fail one or more of them. Cartexel.AI — Page Auditor shows you exactly what each reader gets, scores your page on three axes out of 100, and hands you a prioritized list of concrete fixes.

Click the toolbar icon on any page and get an instant, private audit. No sign-up, no account, no data leaves your machine (the core audit runs entirely in your browser).

WHAT IT SCORES

The auditor models the three consumers of every modern web page and grades each out of 100, plus a weighted overall score:

• LLM & Agent Readability — Can a machine fetch, parse, and navigate your page? It checks raw-HTML content vs. JavaScript-gated content, semantic landmarks and ARIA, heading order, boilerplate ratio, machine-readable metadata, validated structured data, AI-crawler access (including llms.txt), and the accessibility-tree affordances that AI agents rely on to act on a page.

• Content Quality for LLMs — Is your content actually worth citing in an AI answer? It looks at depth, extractable chunking, answer-first structure, factual density (numbers, stats, units), authoritative citations, E-E-A-T signals (author, organization, dates), freshness, linguistic readability (sentence length + Flesch), topic coherence, and whether the content is unique or thin.

• SEO — Classic and modern search readiness: indexability, title and meta tags, heading structure, canonical URLs, structured data for rich results, social cards, mobile-friendliness, HTTPS, Core Web Vitals proxies, accessibility, internal linking, URL hygiene, hreflang, and sitemap.

Because a product page and a blog article should never be judged the same way, a built-in page-type detector (Product / Collection / Article / Docs / Landing) automatically swaps in tuned scoring profiles for the page you're on.


WHY THE SCORES ARE ACCURATE — NOT JUST GUESSED

• Structured data is parsed and validated, not merely detected. A Product page is checked for offers, price, and image; an Article for headline, author, and datePublished — and scored on how complete it really is.

• Accessibility and agent affordances are read from the live accessibility tree — the same tree AI agents use to understand and act on a page — including form-label association, accessible link and button names, landmarks, and heading order.

• Core Web Vitals proxies flag real performance risks: render-blocking scripts, images without dimensions (layout shift), and heavy DOM/HTML weight.

• JavaScript-gating is measured, not assumed. The tool compares the words in your raw HTML to the words in the rendered page and penalizes the gap — so single-page-app shells that look fine to humans but nearly empty to crawlers are correctly scored low.


THE FLAGSHIP: SEE HOW AI ACTUALLY READS YOUR PAGE

This is what sets the auditor apart. Instead of guessing, it shows you what an AI genuinely extracts from your page:

• Extractable facts — the product, price, brand, or the headline, author, and date — each tagged with its source: confirmed from structured data (reliable), guessed from text (may be wrong), or missing entirely (the AI has to infer or skip it). The lesson lands immediately: put it in schema, or the AI guesses.

• "Show the actual text an AI reads" — the cleaned, chrome-stripped text a crawler ingests, with a clear warning when much of it is only available after JavaScript runs.

• Citability test (powered by Claude) — this simulates an AI answer engine end to end. It writes the likely user question, answers it using only your page, lists which facts it's confident about versus unsure, flags hallucination risks and gaps, gives your page a citability score, and tells you exactly what would make it more citable. It's the truest available measure of whether your page is "LLM-ready."


MEASURED CORE WEB VITALS

A dedicated panel measures your page's real LCP, CLS, FCP, and TTFB live from the page using PerformanceObserver and Navigation Timing — color-coded against Google's good / needs-improvement / poor thresholds. The measured result is blended into the SEO performance signal, and poor LCP or CLS raise their own actionable issues. When live measurement isn't available, it falls back to intelligent estimates.


BUILT FOR ECOMMERCE AND MERCHANTS

On product pages, the auditor adds tools built around Cartexel.AI's core domain — getting product data ready for AI-driven commerce:

• Merchant feed readiness — scores the exact fields Google Shopping, Performance Max, and AI shopping agents need: title, description, image, price, currency, availability, brand, GTIN, plus MPN, SKU, condition, ratings, and shipping. Read from JSON-LD Product/Offer data first, with smart text fallbacks.

• Ready-to-paste feed exports — generate a Google Merchant feed row (TSV) and a complete, enriched Product JSON-LD block from your page's real values, with bracketed placeholders only where a value is genuinely missing. One click to copy each.


PRIORITIZED FIXES YOU CAN ACT ON TODAY

Every problem found across all three dimensions is compiled into a single issues-and-fixes list, ranked critical → warning → info, each with a concrete recommendation. A "Copy full report" button exports the entire audit as plain text so you can share it with your team or drop it into a ticket.


OPTIONAL DEEP ANALYSIS WITH CLAUDE

The audit is powerful on its own and completely free. When you want a real editorial judgment of quality — not just structure — one click sends the extracted main content to Claude for a genuine review with strengths, weaknesses, and concrete fixes. It also returns ready-to-paste artifacts: a generated meta description and a complete, valid JSON-LD structured-data block filled from your page's real content. Deep analysis is entirely optional and requires your own Anthropic API key, set once in the options page.


COMPARE AGAINST A COMPETITOR

Paste any competitor URL and the auditor fetches and scores it, then shows a side-by-side comparison — overall score, LLM readability, LLM quality, SEO, feed readiness, word count, and schema validity — with clear win/lose indicators. Both pages are scored from raw HTML for a fair, apples-to-apples comparison. A safety guard blocks private and reserved addresses to prevent misuse.


PRIVATE BY DEFAULT

The core audit runs 100%  in your browser — instant, free, and private. Nothing is sent anywhere unless you explicitly choose to run the optional Claude deep analysis or citability test, which use your own API key.


WHO IT'S FOR

• Ecommerce merchants and Shopify store owners who want their products to show up and be cited by AI shopping assistants.
• SEO specialists and content marketers adapting to answer engines and AI Overviews.
• Developers and technical teams shipping pages that need to be crawlable, structured, and agent-friendly.
• Anyone who wants to know — concretely — whether AI can read, understand, and cite their page, and what to fix first.


HOW IT WORKS

A background service worker fetches your page's raw HTML, robots.txt, and llms.txt — the same bytes AI crawlers mostly see. It reads the rendered DOM to detect JavaScript-gated content and snapshot the accessibility tree. A dependency-free scoring engine then computes all three scores and the prioritized issue list from those inputs.


Install Cartexel.AI — Page Auditor and stop guessing how machines see your pages. Start seeing exactly what search engines, LLMs, and AI agents read — and fix what matters, in priority order.

By Cartexel.AI — product data enrichment for AI-driven commerce. https://www.cartexel.ai