Methodology

Where the records come from

The facility list is read from the Republic of Türkiye Ministry of Health's own public interface for its health tourism portal, healthturkiye.gov.tr. It is a public endpoint: no account, no key, no agreement. Documents are mirrored from the Ministry's own document CDN, which is where the portal itself serves them from.

No record here is typed in by hand, and none is bought from a third party. Where a field did not come from the Ministry, the page says so.

There are two official lists, and they do not match

The Ministry publishes the authorised facilities twice, in two different places, and the two are not the same size.

• The health tourism portal at healthturkiye.gov.tr, which is what this site reads. It held 4,403 facilities when last read.
• The Health Tourism Department's published lists at shgmturizmdb.saglik.gov.tr, issued as four documents: authorised hospitals, medical centres, private practices and other health facilities.

We compared them on 18 September 2026, against department documents dated 16 September 2026. Both sources publish each facility's registered legal name, so the comparison is name against name, normalised only for case, Turkish characters, punctuation and the connector ve. Our copy of the portal's names was checked against the Ministry's own interface first and matches it exactly, so the comparison rests on the register's wording rather than ours.

316 facilities from the department's lists are published here with no portal entry behind them. 147 are in Istanbul, 58 in Antalya, 23 in Ankara and 20 in Izmir, and they are overwhelmingly small: private dental polyclinics and single-doctor practices.

Far more were held back than published, deliberately. A row is not published when it shares an uncommon word with any portal entry, which is how “Özel Simyadent Ağız ve Diş Sağlığı Polikliniği” was kept out once the portal turned out to carry it as “SimyaDent Oral and Dental Health Clinic”. Public hospitals and university hospitals are held back as a group: the department's list writes “Adana Şehir Hastanesi” where the portal writes “Adana Şehir Eğitim ve Araştırma Hastanesi”, which is the same hospital under two official spellings. Comparing registered names cannot catch that, so the whole group is held back rather than published on a name match that happens to fail. Rows whose text did not parse cleanly are held back too. A facility wrongly shown as undocumented is a worse error than one not yet listed.

The figure moved several times before it settled, because the method did. The lists were first read as flat text, which cannot tell a facility's name from the type column beside it, and every comparison built on that was comparing damaged names. They are now read as the table they are, by column position, so the registered name arrives exactly as the Ministry prints it and is published unchanged. A row counts as already in the portal when the registered names match outright, or when the distinctive words in both names overlap and the addresses agree — address alone decides nothing, because a medical building holds many clinics and two unrelated practices share one street number.

The likeliest reason the two lists differ at all is that they record different things. The department's list records that a facility was authorised. The portal is a directory a facility registers itself into, which the 2025 regulation sets out as a separate obligation — if it followed automatically from being authorised, the regulation would not need to require it. That fits the shape of the gap: what is missing from the portal is mostly the smallest facilities, the least likely to complete an extra administrative step.

What this means for this site. A facility absent from here may still hold the authorisation, which is one of the reasons this site never treats absence as proof. The 278 that are listed are marked: in place of an authorisation number they carry Document not verified, in search results and on their own page, and the page says plainly that no document is held and that this is not a finding against the facility. Any of them can send us the certificate, and it is checked against the register before anything changes.

How often it is refreshed

A scheduled job reads the whole register once a day at 03:17 UTC. Each run writes a snapshot row recording the time and how many entries the register returned, so the history of the site's own observations is itself a record. The counters on this site were last refreshed on 2026-09-18.

How a record is matched to the Ministry's entry

Every facility is matched on the register's own numeric identifier, which is the primary key of our table. Records are never matched by name.

This matters because name matching is where directories go wrong. Turkish facility names repeat across cities and branches, differ by accent and abbreviation, and change. Keying on the Ministry's identifier means a clinic that renames itself is still the same record, and two clinics with nearly identical names are never merged.

Names and addresses that change

A change of registered name is recorded as an event against the same identifier, with the old value and the new one and the date we saw it. The old name remains searchable, because that is the name a patient may have been given months earlier.

Addresses come from the register where the register carries them, and are otherwise left empty. 4,403 facilities are listed; an address is held for most but not all of them.

Records that disappear or are suspended

Nothing is deleted. A facility present in one day's register and absent the next is marked as no longer listed, with the date, and its page stays up saying so. That is the whole point of keeping a time series: the state portal shows only the present, so without this, a facility's removal would be invisible.

Absence is not proof. A facility may be missing from a search because it is registered under a different legal name, because a spelling or accent differs, or because it genuinely holds no authorisation. These look identical from the outside and this site does not pretend to tell them apart.

Clinic-declared data

Some of what you see is not from the Ministry. It is marked as such, and it is:

• Websites. 2,959 facilities have a website here. It was not supplied by the clinic: it was derived from the domain of the email address in the clinic's own register entry, then fetched and checked. A domain is published only if it answered, was not a placeholder or a parked page, and the page itself named the facility. 287 domains answered but named nothing we could tie to the facility and are not published; 57 were placeholders and 183 did not answer at all.
• Descriptions. 1,824 facilities carry a description written by the clinic. It is the summary the clinic publishes for its own site to be quoted by (og:description or the page's meta description) and nothing else: no page text is scraped. It is shown only where the register itself carries no description.
• Social accounts. Linked where the clinic publishes them.

What this site verifies

• That the entry existed on a given date. Each snapshot is dated and counted.
• That the published document is the file we downloaded. Every mirrored PDF carries the SHA-256 of its bytes, and 6,402 of them are republished as signed W3C Verifiable Credentials that anyone can check independently of us.
• That a published website resolves and names the facility. See above.

What this site does not verify

• Quality of care. Nothing here is an opinion about treatment, outcomes or safety at any facility. An authorisation is a permission, not a grade.
• That a listed doctor still works there. 29,099 clinicians are on record from the register. People move, and the register is not instant.
• Anything a clinic says about itself. A clinic-written description is reproduced, not endorsed or fact-checked.
• Prices. No price for any treatment appears anywhere on this site.
• Accreditations a clinic claims. Where the register records a count of accreditations or certifications, that count is shown as the register's claim. Where the named documents exist they are published; where only a number exists, only the number is.

Known limits

• Our own history starts recently. This site began observing the register in September 2026. It can say what it has seen since; it cannot describe the years before.
• Document numbers and dates are read by machine. Numbers and dates inside mirrored PDFs are extracted by optical character recognition. Where a value did not read cleanly it is left out rather than printed, because a wrong licence number is worse than a missing one.
• The English wording is the Ministry's. Facility types, treatments and branches use the register's own English text where it exists, not our translation.

If something here is wrong, tell us — that page explains exactly what happens next.