Continuing Professional Development, Evidence Evaluation and Future Preparation

Gains:

• Ability to use AI-supported literature scanning and summarization tools together with the discipline of evidence-based dentistry
• Ability to detect and critically verify fabricated sources, misattributions, and outdated information in AI output
• Ability to create a personal roadmap and learning plan that gradually, safely and ethically integrates AI into the clinic

Dentistry is a rapidly changing science; An application valid today may be updated tomorrow. That's why continuing professional development (CPD) and evidence-based dentistry (practice supported by scientific evidence) are integral to good medicine. AI can be a powerful tool here by accelerating literature scanning, article summarization, and organizing information. But the same AI can fabricate non-existent articles, misattribute, and present outdated information in persuasive language. In this final unit, we will learn how to safely use AI in learning and evidence evaluation and establish a roadmap that gradually integrates AI into your clinic.

Immutable principle: AI accelerates learning but does not verify evidence. Every reference and claim must be verified from an original, current and reliable source.

Evidence screening with AI: strengths and dangers

AI helps with:

• Quickly summarizing the general framework of a topic.
• Translating a complex article into plain language.
• Generating questions that help you think about a question from different perspectives.
• Explain terminology and concepts.

However, it carries the following dangers:

• Hallucination: Fabricating a non-existent article, author, journal or DOI (digital identification number of articles).
• Recency gap: Not knowing new evidence because the training data is up to a certain date.
• Loss of context: Generalizing a finding by detaching it from the context in which it is valid.
• Source mixing: Inadvertently combining results from different studies.

Beware: Even if a language model gives you a real-looking DOI and a complete author list, that article may not exist at all. Do not use each reference without verifying it from the source itself.

Evidence pyramid and verification

In evidence-based dentistry, not all information has equal weight. Systematic reviews and well-designed clinical studies are stronger evidence than expert opinion or single case reports. When AI presents information, ask: what is the level of evidence for this information, is its source current, can it be applied to my patient?

step

Question

Is there a source?

Is the attribution real, verified from the original?

What is the level of evidence?

Review, clinical study, opinion?

Is it up to date?

Have the new guidelines changed this?

Is it applicable?

Does it fit my patient profile?

Mini case 1: Capturing the fake source

A physician requests a “sourced summary” from the AI ​​about a treatment approach. AI yields three articles; two are extremely realistic, with full author, journal and year information. The doctor searches the database for references: one is real, two never exist. The physician uses only the real thing and eliminates the others. Lesson: Every resource given by AI should be verified against the original before being used.

Mini case 2: Outdated recommendation

AI offers a recommendation on a topic that was valid a few years ago but has since been updated. The physician looks at the current professional association guide and sees that the recommendation has changed and shapes his practice according to current evidence. Lesson: AI's "knowledge" is frozen until a date; The physician confirms the current status.

Mini case 3: Gradual integration

Rather than pushing AI everywhere at once, one clinic is making a gradual plan: first low-risk administrative tasks (appointment messaging), then clinical documentation (dictation + verification), then careful radiography pre-screening — all while maintaining physician approval. At each stage, the team is trained, errors are recorded and the process is improved. In six months, both productivity increases and security culture is established. Lesson: safe integration comes gradually, measuredly and with training.

AI integration into the clinic: roadmap

1. Evaluate: Which tasks are in the green/yellow/red zone? (Unit 1)
2. Start (green): Start with low-risk administrative tasks and see quick gains.
3. Expand (yellow): Proceed with physician verification of documentation and communication.
4. Apply with caution (red): Confirm radiography/planning support as second eye only.
5. Educate: Teach the team about privacy, verification, and ethics.
6. Monitor and improve: Record errors, review processes regularly.

Tip: Keep a personal “AI learning diary”: what task did you speed up with AI, what bug did you catch, what rule did you update each month? This journal nourishes both your development and your safety culture.

Copiable templates

Role: Source verification checker. Task: List the following AI citations as "claims that need to be verified". For each, specify what information (author, journal, year, DOI) I should check against the original. Don't assume any of it is "correct". Attributions: [paste]

Role: Article summarizer (cautious).Task: Summarize the following article in plain language: purpose, method, main finding, limitations.Do not add any conclusions that are NOT in the text; Write "[unclear in text]" where you are not sure.Article text: [paste]

Role: Level of evidence classifier. Task: Generate the possible level of evidence (review, clinical study, case, expert opinion) for the following information/claim and the questions I should ask my patient for applicability. DO NOT make a definitive clinical decision. Claim: [write]

Role: AI integration planner.Task: Draft a phased AI adoption roadmap for my clinic: green/yellow/redwhat tasks in order, physician approval and team training points at each stage, tracking metrics.Clinic profile: [size/specialty]

Weak prompt / Strong prompt

Weak: "Give me the most up-to-date evidence and resources on this topic and I'll tell my patients."

Why it's weak: Assumes AI will provide up-to-date and accurate resources; It ignores the risk of hallucinations and actuality deficits.

Strong: "Draw a general framework on this issue and list the key claims I need to verify. I will verify each source you give from the original and current guides; indicate where you are unsure."

Why it's powerful: Positions AI as a framework/initiator, leaves validation to the physician, makes uncertainty visible.

Critical literacy: The most valuable skill in the age of AI

AI tools will change, develop, new ones will come; But what will not change is the physician's critical thinking skills. The most valuable professional competence in the age of AI is the ability to evaluate whether an output is true or false, in which case it is reliable and in which case it is questionable. This skill is much more permanent than the tool itself. If a physician understands at a basic level how AI works (knows that it recognizes patterns, does not produce evidence, has limited timeliness, can hallucinate), he can position it correctly, no matter what tool comes along.

So learning AI is not about memorizing the buttons of individual tools, but about grasping the principles: use according to risk level, multi-layer authentication, privacy, transparency, and ultimate responsibility with the physician. These principles are the backbone of this module and can be applied to any new technology. The best way to prepare for the future is not to become an expert on a particular tool; is to develop a mental discipline that questions and verifies, based on these principles.

Mini case 4: The tool has changed, the principle remains

A clinic replaces its radiography AI tool with a newer version after one year. Although the interface and features are different, the team seamlessly applies the same verification discipline (physician reading first, comparison later, clinical correlation, physician approval on critical decision) to the new tool. The transition was painless because the team was committed to the principle, not the tool. Lesson: tools are temporary, principles of verification and accountability are permanent; Preparation for the future is to internalize these principles.

Common mistakes

• Using AI-generated citations without verifying them from the original.
• Assuming the AI's knowledge is up to date.
• Taking all information with equal weight without questioning the level of evidence.
• Introducing AI into the entire clinic at once, without training.
• Not recording errors and not improving processes.

In summary

AI accelerates continuing professional development and evidence screening, but does not verify evidence. Hallucination, topicality deficit, and loss of context are real risks; Every reference and claim must be verified from an original, current and reliable source. Always question the level of evidence and applicability to the patient. Integrate AI into the clinic gradually, from green to red, with training and maintaining physician approval; record errors and continuously improve processes. The unchanging principle applies throughout this module: the decision to diagnose and treat belongs to the human, competent dentist; AI output does not replace this responsibility.

Application task

Write a 6-month AI integration and learning roadmap for yourself: which task (green/yellow/red) you will deploy in which month, which validation and training step you will implement at each stage. Additionally, request a summary from AI on a topic and confirm all the citations it gives with the "source verification" prompt; Report how many are real and how many are fabricated.

checklist

• [ ] I verified every attribution given by AI from the original.
• [ ] I checked the up-to-dateness of the information with professional organization guidelines.
• [ ] I questioned the level of evidence and applicability to the patient.
• [ ] I established a plan to integrate AI into the clinic gradually and with training.
• [ ] I got into the habit of recording errors and improving processes.

Module Exam

1. A dentist is evaluating a suspicious periapical lesion marked by the AI scanning tool on a panoramic radiograph. Which is the most correct approach?

• A) Evaluate the AI finding along with clinical examination, patient history and additional images when necessary and give the final comment as a physician ✔
• B) Since AI marks the lesion, start treatment directly
• C) Skipping the clinical examination if the AI did not flag anything
• D) If the AI confidence score is high, write a report without examining the patient

Description: Radiographic diagnosis is a safety-critical decision. AI is useful as a preliminary assessment and second set of eyes; However, the final interpretation should be given by the qualified dentist along with clinical examination, patient history and additional images when necessary. AI output does not replace physician approval.

2. What is the most significant clinical risk of a 'false negative' AI radiograph print in dentistry?

• A) The risk of not marking an actually existing pathology and of the physician relying on it and overlooking it ✔
• B) Incorrect marking of a healthy tooth and unnecessary treatment
• C) The image has high resolution
• D) Faster production of the report

Explanation: A false negative is when an actually existing pathology (e.g. a bruise or lesion) is not flagged by the AI. The biggest risk is that the physician will rely too much on the AI ​​and miss overlooked pathology; so just because the AI ​​doesn't flag anything doesn't mean 'clear', clinical examination is a must.

3. What information must be verified manually when structuring patient history and examination notes with AI?

• A) Format of the appointment time
• B) Address of the clinic
• C) Font of the note
• D) Critical medical information such as allergies, medication use, bleeding disorders and systemic diseases ✔

Description: Critical medical information such as allergies, medications used, bleeding disorders and systemic disease directly affects the safety of treatment. Since voice-text and summarization tools can make errors, this information must be manually verified by the physician.

4. What is the proper use of an AI treatment plan outline?

• A) Presenting the draft as a definitive plan to be applied directly to the patient
• B) Using the draft as a starting point that the physician reviews and decides on according to indications, contraindications and patient preference ✔
• C) Adding the draft to the patient file without reading it at all
• D) Accept it as correct if the AI gave the same plan more than once

Description: AI can quickly outline options and stages; However, the diagnosis, indication and final treatment decision belong to the competent physician based on clinical examination and evidence. The outline is a starting point that the physician reviews and approves.

5. What is the most critical control when preparing the informed consent text with AI?

• A) The text should be colorful and visual
• B) Keep the text as short as possible
• C) Full coverage of treatment-specific risks, alternatives and legal obligations with physician approval ✔
• D) The text has been translated from English

Explanation: The consent text is a legal and ethical document. AI can speed up the draft, but treatment-specific risks, alternatives, and legal obligations must be checked and completed by the physician. Incomplete risk information or exaggerated promises are unacceptable.

6. What limits should be maintained when sending automated patient messages with AI in clinical management?

• A) Adding the patient's entire treatment history to each message
• B) Sending messages from public social media
• C) Giving medical advice via automatic message
• D) Messages do not contain personalized diagnosis and medical advice and patient privacy is protected ✔

Explanation: Automatic reminder and notification messages are for administrative purposes; It should not contain personalized medical advice or diagnosis. In addition, patient privacy should be protected in messages, and sensitive health information should not be shared on channels visible to third parties.

7. What is the correct approach for an AI-recommended crown design in a CAD/CAM workflow?

• A) Sending the proposal to production without reviewing it
• B) Cementing the patient without rehearsal because AI recommends it
• C) Waiting for production without sending any notes to the laboratory
• D) Check and confirm occlusion, marginal harmony and aesthetics as a physician, and correct if necessary ✔

Description: AI can quickly suggest marginal lines and morphology; However, occlusion, marginal fit, contact points and aesthetics must ultimately be checked and approved by the physician. Design automation does not replace physician supervision.

8. What is the best approach to planning an implant using AI-assisted segmentation and measurement on the CBCT image?

• A) Verify AI measurements with anatomical safety distances and clinical judgment and approve the plan as a physician ✔
• B) Going to surgery based on the AI measurement without ever checking the measurements
• C) Not examining the anatomy because the segmentation is automatic.
• D) Presenting the AI plan to the patient as the only unchangeable option

Description: Implant planning is safety-critical; Distances to the mandibular canal, sinus and adjacent roots determine surgical safety. AI speeds up measurements, but anatomical safety distances and the final plan must be verified by a qualified physician.

9. What is true in terms of clinical advertising and ethics when producing patient education content with AI?

• A) Attracting patients with promises such as 'guaranteed results' and 'painless definitive treatment'
• B) Keep content accurate, evidence-based and balanced; Avoid exaggerated promises and publish with physician approval ✔
• C) Making comparisons that denigrate the rival clinic.
• D) Using before and after treatment images without permission and in a misleading manner

Statement: Exaggerated promises, guaranteed results and misleading comparisons in health content are prohibited and are against professional ethics. Content must be accurate, evidence-based and balanced; It should be checked by the physician for currentness and accuracy.

10. What is the most important criterion in terms of KVKK when choosing an AI tool that works with patient health data?

• A) The tool is the most popular and free
• B) The interface of the vehicle is beautiful
• C) KVKK assurances such as data processing conditions, storage location, international transfer and data processing agreement ✔
• D) The vehicle has the most features

Description: Health data is special personal data and requires high protection. The data processing conditions of the vehicle, where the data is stored, international transfer, data processing agreement and security measures are critical criteria. Just because it's free doesn't mean it's safe.

11. What is the primary way to maintain privacy when uploading patient radiography or information to an AI tool?

• A) To get better results by adding the patient name and TR ID number
• B) Remove and anonymize identifying information and use only KVKK compliant, contracted tools ✔
• C) Uploading the entire patient file to any free tool
• D) Upload the data to a public forum and ask for comments.

Explanation: Descriptive patient information (name, ID, date of birth, contact) should not be uploaded to tools that are not secure or do not have a data processing agreement. If possible, the context is anonymized; Health data is processed only in KVKK compliant, contracted and secure systems.

12. What should be done against the risk of fabricated sources (hallucinations) when you ask AI for a literature summary on a clinical topic?

• A) Verify each reference given from the original source and use only confirmed information ✔
• B) Using sources without verifying because AI writes with confidence
• C) Presenting the resource list to the patient as it is
• D) Accepting the article as correct without ever opening it if it has a DOI number

Description: General-purpose language models can invent articles, authors, or DOIs that do not exist. Each citation given must be verified from the original source; Clinical decision should be based only on validated, current and evidence-based literature.

13. Who is responsible for the diagnosis and treatment decisions in the use of AI in dentistry?

• A) To the company that developed the software
• B) The AI model itself
• C) To a competent dentist who evaluates with clinical examination and evidence ✔
• D) To the secretary who made the appointment

Description: AI is a decision support and efficiency tool. The diagnosis, indication and treatment decision belongs to the competent dentist based on clinical examination, patient history and evidence. Legal and ethical responsibility lies with the physician; AI output does not absolve this responsibility.

14. Which of the following is the most common technical problem that impairs the reliability of AI output in panoramic radiography?

• A) The image is in digital format
• B) Incorrect positioning, motion blur, artifact and poor image quality ✔
• C) The image is in color
• D) The radiography is archived

Description: Incorrect patient positioning, motion blur, ghost shadows and poor image quality cause AI to incorrectly mark. In case of a poor quality image, the AI ​​output should not be trusted; the image should be taken again or interpreted together with the clinical examination.