To measure answer engine optimization, choose 20 to 50 questions your customers ask, run each one many times on several AI platforms, and record how often your business is named, linked, and described correctly. Then add referral traffic from AI sites and branded search to your analytics. The bottom line: because AI answers change from run to run, the unit of measurement is the share of many runs, never a single answer. This guide supports our main explainer on answer engine optimization.
Why single answers prove little
SparkToro and Gumshoe tested AI recommendations and found less than a 1 in 100 chance that two responses returned the same list of brands. A business named in one answer may be missing from the next. That is why comparison guides describe tools that repeat each prompt on a schedule, and why a manual check once a quarter is not a measurement.
The metrics
| Metric | What it is | How to calculate it |
|---|---|---|
| Mention rate | How often an answer names your business, with or without a link | Runs that name you divided by total runs |
| Citation rate | How often an answer links to your site | Runs that link to you divided by total runs |
| Share of voice | Your mentions compared with named competitors | Your mentions divided by all brand mentions in the same runs |
| Accuracy | Whether the answer describes you correctly | Correct answers divided by runs that name you |
| Sentiment | How the answer describes you | Positive, neutral, or negative, by reading the answer |
A worked example
Suppose a dental practice picks 30 questions, such as "who is the best family dentist in Plano" and "how much does an emergency dental visit cost," and runs each one 5 times on 3 platforms. That is 450 runs (30 times 5 times 3). If the practice is named in 90 runs, the mention rate is 20 percent (90 divided by 450). If it is linked in 36 of those, the citation rate is 8 percent (36 divided by 450). After two months of work on its pages and reviews, the same 450 runs show 135 mentions. The mention rate is now 30 percent, a 10 point gain. The example is illustrative, with numbers chosen to show the arithmetic.
Build the prompt set
- Collect real questions. Use the questions customers ask on calls, in emails, and in reviews.
- Cover the buying journey. Include "what is," "how much," "best," and "near me" questions.
- Include your category and your name. Track both unbranded questions and questions that name your business.
- Fix the wording. Use the same prompts each month so the trend is comparable.
- Add your competitors. Record who else is named.
A tracking sheet
| Date | Platform | Prompt | Run number | Named? | Linked? | Described correctly? | Competitors named | Notes |
|---|---|---|---|---|---|---|---|---|
| October 4, 2026 | ChatGPT with search | best family dentist in Plano | 1 | Yes | No | Yes | Two others | Hours were wrong |
| October 4, 2026 | Perplexity | best family dentist in Plano | 1 | No | No | n/a | Three others |
Add analytics and search data
- Referral traffic. Look for visits from chatgpt.com, claude.ai, and perplexity.ai, which are the domains the operators publish. Referral volume is usually small, but it shows which pages get clicked.
- Branded search. A rise in searches for your name can follow more AI mentions, though it has other causes.
- Google Search Console. Track impressions and position for your target queries, as you would for any SEO work.
- Crawler logs. Check your server or CDN logs to confirm
OAI-SearchBot,Claude-SearchBot, and PerplexityBot are reaching your pages. - Leads. Add "how did you hear about us" to forms and calls, with ChatGPT and Google AI as options.
How to read the results
Compare the same prompt set month over month, and look at the trend across at least three months. A gain in mention rate with no gain in leads means the prompts may not match buying questions. A gain in citations with a drop in accuracy means the engines are quoting an outdated page. Our guides to answer engine optimization tools and the answer engine optimization checklist help with the next step, and you can run our free answer engine optimization check for a baseline.
Questions people ask
How do you measure answer engine optimization?
Run a fixed set of customer questions many times on several platforms, and track the share of runs that name, link, and correctly describe your business. Add referral traffic and branded search.
Why run each prompt more than once?
Answers vary. SparkToro and Gumshoe found less than a 1 in 100 chance that two responses returned the same brand list.
What is a good mention rate?
It depends on your category and competitors. Compare your own trend and your share of voice against named competitors.
Can I measure it for free?
Yes, with a spreadsheet and a fixed list of prompts, though it takes time. HubSpot's free AEO Grader gives a one time view.