AI Visibility · Private Practices

How do you measure if AEO is working?

Last updated: 2026-06-06
By Dev Sardana, Founder, Tenva
The direct answer

You measure whether AEO (answer engine optimization) is working by re-asking the same fixed set of patient questions to each AI engine on a schedule, then counting how often those answers cite or name your practice. Compare every count against a dated starting baseline. Movement against the same questions, not screenshots, proves progress.

What should you actually count?

Count two distinct things separately, engine by engine: whether an answer links to your website, and whether an answer says your practice name without a link. A citation carries a clickable source; a mention is recognition without a link. Both matter, and they move at different speeds, so collapsing them into one number hides what is changing.

Tie every count to a frozen list of the questions real patients ask ChatGPT, Claude, Perplexity, and Gemini before they book. AI share of voice is the share of AI answers that cite or mention your practice. If the question list drifts between checks, the number is no longer comparable and the measurement is worthless.

How often should you re-measure?

Re-probe monthly with the identical question set. AI answers vary between runs, so a single check on a single day can mislead you in either direction. A monthly cadence smooths that noise and turns a twitchy snapshot into a trend you can actually read.

Run every question across all four engines each cycle and record the date. Tenva re-measures itself this way, and the published numbers show why one reading is never enough.

What does real progress look like?

Real progress starts with an honest, dated zero. Before optimization, Tenva published its own baseline of zero citations across forty AI answers to its buyer questions, then tracked the same questions forward. A credible result is movement against that fixed set over weeks, not a lucky answer captured once.

Watch the direction of citations and mentions against your baseline, per engine. Gaining ground in Perplexity while staying flat in ChatGPT is a finding, not a failure, and you only see it because the questions and engines never changed between runs.

What should you refuse to trust?

Distrust one-off screenshots and any promise of an AI ranking. No vendor controls what ChatGPT or Gemini says. Guarantees of a fixed position or a set number of new patients describe an outcome nobody can deliver. The measurement surface itself has no authoritative source yet.

Tenva's probes found that twelve of sixteen buyer questions about AI visibility have no source the engines agree on. That vacuum is exactly why your own repeated measurement, against your own questions, is the only honest scoreboard available right now.

Frequently asked questions

What is the single best metric for AEO?
There is no single metric. Count whether an answer links to your website and whether it says your practice name without a link. Do this per engine against a fixed question set, and watch both move over time.
How is measuring AEO different from checking Google rankings?
AI answers have no stable ranked list to check. Instead of one position, you re-ask each engine the same patient questions and count how often it cites or names your practice. The unit is answer presence over runs, not a search position.
Why measure monthly instead of just once?
AI answers shift between runs, so one check can read high or low by chance. Re-probe monthly with the identical question set so noise averages out and you see a real trend rather than a single misleading moment in time.
Do I need a starting baseline before optimizing?
Yes. Record a dated baseline first, then track the same questions forward. Tenva published its own zero before optimizing. Without a frozen starting point you cannot tell improvement from the normal run-to-run variation in AI answers.
Can a vendor guarantee my practice will rank in AI answers?
No. No vendor controls what ChatGPT or Gemini says, so any guaranteed ranking or promised patient count is not credible. Honest AEO work improves the odds that engines recommend your practice and proves it with repeated measurement.
Why is there no off-the-shelf AEO scoreboard yet?
The measurement surface is new and unsettled. Tenva found twelve of sixteen buyer questions about AI visibility have no source the engines agree on, so your own repeated measurement against your own questions is the most reliable scoreboard today.

See what AI says about your practice.

Tell us your practice type and city. We run your practice through the same multi-engine check used on this page and walk you through the results on a call.

Check my practice