How to measure your GEO and AI visibility

Executive summary

GEO measurement tracks a fixed set of buyer prompts across AI engines and records whether your brand is cited, where it sits, and how it is characterised. Run on a schedule, it turns AI visibility from an anecdote into a trend you can manage against competitors.

What this helps you decide

Whether your AI visibility is rising or falling and which prompts need work.

Business problem

AI assistants increasingly answer buyer questions without a click, yet most teams have no idea whether they are cited, ignored or misrepresented. What is not measured cannot be improved or defended.

Step-by-step process

1. 1
   Define the prompt set

   Assemble the decision-stage questions real buyers ask - 'best tool for X', 'X alternatives', 'is X good for Y'. Freeze the set so results are comparable over time.

2. 2
   Choose the engines

   Cover the assistants your buyers actually use. Measuring the same prompts across several engines exposes where you are strong and where you are absent.

3. 3
   Record citation, position and sentiment

   For each prompt note whether you appear, how prominently, and whether the mention is positive, neutral or wrong. A cited-but-misrepresented brand is a different problem from an absent one.

4. 4
   Weight by decision stage

   Score decision-stage prompts higher than definitional ones, and exclude brand-name questions so the headline rate reflects genuine competitive visibility.

5. 5
   Benchmark against competitors

   Run the same prompts for rivals so your share of citations is relative, not absolute. Being cited 40% of the time means little without knowing the leader's rate.

6. 6
   Trend and alert

   Store each run and chart the citation rate over time. Set thresholds that flag sudden drops or a competitor overtaking you.

7. 7
   Feed the improvement loop

   Route low-citation prompts to the content and evidence work that raises them, then re-measure to confirm the intervention worked.

Worked example

Checklist

• A frozen, decision-stage prompt set exists
• Multiple relevant engines are covered
• Citation, position and sentiment are all recorded
• Brand-name prompts are excluded from the headline rate
• Competitor benchmarks are run on the same prompts
• Results are trended and alerting is in place

Common mistakes

• Changing the prompt set between runs so trends are meaningless
• Measuring citation presence but ignoring sentiment and accuracy
• Reporting an absolute rate with no competitor benchmark

30-minute experiment

KPIs to track

• AI citation rate on decision-stage prompts
• Share of citations versus competitors
• Sentiment and accuracy of mentions

FAQs