Consumers ask AI, not search engines.
Let your embassy speak for your answer.

Scale

The national roster currently lists 7,113 dental clinics, of which 702 already have an embassy page. Sentence-level verification is complete for 27,401 content-verification pages that anyone can read one by one, drawn from 117,982 crawled website pages in total. Institutions that are on the roster but not yet verified stay in the registry with their status stated honestly — dropping them until they are verified would make the word "registry" meaningless.

Volume verified

These pages yield 123,002 checkable outward claims in total. Each one is compared against the medical research abstracts and then assigned one of four states according to the strength of the evidence:

• 1,158Provable
• 16,353Partly supported
• 105,442Pending
• 49Contradicted

Provable

The sentence is directly supported by medical research; the evidence is attached and the source can be opened.

Partly supported

The literature supports part of the sentence, but the scope, conditions or population do not fully match; the attached evidence states how far the support goes.

Pending

The literature was searched and nothing strong enough to support or refute the sentence has been found yet. This is a neutral state and carries no evidence.

Contradicted

Evidence was found pointing the other way. These are usually very few, and we list them anyway rather than hide what looks bad.

"Pending" is usually the largest state, and we show it as it is rather than hiding it. It means we searched the literature for this sentence and have not yet found evidence strong enough to support or refute it — not that the sentence is wrong, and not that anything is wrong with the clinic. Hiding pending items would make the scores look better, but that improvement would be fake.

Where the data comes from

The list of institutions comes from Taiwan's National Health Insurance contracted medical institutions roster. It is an official public roster, and it decides which institutions exist. We do not compile our own list, and we do not select who gets in.

The claims come from each institution's own website. We crawl public pages only and keep every sentence verbatim: no rewriting, no polishing, and nothing put into its mouth. Every sentence on a verification page shows the URL it came from.

The literature used for comparison comes from the PubMed abstract corpus. Supporting evidence carries its PMID, so readers can open the source themselves rather than take our word for it.

Method and limits

What we verify is whether a sentence is supported by literature, not whether a clinic is good at dentistry. This site does not judge clinical outcomes, does not rank practitioners, does not generate recommendations, and does not let anyone pay to change the verification result of a single sentence.

Supporting evidence is attached only in the "provable" and "partly supported" states; "pending" and "contradicted" never carry evidence. This keeps "candidate papers we looked at" from being read as "evidence that supports this sentence" — the two look alike to a reader and are entirely different as evidence.

Fields with no data are left blank or marked as accumulating; we never fill them with a guess. Every number on this page is queried live rather than written into the copy, so it changes on its own as the data grows.

Full scoring methodology →

How to read an embassy

The front page of each embassy is the institution file: name, address, phone, services and identifiers, each field labelled with where it came from, and honestly left blank when there is nothing.

Below the file is the list of content-verification pages, one per crawled website page. Open any of them and you see every sentence extracted from that page, the state each sentence falls into, and which paper is attached to the provable and partly supported ones. This is the only thing this site asks readers to trust: that you can check it yourself, sentence by sentence.

Update cadence

The roster and the verification results are updated incrementally every day; the most recent change to the data was 09/28/2026. Newly crawled website pages enter the verification queue and appear in the corresponding embassy's list of verification pages once they finish — no manual publishing step.