Click Coded · AI-operated, human-reviewed

GEO Is Not Guesswork

Most tools that call themselves "AI visibility checkers" work the same way: ask ChatGPT or Perplexity what it thinks of a site, write down the answer, call it a score. Ask again tomorrow and the answer is different. That's not a measurement. Here's what we check instead, and why it produces a number you can actually act on.

The live-prompt problem

We looked at the AI-visibility tooling landscape directly this week, enterprise platforms charging $29–$250+/mo and several new free entrants, including tools from Ahrefs and Semrush. Nearly all of them share one mechanism: they send a batch of sample prompts to an LLM, read whatever comes back, and report a "visibility score" or "share of voice" built on that answer.

The problem isn't the intent, it's the physics. An LLM's answer to the same prompt varies by the day, the user, the model version, and sampling temperature. A score built on that is a snapshot of a mood, not a fact about your website. Ask twice, get two different numbers, and neither one tells you what to fix.

What we check instead

The AI Visibility Score run by this studio never calls an LLM to ask its opinion. It checks nine technical conditions directly, the same way a browser's "view source" does: a file exists or it doesn't, a schema block validates or it doesn't, a robots.txt entry blocks a crawler or it doesn't.

CheckWhat it verifies
llms.txt presentA plain-text index telling an AI agent what the site offers, without parsing HTML
Structured data (schema.org)JSON-LD or microdata turning prose into machine-extractable facts
robots.txt allows AI agentsNo blanket block on GPTBot, ClaudeBot, PerplexityBot, or similar
Readable without JavaScriptContent present in raw HTML, not rendered only client-side
Answer-first structureKey facts stated directly, not buried after paragraphs of preamble

Full rubric, all 9 checks and exact weights: the published methodology. Same weights power the free checker, every industry benchmark, and every paid audit this studio sells — one spec, checkable by anyone.

Why this matters more, not less, as the tooling gets crowded

A prompt-based score can't tell you what to fix, because there's no fixed thing to point at, just an LLM's answer that happened to come back a certain way. An infrastructure-based score can: "add an llms.txt file" is a concrete instruction. "Get the AI to think of you more favorably" is not.

This is the same discipline this studio's benchmarks apply to every industry audited: same input, same method, every time, no live AI query anywhere in the pipeline. Even the AI companies' own homepages get checked this way, and 30% of them blocked the fetcher outright.

Curious how this stacks up against named tools, not just "enterprise platforms": Fokal, Profound, Otterly, and 3 more, compared honestly.

Try it on your own site

9 checks, 0 API calls Paste a URL into the free checker and get the same rubric described above, run against your own site, in your browser, in about a minute.

Run the free check

Start here, no signup. If you want the full method to fix what it finds: read the rubric or see how 158 real business homepages scored against it. Prefer a plainer walkthrough first: Can AI Assistants Actually Read Your Website?

Want to hear when we ship something new?

We add new free tools and benchmarks periodically. Get an email when we do, no spam, unsubscribe anytime.

Built by Click Coded, AI-operated and human-reviewed. Contact: alexander.k.eliot@gmail.com