How to Measure GEO: The Metrics That Actually Matter

Measuring GEO means tracking a system, not one citation count. You can still measure proxies for answerability, discoverability, and trust over time.

By MAIO AI. Published 2026-03-25. Updated 2026-11-06.

Why does GEO measurement use several signals?

There is no single number that tells you how an AI system treats your site. Each platform decides what to cite, answers vary from one run to the next, and most platforms do not report citations back to site owners. So you combine four signals that each answer a different question: is the page ready, are AI crawlers reaching it, do AI tools send visitors, and do AI answers mention you.

None of the four proves that a change caused a result. Together they show whether you are moving in the right direction.

Signal 1: Is your content ready to be cited?

Readiness is about the page itself: a direct answer near the top, headings that match questions, structured data that matches the visible text, and clear trust signals such as authorship and update dates. You control all of it, which makes it the best place to start.

Scan the same set of pages on a schedule and record the score and the fixes it lists. A readiness score describes how well a page is prepared. It does not predict rankings or citations. Compare a score only with the same tool over time, because each tool uses its own formula.

Signal 2: Are AI crawlers reaching your pages?

An AI crawler is an automated agent that fetches pages for an AI service. Its purpose varies: collecting training data, building a search index, or fetching a page live for a user's question. Your server logs, your CDN analytics or a plugin can show which crawlers visit and which pages they read.

Count visits by crawler name and note which pages they read. A page that no crawler reads is the first one to look at. A visit is not a citation, and some live fetchers may not follow robots.txt rules, so check the names in your logs and set a robots.txt rule for each crawler on purpose.

Signal 3: Do AI tools send you visitors?

AI referral traffic is visits that arrive from links in AI assistants or AI search. Look in your analytics for AI tool domains among your referrers, and watch the trend.

Some of these visits arrive without a source, so analytics can undercount them. A visitor who clicks a link inside a chat answer may be labelled with the AI tool's domain, or as direct. Treat the number as a floor, not the total.

Signal 4: Do AI answers mention you?

Prompt tracking means asking the same questions to AI tools again and again and recording how the answers change. Pick 10 to 20 prompts your customers really use, run them on a fixed schedule and keep the answers.

For each answer, note whether your brand is named, whether your site is cited and how you are described. Answers vary between runs, so read trends across many runs, not one answer.

To compare with competitors, count mentions across the whole prompt set. If five brands are named 200 times and yours 50 times, your share of voice is 25 percent. Fix the prompt set and the competitor list first, and measure again later with the same set.

What does a simple monthly routine look like?

Keep one page of notes with four rows: the readiness score of your key pages, AI crawler visits, AI referral visits, and the mention rate on your prompt set. Fill it in on the same day each month.

Change one thing at a time, for example adding an FAQ to one page or fixing a robots.txt rule. Give the change a few weeks, then compare the same measures before and after.

What should you not read into the numbers?

Scores from different tools are not comparable, so do not mix them. A rise in crawler visits is not a rise in citations. A small prompt set can swing from one run to the next. And an AI platform can change how it answers without any change on your site.

Use the numbers to decide what to check next, not to promise a result. No tool can guarantee a citation.

Questions

What is the most important metric to start with?

Start with the readiness of your key pages, because you control it directly. Add crawler visits next, then referral traffic and prompt tracking.

How do I know whether my changes improved AI answers?

Compare the same prompt set before and after a change, over several runs. Look at whether your brand is named and cited more often, and read that alongside crawler visits and referral traffic.

Do I need a paid tool to measure GEO?

No. You can scan pages for free, read crawler names in your server logs, check referrers in your analytics and run a prompt set by hand. A tool saves time and keeps a history.

How long should an experiment run?

A few weeks is a practical window for most changes. Give crawlers time to revisit the pages, then compare before and after with the same method.