Ask ChatGPT where to find a good accountant in your city and it will name three or four companies. It will not show ten blue links. It will recommend. Perplexity does the same and adds footnotes. Google's AI Overviews increasingly answer the question before anyone scrolls to the results.
If your website is not among the sources these engines can see, read and quote, you are not "ranking lower". You are simply not in the conversation.
This is what GEO - Generative Engine Optimization - is about: making your website visible and citable for AI answer engines, the same way SEO made it visible for classic search.
AI engines don't browse like people do
An AI assistant answering a customer's question doesn't render your animations or admire your design. Behind the scenes, a crawler fetches your pages as plain HTML and the engine decides whether your content is worth quoting. That pipeline fails far more often than site owners expect, usually in one of five places:
1. The crawler never gets in. robots.txt rules written years ago (or copied from a template) often block bots like GPTBot, OAI-SearchBot, PerplexityBot or ClaudeBot - sometimes deliberately, sometimes by a blanket Disallow. A firewall or anti-bot service can do the same damage silently, returning 403 to anything with an unusual user-agent. Blocked crawler, no citation. It is the GEO equivalent of a noindex on your homepage.
By the way, "I don't want AI trained on my content" and "I want AI to recommend me" are compatible goals. Training crawlers (GPTBot, CCBot, Google-Extended...) and search/assistant crawlers (OAI-SearchBot, ChatGPT-User, Perplexity-User...) are different bots. You can block the first group and welcome the second - it takes a few explicit lines in robots.txt.
2. There is no map for machines. llms.txt is a simple markdown file at your site root that tells AI engines what your site is about and which pages matter. It is young, it is not a magic ranking switch, but it costs nothing and the engines that look for it get exactly the curated summary you want them to have.
3. The engine has to guess who you are. Structured data (JSON-LD: Organization, LocalBusiness, Product, FAQPage...) turns "some website that mentions plumbing" into "a plumbing company in Ostrava, open weekdays, serving these areas". Engines strongly prefer sources they can identify. Most small-business sites have no structured data at all.
4. The content only exists after JavaScript runs. Most AI crawlers do not execute JavaScript. If your site is a single-page app that renders everything client-side, a crawler sees an empty shell. The fix is server-side rendering or prerendering of the pages that matter - the same advice as classic SEO, just with higher stakes.
5. There is nothing quotable. AI answers are built from short, direct passages: a question as a heading, a concrete answer in the first paragraph under it, a visible date, an identifiable author. A wall of marketing adjectives gives the engine nothing to lift. An FAQ gives it everything.
Measure it instead of guessing
All five points above are checkable by machine - no opinions needed. That is exactly what our free GEO test does:
- fetches your robots.txt and evaluates 17 real AI user-agents (allowed / blocked / default),
- probes whether a firewall treats an AI crawler differently from a normal visitor,
- checks llms.txt, structured data types and their required fields,
- estimates how much of your content survives without JavaScript,
- scores how quotable your pages are (question headings, FAQ, dates, author signals),
- and then does the thing everyone actually wants to know: asks a live AI engine a realistic customer query, with web search on, and checks whether your site shows up in the answer.
You get a GEO score and the top findings on screen. The full PDF report - free, by email - includes ready-to-upload fixes: the exact robots.txt lines (including the "visible in AI answers, but no training" variant), a generated llms.txt built from your real pages, and a JSON-LD block for your <head>.
[Run the free GEO test](/geo) - one test per domain per day, no sign-up.
And then keep watching it
A GEO check is a snapshot. Robots files get overwritten by deploys, plugins rewrite headers, and AI answers themselves shift week to week as engines re-pick their sources.
That is why GEO is now the fifth watcher in SiteBot24's Full Service, next to uptime, SSL, SEO and content watch: once a week it re-checks crawler access and structured data, asks AI engines the same 10 realistic customer queries about your market, and your Monday report says in plain words what changed - "Perplexity cited you for 2 queries last week, 0 this week, and here is the likely reason."
One URL. Five watchers. No dashboards to decode.