Scannerby warOnSaaS

Freefree account to scanopen source

Can AI assistants read your website?

Put in your website address. In about thirty seconds you see what ChatGPT, Claude and Google can read on your home page, a score out of 100, and the fix for every gap.

Try waronsaas.com or example.com. We fetch your home page and six public files, the way any crawler does. Reports are open to everyone; starting a scan takes a free warOnSaaS account.

scan waronsaas.com Example 0.0s

    What the scan checks

    Twenty-nine checks in four areas, one score out of 100.

    No AI model is involved, so the same website gets the same score every time. Each area counts for a fixed share of the score, and each check inside it carries a weight. The report shows what every fix is worth in points.

    What AI assistants can read

    Whether ChatGPT, Claude, Perplexity and the like can read the site and quote it correctly.

    • A file assistants can readA plain summary of the site at /llms.txt. Cheap to add. Evidence that AI tools read it is still mixed.
    • Nothing pretending to be an answerSome sites return a normal web page where the file should be. An assistant reads that as broken.
    • Assistants are allowed inIf robots.txt blocks AI crawlers by name, assistants answer about you from other people's pages.
    • Structured data that parsesMachine-readable facts (name, products, prices, hours) that search engines and assistants can quote exactly.
    • The words are in the pageIf the text only appears after scripts run, many crawlers and assistants see an empty page.
    • A sitemap to crawlA list of every page, so crawlers find all of them, not just what the menu links to.

    Search basics

    The basics a search engine reads first: title, description, headings and links.

    • Title tagThe title is the headline in search results and the first thing an assistant reads.
    • Meta descriptionThe summary under the headline in search. Too long and it gets cut off.
    • One H1One main heading tells search engines what the page is about.
    • Title and H1 agreeWhen the title and the main heading share words, the page states its subject twice.
    • Title matches the pageA title that names what the page is mostly about ranks better and gets quoted correctly.
    • Canonical URLTells search engines which address is the real one, so duplicates do not split the ranking.
    • Social cardControls the preview image and title when the page is shared or linked.
    • Enough text to rank onSearch engines and assistants need words to understand a page. Images and video alone do not count.
    • Internal linksLinks between your own pages help crawlers find and weigh them.
    • Images describedAlt text describes images for screen readers and for search.

    Speed

    How much the page costs to load, and Google's own speed test on a phone when it is available.

    • Server responseHow long the server takes to start sending the page. Slow starts slow everything.
    • HTML weightA heavy page takes longer to arrive, especially on a phone.
    • Third-party codeEvery outside script is something you do not control that slows the page down.
    • Render-blocking scriptsScripts that stop the page from showing until they load.
    • Modern image formatsWebP and AVIF images are much lighter than JPEG and PNG at the same quality.
    • Google PageSpeed, mobileGoogle's own speed test on a phone. Speed affects ranking and how many visitors stay.

    Security basics

    Encryption, a few security headers, language and phone layout.

    • HTTPSAn encrypted connection. Browsers warn visitors away from sites without it.
    • HSTSTells browsers to always use the encrypted connection.
    • MIME sniffing offA one-line header that stops browsers guessing file types, a small security gap.
    • Content Security PolicyLimits which scripts can run on the page, which blocks a common kind of attack.
    • Language declaredTells browsers, screen readers and search engines what language the page is in.
    • Mobile viewportWithout it, the page shows as a shrunken desktop page on a phone.
    • AccessibilityWhether people using screen readers and keyboards can use the site.

    Use it from your agent

    Run scans from Claude, ChatGPT, Claude Code or Codex.

    The scanner is also an MCP server. Add it once and ask your assistant to scan a site, read a report or compare you with a competitor. Everything this page does, an agent can do too: scan_site, get_scan_status, get_report, compare_sites, list_recent_scans and list_checks.

    Claude

    Web, desktop and phone

    Settings, then Connectors, then Add custom connector. Paste this address.

    https://scanner.waronsaas.com/mcp

    ChatGPT

    Developer mode

    Settings, then Apps and Connectors, then Create. Paste this address and sign in with your warOnSaaS account when it asks.

    https://scanner.waronsaas.com/mcp

    Claude Code

    Terminal

    One command, then ask it to scan a site.

    claude mcp add --transport http scanner https://scanner.waronsaas.com/mcp

    Codex

    Terminal

    One command, then ask it to scan a site.

    codex mcp add scanner --url https://scanner.waronsaas.com/mcp

    On your own computer

    Command line

    Scan from your machine, print the report or save it as JSON.

    npx github:warOnSaaS/scanner example.com

    From your code

    HTTP

    Reports as JSON, open to anyone. Scanning from code signs in the same way an AI app does. Every route.

    curl "https://scanner.waronsaas.com/api/report?url=example.com"

    Your assistant signs in once, with your warOnSaaS account, and then acts as you. You can see it and sign it out at account.waronsaas.com.

    Try saying: Scan mysite.com and tell me the three fixes worth the most. Compare mysite.com with competitor.com.

    What it does not do

    A first look, not an audit.