Dental FAQs

Can I trust what ChatGPT or Google's AI says about dental work overseas?

Trust it for general explanations, and check it for anything you will act on. In studies where dentists graded chatbot answers to common patient questions about implants and dentures, the newer chatbots scored well on general questions. They did worse on specific cases: in a 2026 study of implant complications, the share of answers rated safe ranged from 56.7% to 86.7% across general-purpose chatbots, and the share free of made-up content from 36.7% to 93.3%. References are the weakest part: in one study 47% of the references ChatGPT gave were invented and only 7% were real and correct. For a trip abroad, AI answers also lean on the pages they can find, which are often booking sites, clinic marketing and forum threads, and they cannot know a clinic's current price or what your X-rays show. Use AI to learn the words and the questions to ask. Then check every number against a written quote, every claim against its source, and every plan with a dentist who has seen your mouth.

  • Answers rated safe: 56.7% to 86.7% across general-purpose chatbots, on implant complication cases (Mayer 2026).
  • References given by ChatGPT 3.5: 47% invented, 46% real but wrong in some detail, 7% real and correct (Bhattacharyya 2023).
  • Implant questions, common ones: ChatGPT-o1 averaged 4.96 out of 5 for accuracy in a dentist-graded test of 45 questions (Tuzlalı 2025).

What AI gets right and wrong, at a glance

Kind of questionHow AI usually doesWhat to do
"What is an implant abutment?" General explanationsUsually good with newer chatbots (Tuzlalı 2025)Use it to learn the words
"What should I ask the clinic?"Often useful listsAdd the questions on our checklists
"My implant feels loose, what is wrong?" Specific casesMixed: safe answers ranged from 56.7% to 86.7% (Mayer 2026)See a dentist; AI cannot examine you
"Show me the studies"Weak: many references are invented or wrong (Bhattacharyya 2023)Open every reference yourself
"How much does it cost in Vietnam?"Repeats ranges from pages it finds, often with no date or sourceCheck against a written, itemised quote
"Which clinic should I choose?"Repeats what booking sites and clinics say about themselvesCheck licences, plans and records yourself

AI chatbots and AI search summaries write a fluent answer from patterns in their training and from pages they can find. They do not examine you, call a clinic, or read your X-rays the way a dentist does, and they rarely tell you how sure they are.

What the accuracy studies found

Dentists have tested chatbots on dental questions several times since 2023. The results depend on the chatbot, its version, and above all the kind of question.

StudyWhat was testedWhat it found
Molena 2024 (Cureus)ChatGPT 3.5 on 30 questions written by 10 experts in six dental fieldsGood accuracy on average, but some wrong answers persisted, and its references were more often wrong than right
Özcivelek 2025 (BMC Oral Health)Four chatbots on 35 common patient questions about dental and facial prostheses, such as denturesAccuracy differed between chatbots; a dental-trained chatbot and ChatGPT-o1 scored highest, ChatGPT-4 lowest
Akpınar 2025 (BMC Oral Health)ChatGPT 4.0, OpenEvidence and MediSearch on 20 All-on-4 questions, graded by oral surgeonsSignificant differences between chatbots on both patient and technical questions
Tuzlalı 2025 (BMC Oral Health)Five chatbots on 45 common implant questionsChatGPT-o1 scored highest, with mean accuracy of 4.96 out of 5; Gemini Advanced and Perplexity Pro were weaker on completeness and clarity
Mayer 2026 (Journal of Dentistry)Eleven AI platforms on 30 cases of implant complications, scored blind by four experienced implant dentistsSafe answers 56.7% to 86.7% and answers free of made-up content 36.7% to 93.3% among general-purpose chatbots; a medical tool grounded in published studies had every answer rated safe and free of made-up content

Two patterns run through these studies. First, answers to common questions are often good, and newer versions have tended to score higher. Second, the more specific and clinical the question, the more the chatbots differ and the more often an answer is unsafe or contains something made up. The authors of the 2026 study say all platforms should support a dentist's judgement, not replace it.

These studies are small, use set questions rather than real conversations, and are graded by a few experts. Chatbots also change every few months. So the numbers show the kind of errors to expect, not a fixed score for any product today.

Invented references

Asking an AI for its sources does not settle the question by itself, because a chatbot can write a reference that looks real and is not.

  • Bhattacharyya 2023: ChatGPT 3.5 wrote 30 short medical papers with 115 references. 47% were invented, 46% were real but inaccurate, and only 7% were real and accurate.
  • Gravel 2023: ChatGPT answered 20 medical questions with references. Of 59 references checked, 41 (69%) were invented, though they appeared real. The experts also found 5 major and 7 minor factual errors in 17 answers.
  • Molena 2024: in dentistry, ChatGPT's references were more often incorrect than correct.

AI tools that search the web while answering, and link the pages they used, do better on this, because you can click through. The link still needs a look: the page may not say what the summary says it does.

Why AI answers about dental trips repeat marketing

An AI answer can only be as good as the pages behind it. For dental work abroad, most pages are written by people selling something: clinics, booking sites, directories and agencies, including us. Forum threads add real stories but rarely check them.

  • When we ran five test searches about dental work in Vietnam in October 2026, the sources behind the AI-style answers were mostly booking sites, directories, travel articles and forum threads.
  • One large booking site came up for almost every cost question. When we read its implant page, its own price ranges had errors.
  • Different AI tools draw on different sources and search in different ways, so two tools can give two different answers to the same question.
  • Google's AI Overviews have been criticised for health errors. In January 2026 a Guardian investigation reported misleading answers on liver blood tests and on diet for pancreatic cancer, and Google reportedly removed some of those summaries (Euronews, 12 January 2026).

None of this makes AI useless. It means an AI answer about a clinic, a price or a country is a summary of other people's claims, and it should be checked like one.

Prices from AI: why they are often wrong

  • Prices change. A page an AI read last year may show last year's price, and the answer rarely says how old it is.
  • Ranges hide the brand. An implant price depends on the implant brand, the crown, and whether you need a bone graft, and a single range mixes them all.
  • The total depends on your mouth. Only a dentist who has seen your X-rays or 3D scan can say how many implants or crowns you need.
  • Currency is often mixed up. Prices in US, Australian and New Zealand dollars look alike.

Our own prices are on the pricing page, with each brand and what it includes, and they are dated. The figure that counts is the one on a written, itemised plan for your mouth.

Can AI read my X-rays or photos?

Many chatbots now accept images, and some patients upload an X-ray or a photo of their teeth to ask what is wrong. The first tests are not encouraging for this use.

  • Mine 2026 (Journal of Dentistry) asked three general chatbots that read images to find implants on 164 panoramic X-rays, the wide X-ray that shows both jaws. They found 17% to 69% of the 297 implants present.
  • On X-rays with no implants at all, they still marked implants that were not there: they correctly found none on only 32.7% to 68.5% of those images. Fillings and crowns, which also show white on an X-ray, made this worse.
  • The same chatbot gave different answers on different runs. Even the more consistent models found the same implant in all five runs for only about a quarter of implants.

The authors concluded that these errors limit use without a clinician. A dentist reading your X-ray or 3D scan also looks at bone, gums and the bite, which a chatbot does not do reliably. Our page on how to read a dental X-ray explains what a dentist looks for.

How to check what AI tells you

  1. Ask the AI for its sources, then open them. If a link does not work or the page does not say what the answer claims, do not rely on that part.
  2. For a study, look up the title on PubMed or Google Scholar. If you cannot find it, treat it as invented.
  3. For a price, check the date and the currency, then compare it with a written quote that lists each treatment, each tooth and each brand.
  4. For a clinic, check what you can yourself: the licence, the written plan, the brands, and what happens after you fly home.
  5. For your own symptoms or X-rays, ask a dentist. An AI cannot examine you.
  6. Ask the same question in two different tools. If they disagree, that is the part to check with a person.

Good ways to use AI before a dental trip

Use it forWhy it works
Explaining dental words on a quoteGeneral explanations are where chatbots score well
Drafting questions for the clinicA longer list is easy to trim
Summarising a long page you have readYou can check the summary against the page
Planning the non-dental parts of the tripLow stakes, easy to check
Comparing two written plans side by sideIt can spot differences; a dentist decides which plan fits you

Common beliefs that are not quite right

What people hearWhat the evidence says
"AI answers are neutral."They summarise the pages they can find, and for dental trips many of those pages are marketing.
"If it gives a reference, it is true."In one study 47% of references were invented and only 7% were real and correct (Bhattacharyya 2023).
"All chatbots are about the same."Studies found large differences between chatbots and versions (Özcivelek 2025, Mayer 2026).
"AI is always wrong about health."On common implant questions, newer chatbots scored highly (Tuzlalı 2025). The risk is in specific cases and sources.
"The AI price is what I will pay."Only a written plan for your mouth gives your price.

What The Hudson does

We write our pages with named sources and dates, so you or an AI tool can check them. We do not ask you to trust a summary: send us your X-rays and we get a written, itemised plan from our partner clinic, with each tooth, each brand and each price, before you book anything.

How we make money

You don't pay us anything. When you go ahead with treatment, the clinic pays us a referral fee for each patient. It doesn't change your price: you pay the clinic's normal price, the same as if you walked in yourself.

Quick answers

Is ChatGPT accurate about dental implants?

On common questions, newer versions scored highly in dentist-graded studies (Tuzlalı 2025). On specific complication cases, safe answers ranged from 56.7% to 86.7% across general chatbots (Mayer 2026).

Are the references ChatGPT gives real?

Often not. In one study, 47% of ChatGPT's medical references were invented and only 7% were real and accurate (Bhattacharyya 2023). Open every reference yourself.

Can I trust AI prices for dental work in Vietnam?

Treat them as rough. They come from pages that may be old, mix brands and currencies, and cannot reflect your X-rays. Compare with a written, itemised quote.

Can AI tell me which dental clinic to choose?

It repeats what it finds, which for dental trips is mostly booking sites, directories, clinic marketing and forums. Check licences, the written plan and aftercare yourself.

What is AI good for before a dental trip?

Explaining dental words, drafting questions for the clinic, and summarising pages you can check. Leave diagnosis and choosing a plan to a dentist.

Want an answer you can check?

Send us your X-rays or your quote. We get a written, itemised plan from our partner clinic, with each tooth, brand and price. Free.

Get my written plan