Summary
The seven GEO metrics that actually matter are citation rate, share of voice, sentiment, prompt coverage, source and position, answer accuracy, and trend over time. Together they tell you whether AI engines mention you, how often versus competitors, in what tone, across which questions, from which sources, with what accuracy, and which direction it is all moving. This guide defines each one and explains how to measure it.
Why this matters now
Classic SEO metrics — impressions, rankings, clicks — say nothing about whether ChatGPT, Gemini or Perplexity recommend your brand. When buyers ask an AI engine instead of scrolling a results page, you need a different scoreboard. These seven metrics are that scoreboard.
What Makes a GEO Metric Worth Tracking
A useful GEO metric has to answer a decision, not just decorate a dashboard. The bar is simple: it should change when your AI visibility changes, it should be comparable over time, and it should point to an action. Vanity numbers that move with model updates you cannot influence are noise. The seven below clear that bar — each maps to a lever you can actually pull.
Citation Rate
Citation rate is the percentage of relevant prompts where an AI engine mentions or cites your brand. It is the single most important GEO metric because it measures presence at all: if you are not cited, nothing else matters. Measure it by running a fixed set of buyer questions across each engine and counting the share where your brand appears. Track it per engine, because being strong in Perplexity and absent in ChatGPT is a very different problem from being weak everywhere.
Share of Voice
Share of voice is your citation rate relative to competitors on the same prompts. A 40% citation rate sounds great until you learn a rival sits at 80% on the questions that drive revenue. Measure it by tracking the same prompt set for you and your named competitors, then expressing your mentions as a share of all brand mentions. This is the metric executives care about most, because it frames AI visibility as a competitive race rather than an absolute score.
Sentiment and Framing
Being mentioned is not the same as being recommended. Sentiment captures whether the engine frames you positively, neutrally or with caveats — described as a leader, listed as one option among many, or flagged with a weakness. Measure it by classifying the tone of each mention and watching how framing shifts as you publish or as reviews accumulate. A rise in neutral or negative framing is an early warning that your reputation signals need work.
Prompt Coverage
Prompt coverage is the breadth of questions where you show up, across the full buyer journey. You might dominate "best tool for X" yet vanish on "X alternatives" or "how to do X". Measure it by building a prompt universe — a structured map of the questions your audience asks, from problem-aware to purchase-ready — and tracking citation rate per cluster. Coverage gaps tell you exactly which content to create next.
Measure all seven with Refine
Refine tracks citation rate, share of voice, sentiment and prompt coverage across ChatGPT, Gemini, Perplexity, Claude, Copilot and Google AI surfaces — on a schedule, so the numbers are comparable week over week. Instead of spot-checking prompts by hand, you get one scoreboard that shows where you win, where competitors beat you, and what to fix.
Source and Position
When an engine cites you, where does it pull from, and how prominently? Source tells you which pages, reviews or third-party sites the engine trusts enough to quote — your own site, a review platform, a Reddit thread. Position tells you whether you are the lead recommendation or a footnote. Measure both by inspecting the citations behind each answer. They point you to the assets worth strengthening and the placements worth earning.
Answer Accuracy
AI engines confidently state wrong facts — outdated pricing, a feature you shipped a year ago, a positioning that no longer fits. Answer accuracy tracks whether what the engine says about you is actually true. Measure it by reviewing the substance of each mention against reality and logging errors. Inaccuracies are fixable: they usually trace back to stale or thin sources the engine is leaning on, which you can refresh or correct.
Trend Over Time
Every metric above is far more useful as a trend line than as a one-off snapshot. AI answers shift as models update and as you publish, so a single reading tells you little. Measure trend by running the same prompts on a fixed cadence and tying each movement back to what you shipped. The teams that compound are the ones watching direction, not just today is number.
- Citation rate — are you mentioned at all?
- Share of voice — how do you compare to competitors?
- Sentiment — are you recommended or merely listed?
- Prompt coverage — across which questions do you appear?
- Source and position — from where, and how prominently?
- Answer accuracy — is what the engine says about you true?
- Trend over time — which direction is it all moving?
The Bottom Line
GEO is not measured in impressions or rankings. Start with citation rate and share of voice to know where you stand, layer in sentiment, coverage, source, position and accuracy to know why, and watch all of it as a trend so you know whether your work is paying off. Pick the prompts that matter, measure on a schedule, and act on the gaps.
Short on time? Have an assistant summarise this page for you.