How hospital AI search visibility actually moved

Before reading — what the number is

• · Visibility = the share of a hospital's fixed question set on which the hospital is named in the answer, across ChatGPT, Claude, Gemini and Perplexity. Questions containing the hospital's own name are excluded from the denominator.
• · Biweekly. AI answers vary run to run, so weekly changes are mostly noise.
• · The question set is frozen. Changing it breaks comparison with earlier rounds; sets whose questions changed are excluded from this record.
• · Work is shown as status only (in progress / not started). What was done, where and in what order is not published.
• · One hospital can be several rows — branches and area question sets are separate sets.

All records

One observation with a control

Two dental clinics with the same starting point (about 10.9%). One had its homepage remediated; the other did not. Two weeks later the remediated one measured about 26.9% and the other about 10.3%. A single round with two clinics — not a design that supports a causal claim.

What repeats across the records

1. Lower starting points moved more. The 5 tracked sets that started below 5% are now between 11.3% and 34.6%; higher starts tend to move less.
2. The first month was usually quiet. One hospital sat at 0.6% for three rounds before reaching 16% in the fourth. Reading early silence as failure hides what comes after.
3. Nothing rose in a straight line. All 4 sets measured six times had at least one round below the previous one. A single round proves nothing.
4. Some went down. One set fell within two rounds before remediation started; another, mid-remediation, is below its first measurement after four rounds. Both stay in the record.
5. External events moved everything at once. In early August 2026 one engine changed how it grounds answers in web search and several hospitals shifted in the same direction.

Limits of this record

• · API-based measurement; an individual user's app view may differ.
• · One measurement model was changed on 2026-08-19; values before and after are not directly comparable.
• · Question sets differ per hospital, so absolute values are not comparable between hospitals. Compare each set only with its own earlier rounds.
• · Small sample, one control observation — no causal claim between the work and the numbers.
• · All contracted sets under biweekly measurement are included; one-off snapshots without a contract and sets whose questions changed are not.

FAQ

Q. Are there real case records for hospital GEO/AEO work?

A. Yes — Boily publishes the round-by-round visibility of all 23 contracted hospital sets it measures biweekly. Of the 15 sets with two or more rounds, 13 are above their first measurement and 2 are below. Choosing only the ones that rose would itself be an exaggeration, so the declines are kept. Hospitals are identified by area and specialty only.

Q. How much does visibility rise?

A. It varies, and the starting point matters most. The 5 tracked sets that started below 5% are now between 11.3% and 34.6%; higher starting points tend to move less. Nothing rose in a straight line: all 4 sets measured six times had at least one round lower than the previous one. No number is guaranteed; what Boily provides is a before/after record on the same question set.

Q. How long until anything changes?

A. Often nothing moves in the first two rounds (about a month). One general hospital sat at 0.6% for three rounds and then reached 16.3% in the fourth; one dental clinic dropped in round 2 and then climbed over the next four rounds. That is why judgement is made on rounds 3–6, not 1–2.

Q. Is there a control group?

A. One observation. Two dental clinics with the same starting point (about 10.9%): one had its homepage remediated, the other did not. Two weeks later the remediated one was at about 26.9% and the other at about 10.3%. It is a single round with two clinics — not a design that supports causal claims.

Q. Why are the work details missing?

A. Each record states only whether remediation was in progress or not started. What was done, where and in what order is not published. The page exists to show how numbers measured on a fixed question set moved over time, not to describe the method.

Q. Can these numbers be taken at face value?

A. With three caveats. Measurements are API-based and may differ from what a user sees in the app. One measurement model was changed on 2026-08-19, so values before and after are not directly comparable. And in early August 2026 one engine changed how it grounds answers in web search, moving every hospital's numbers at once. These external factors are outside Boily's control and are stated rather than hidden.