Short answer
Use a fixed question set and record the platform, date, and search mode. Save mentions, citations, and factual errors separately. Repeat the checks. Use analytics for recorded visits and customer actions. One answer is a sample.
Steps
- 1Write 5 to 10 promptsUse the questions a buyer asks before choosing, in their words, such as best rank tracker for a small agency.
- 2Run each prompt in each assistantUse a fresh chat with no history so earlier answers do not steer the next one.
- 3Record the outcomeSave the answer, brand mentions, source URLs, and factual errors.
- 4Repeat on a scheduleRepeat the same question set. Compare samples with the same platform and mode.
- 5Improve the pages that should be citedAnswer the question plainly on the page, with the facts an assistant can quote.
What you will learn
You will create a repeatable AI visibility record. You will separate answer exposure from visits and customer actions.
Read SEO, AEO, and GEO first if the terms are new to you.
Define the question set
Choose questions a potential customer would ask. Include the service, market, and decision when relevant.
| Broad input | More useful test question |
|---|---|
| bike repair | Which bike shops repair brakes in my city? |
| SEO tool | Which tool can track keywords for a small agency? |
| coffee | Where can I buy coffee beans online in Jakarta? |
Avoid your own brand name when testing discovery. Use separate brand questions when you want to test factual accuracy.
This collection is a prompt set. It represents selected questions, rather than every customer's behavior.
Keep the conditions in the record
- Save the exact question.
- Record the platform and date.
- Record the model or mode when available.
- Note whether web search was active.
- Save the answer and source links.
Use a fresh conversation to reduce prior-chat effects. Account settings, location, and model changes can still affect results.
Repeat checks under comparable conditions. One response is a sample, and a small difference may reflect answer variability.
Separate the outcomes
| Outcome | What to record |
|---|---|
| Mention | Whether the answer names the business |
| Citation | Whether the answer displays a source URL |
| Accuracy | Whether the stated service, price, or limit is correct |
| Referral | Whether analytics records a visit from the AI service |
| Conversion | Whether a visitor completes the intended action |
A citation does not guarantee a click. A mention without a link can still affect a decision. An inaccurate mention needs investigation.
For a sampled mention rate, divide answers naming your brand by valid answers in the defined set. Keep the denominator and dates visible.
Understand platform reports
Read the current documentation and the fields available in your account. Reports cover different experiences and counting rules.
Google documents a Generative AI performance report for its Search features. Do not present it as a report of all assistant activity.
Bing's AI Performance reports describe citations within supported experiences. Citation share is an observation within the measured set, rather than traffic share.
Analytics can show recorded referral visits. Referrer loss and attribution limits mean it cannot count all AI exposure.
Work through an example
Suppose 6 of 20 valid answers name a workshop. Two answers cite its service page. One answer gives an outdated opening time.
Record 30 percent sampled mentions and the two citations separately. Correct the opening time at its likely sources.
The sample does not prove that 30 percent of all AI users see the workshop. These numbers are a fictional teaching example.
Improve and check again
Correct inaccurate facts. Add missing information that the customer needs. Check crawler access when retrieval may be blocked.
Keep a dated record of the change. Repeat the defined question set. Compare the same outcomes and state the limits of the observation.
Use Track AI in RankMySEO
Track AI records scheduled prompt results for supported assistants. Review the saved mentions, rivals, and cited pages.
Check factual accuracy yourself. Use analytics separately for visits and customer actions. See how AI Visibility works and compare plans for product details.
Connecting an assistant to your rank history is a separate task. Use the assistant connection guide for that workflow.
