flowchart LR
accTitle: Two ways a tool arrives at a citation
accDescr: Two flowcharts side by side. Answering without search, your question leads to generate answer text, then generate citation text, then something that looks like a reference. Search-first tool, your question leads to search PubMed or an index, then real papers, then generate an answer about those papers, then cites what it found.
subgraph G["Answering without search"]
direction TB
q1[Your question] --> gen1[Generate answer text]
gen1 --> cite1[Generate citation text]
cite1 --> out1[Looks like a reference]
end
subgraph R["Search-first tool"]
direction TB
q2[Your question] --> search[Search PubMed or an index]
search --> docs[Real papers]
docs --> gen2[Generate answer about those papers]
gen2 --> out2[Cites what it found]
end
4 Critical appraisal: checking AI tools, their answers, and the evidence about both
About 70 minutes
TipIn short
- The idea. Asked as a clinician, with web search on, a chatbot such as ChatGPT Edu usually gives you real references, and a minute in PubMed per reference tells you which ones to trust. The same habit works on a vendor’s slide: ask where each number came from, and what kind of study produced it. Cleared by the FDA is not the same as tested on patients like yours.
- What you do. Ask ChatGPT Edu, with web search on, for three references on a question in your specialty and check each one in PubMed. Then sort ten claims about AI scribes by the kind of study behind them.
- What you keep. A routine for checking any reference in a few minutes, and six questions to ask about any AI tool before your program buys it.
- Short on time? Read “Three things that can be wrong” and “But it’s FDA-cleared, isn’t it fine?”, then do Part A (25 minutes).
NoteWhat you’ll learn
By the end of this chapter you will be able to:
- Ask a chatbot with web search for references, check each one against PubMed, and score it on a simple accuracy scale, instead of judging by whether it looks real.
- Explain why made-up references follow from how these systems write text, and why a tool that searches real papers first is a different kind of tool, not just a better one.
- Describe how the wording of a prompt changed the rate of made-up references in one study, and why one test with one kind of wording is not enough.
- Sort claims about a clinical AI tool into randomized, observational, and vendor or trade-press evidence, and explain why the three tiers disagree in the same direction.
- Explain why FDA clearance is not evidence of clinical performance, using a published analysis of the devices the FDA has authorized.
- Tell a wrong reference from a wrong claim from a wrong attribution, and say which of the three a search-first tool actually fixes.
- Ask six questions about a tool you have never seen, and say why each one is the question to ask.
Course competencies: AIM-2 (critical appraisal), AIM-1 (foundations), AIM-4 (ethics, law, and regulation), AIM-9 (research and scholarship).
What you need. Your university ChatGPT Edu or Copilot account, and a PubMed tab. Nothing to install. Made-up patients only. Your Edu account is created after a short training course in Canvas. Until then, use Copilot with your university login, not a personal account. This is a menu chapter: read it if critical appraisal interests you.
4.1 Lunch with a vendor
It is noon conference in your second month of intern year. The sandwiches are better than usual, because a company is paying for them. Their representative is showing a slide about an ambient scribe: a phone app that listens to the visit and writes the note. The slide says documentation time fell 40 percent and burnout fell by a third at a large health system. Your program director is nodding. The chief resident leans over and asks what you think, because you are the intern, and interns get asked.
Look at the slide for a moment before reading on. Every number on it is the kind of number you will be shown. Nothing on it tells you where any of them came from.
Earlier that morning, a senior resident handed you a reference for the case presentation you give on Thursday. She had pulled it from ChatGPT. It has a plausible first author, a real journal, a well-formed DOI, and a finding that fits your case nicely. You have not opened it.
These are the same problem at two scales. In both cases someone has handed you a claim that looks right. You are the person whose name will be on whatever that claim supports. The slide and the reference need the same question, which this chapter will ask twice: what would have to be true for this to be trustworthy, and did anyone check?
Clinicians have a word for that habit when it is applied to a paper. They call it critical appraisal. This chapter extends it to two things a paper never used to do: write its own references, and get sold to your department over lunch.
4.2 Why this matters
You will use AI tools to find evidence within the year, and most of what a good one gives you will be right. In a 2026 study, ChatGPT with web search, asked as a clinician, made up about 2 in 100 of its references. Its commonest error was a real paper with a wrong PMID or DOI (McLaughlin et al., 2026). A check in PubMed takes about a minute a reference and tells you which ones to trust.
Your program will meet the vendor problem, and so will you. Residents sit on the committees that pilot these tools. The vendor’s slide, the health system’s press release, and the randomized trial will say different things about the same product. Deciding which to believe is a skill you will still need in fifteen years, when every product named in this chapter is gone.
4.3 How it works
4.3.1 Two skills that get treated as one
Appraising an AI tool is two separate skills. The first is checking what a model just told you: does the resident’s citation exist, and does it say what the model claims? The second is checking what the literature says about a type of tool: does the scribe on the lunch slide actually give you your evenings back?
Most people arrive assuming the first skill is the hard one and the second is someone else’s job. This chapter reverses that. Checking three references takes about ten minutes and a PubMed tab, and the answer is clear. Weighing a large study against a health system’s press release takes judgment.
The two halves connect through the question from the lunch room, asked twice. In the first half, you are the one who checks. In the second, you grade how well other people checked.
4.3.2 Where a citation comes from
A language model that answers without searching produces a citation the same way it produces any other text: by writing the most likely next words, one token at a time. A token is a word or piece of a word. The model is doing, at enormous scale, what your phone’s keyboard does when it suggests the next word. A well-formed author list, a real journal name, and a DOI in the right format are what likely next words look like. The model is not looking in a database and failing. It never had one.
A search-first tool does this the other way round. It searches a database of real documents first, then writes about what it found. ChatGPT Edu and Copilot work this way when they search the web, and the answer then shows its sources. Some medical products, such as OpenEvidence, always search first. The retrieval chapter shows how to tell whether a given tool really does. Figure 4.2 shows the two paths.
This is why a search-first tool’s rate of made-up references is different in kind, not just a bit lower. It can still misread a paper it found. It cannot invent one it did not find. A chatbot that searches only some of the time can mix the two paths in one answer, which is why you still check.
4.3.3 Three things that can be wrong
A citation that exists is not the end of the check. Three things can be wrong, and each can fail on its own:
| What can be wrong | The question | Who fixes it |
|---|---|---|
| The reference | Does this paper exist? | A search-first tool, almost completely |
| The attribution | Does this paper say what the tool says it says? | A search-first tool, partly; it can still misread what it found |
| The claim | Is the answer clinically right? | Nobody but you |
The third row is the one that matters on the wards. When 33 physicians graded chatbot answers to 284 of their own questions, the median answer was almost completely correct, but 36 answers were rated mostly or completely wrong (Goodman et al., 2023).
4.3.4 What the published rates look like
Three studies show how far the rates vary.
| Study | Setting | Result |
|---|---|---|
| Gravel et al. (2023) | 20 medical questions, ChatGPT | 41 of 59 references fabricated (69%) |
| McGowan et al. (2023) | Psychiatry, ChatGPT | 2 of 35 citations real |
| McLaughlin et al. (2026) | Spine surgery, five tools | OpenEvidence 0 fabrications; Claude 78.0% accurate; Gemini 26.2% fabricated |
The same 2026 study contains the finding this chapter depends on most. The fabrication rate for ChatGPT rose from 2.1% when it was asked as a spine surgeon, in clinical words, to 20.0% when it was asked as a patient, in everyday words (McLaughlin et al., 2026). The study changed who was asking and the wording together, on ten matching questions. So the finding is about the whole prompt: whether it sounds like a doctor or like a patient.
ImportantThe practical consequence
The version of a tool you trust at work may behave differently when you, or your patient, ask the same question in everyday words. Testing a tool once, with one kind of wording, is not enough.
The study did not test people writing in a second language. The habit that follows is the same for everyone: say who you are, and ask the way you would present on rounds.
There is an equity finding in this number, and the authors point to it. The people least able to check an answer, patients asking in their own words, got the answers most likely to cite something that does not exist.
NoteThese numbers move
Fabrication rates depend on the tool and the month, not on the technology. A second 2026 study, on questions about facial cosmetic surgery, reported the GPT-versus-Gemini ordering reversed: 36% of GPT-4’s citations could not be verified, against 14% for Gemini and 8.8% for DeepSeek (Frutuoso Maia et al., 2026). When your own results below disagree with Table 4.2, that is expected. How the tool works, shown in Figure 4.2, is the durable part. The percentages are not.
4.3.5 Grading the evidence about a tool
The second skill applies the same question to the literature about a type of tool. Ambient documentation, or the “AI scribe”, is the right example. The evidence is large and recent, and it disagrees with itself in a consistent direction.
Claims about scribes come in three tiers:
- Randomized trials. Clinicians or clinics assigned by chance to the tool or to a comparison.
- Observational studies. Real-world data from clinicians, large or small, without randomization. This includes pre/post designs, which measure the same clinicians before and after they start using the tool. It also includes sites that adopt the tool one after another without randomization, and surveys of people who chose to use it.
- Vendor or trade press. A claim that comes from a company, a press release, or industry news (the “trade press”), however specific the number sounds.
Here is what one study looks like when each of the three tiers reports it to you. All three headlines are about the same paper, the five-system study cited below.
EurekAlert! · press release
AI scribes linked to modest reductions in electronic health record use and clinical documentation time
April 2026
Same paper, three versions: the institution’s press office, a journalist, and a professional society. None of them is wrong. Which one would the vendor at lunch have put on the slide?
The three tiers disagree in the same direction. Pre/post studies of health-system rollouts report large drops in burnout. The randomized trials mostly report smaller gains, and the one with a large drop was very small. The trials that measured after-hours charting found no change that held up, and that is the outcome that justifies the whole product category. In the largest study to date, across five academic systems, adopting a scribe was linked to modest drops in EHR and documentation time and slightly more visits a week, but not to less EHR time after hours (Rotenstein et al., 2026). You will work out why in the exercise, before reading the explanation.
The lesson to keep: the published number is usually the optimistic one. The extra visits may show another: time saved tends to get filled with more work. The workforce chapter looks at where saved time goes.
4.3.6 “But it’s FDA-cleared, isn’t it fine?”
Clearance answers a different question from the one a clinician is asking. Most AI devices the FDA has authorized are radiology tools, so if you plan to go into radiology this section is about your daily work. An analysis of the 903 AI-enabled devices the FDA had authorized by August 2024 found a clinical performance study reported for about half of them (56 percent); a quarter (24 percent) said none was done. Fewer than a third of the clinical studies (29 percent) reported results for women and men separately (Windecker et al., 2025). A test score standing in for performance on patients is a pattern the bias and equity chapter explains in full.
ImportantCleared is not validated
Clearance says the device is similar enough to one already on the market (the FDA’s term is “substantial equivalence”) and that the manufacturer has a quality system. It does not certify that a device improves outcomes, that it was tested on patients like yours, or, often, that clinical performance was measured at all. If you remember one sentence from this chapter, remember this one, and the paper with it.
4.3.7 What a trustworthy tool looks like
Everything so far grades evidence that already exists. The tool at lunch is new, and there may be no literature yet. You still need something to ask.
The most useful checklist is FUTURE-AI, an international consensus guideline published in the BMJ in 2025. It reduces trustworthy clinical AI to six principles (Lekadir et al., 2025). The six principles, and the question each one becomes when a vendor is standing in front of you, are in Table 4.3.
| Principle | What it means | The question for the vendor |
|---|---|---|
| Fairness | Performance does not depend on who the patient is | Was it tested on patients like mine, and did performance differ by sex, race, language, or age? |
| Universality | It works outside the place that built it | Has it been validated at a site that is not yours, on a different EHR? |
| Traceability | You can see where it came from and where an output came from | What was it trained on, and can I trace this note, or this recommendation, back to its source? |
| Usability | The people who use it were involved, and kept using it | Who designed how it fits into the clinic day, and what fraction of clinicians offered it were still using it at six months? |
| Robustness | Its performance drops a little at a time, not all at once | What happens with a bad microphone, an interpreter in the room, a rare presentation, an unusual chart? |
| Explainability | It can say why, in a form you could defend | If this is wrong in a chart review, what will I be able to show about why I trusted it? |
Two things to notice. Traceability asks where things came from, and that is what this chapter’s first half is about. A tool that shows you its sources next to its answer has answered it. A tool that writes a reference list has not. And none of the six asks whether the tool is accurate on a benchmark, a fixed test set used to compare tools. Accuracy is necessary, and it appears under robustness. It is not the checklist, because a tool can be accurate on the population it was built on and fail every other row.
There is a seventh question the six do not ask. It belongs to the patient in the room rather than the clinician: does the patient know the visit is being recorded, and did they agree? No tier of evidence in Part B answers it, and no vendor slide will. Ask it anyway.
4.4 Watch or listen (optional)
Podcast. NEJM AI Grand Rounds, “The OpenEvidence Episode: Dr. Travis Zack on the Future of Clinical Evidence” (May 20, 2026; 66 min). https://ai-podcast.nejm.org/e/the-openevidence-episode-dr-travis-zack-on-the-future-of-clinical-evidence/
Zack is OpenEvidence’s chief medical officer, so this is the search-first side of Figure 4.2 described by someone who builds it. Listen for how the tool decides what counts as evidence, and for what he says trust depends on. Then ask the question this chapter asks of everything: what would have to be true for that to hold, and who has checked? The first thirty minutes are enough.
Video alternative. “Travis Zack on OpenEvidence and the Future of Medical AI”, Stanford Department of Medicine, YouTube (July 2026; 44 min). https://www.youtube.com/watch?v=nrfGSE__7po
4.5 Do
Made-up patients only. What counts as patient information.
TipPubMed is the verification tool
Use PubMed and nothing else. The exercise depends on searching an index that without question contains the real literature. Google Scholar’s loose matching will find you a different paper for a made-up citation, and you will score it as real.
4.5.1 Part A: ask, then check (about 25 minutes)
Step 1. Pick a question in your specialty (2 minutes). You want a management question with a real literature. If you have no specialty yet, use anticoagulant choice in atrial fibrillation with stage 4 chronic kidney disease.
Step 2. Ask, with web search on (3 minutes). Use ChatGPT Edu, in a Temporary Chat or a chat outside your course project. If you see a web search option beside the message box, turn it on; the prompt also asks for a search. The DOI is the publisher’s permanent link for a paper, and the PMID is its PubMed ID number.
I am a physician managing
<your clinical question>. Search the web and PubMed, summarize the current evidence in five sentences, and give me three references from the peer-reviewed literature with authors, journal, year, and DOI or PMID. Show the source you used for each.
Step 3. Check each reference (10 minutes). Search PubMed by title first, then by first author plus year, then by DOI or PMID. Title first, because a real paper with a wrong PMID or DOI is a common error. Give each reference one score from Table 4.4, and note whether the answer showed a source for it.
Expect most of them to check out. That is the routine working: ask, then check. Most of a good answer is right; your job is to find out which parts.
| Score | Meaning | How you decide |
|---|---|---|
| 2, accurate | The reference exists as given and supports the claim the model attached to it. | Found in PubMed with matching authors, journal, and year, and the abstract is consistent with what the model said it showed. |
| 1, misrepresented | The reference exists, but the model’s description of it is wrong. | Found in PubMed, but it is a different design, population, or conclusion, or the model attributed a finding it does not contain. |
| 0, fabricated | The reference does not exist as given. | Not found by title, author plus year, or DOI/PMID after a genuine search. A real DOI pointing at a different paper than the one named scores 0. |
TipThree examples, one of each
The clinician prompt was run on anticoagulation in atrial fibrillation with stage 4 chronic kidney disease, in a general-purpose model without search (Claude, 2026-09-07), and every reference was checked in PubMed. All five were real, even without search. Expect most of yours to be real too. The three examples below show what each score looks like. The first is from that run. The second and third are made up to show the pattern, and are labeled as such.
Score 2, accurate. Stanifer JW et al. Apixaban versus warfarin in patients with atrial fibrillation and advanced chronic kidney disease. Circulation 2020;141:1384–1392. The tool said it showed less bleeding with apixaban in creatinine clearance 25 to 30. PubMed: exists, authors and journal match, and the abstract reports major bleeding hazard ratio 0.34 in exactly that subgroup (Stanifer et al., 2020). Score 2.
Score 1, misrepresented (made up). Pokorney SD et al. Apixaban for patients with atrial fibrillation on hemodialysis. Circulation 2022;146:1735–1745, described by the tool as “a randomized trial showing apixaban is safer than warfarin in dialysis patients.” PubMed: the paper exists (Pokorney et al., 2022). Its own conclusion is that it stopped early and had inadequate power to draw any conclusion about bleeding. The reference is real. The sentence attached to it is the opposite of what the authors wrote. Score 1. This is the failure a search-first tool can still make, and the one only reading the abstract catches.
Score 0, fabricated (made up). Hernandez AV et al. Direct oral anticoagulants in stage 4–5 CKD: a randomized comparison. J Am Coll Cardiol 2021;77(9):1123–1131. doi:10.1016/j.jacc.2020.12.041. Every feature is plausible: a real journal, a common surname, a year in range, a DOI with the journal’s correct prefix. Search PubMed by title: nothing. By author and year: the real Hernandez AV publishes on other topics. By DOI: it resolves to a different paper or to nothing. Score 0. What reveals it is not any single feature. It is that no feature can be found.
Step 4. The contrast (8 minutes). Open another Temporary Chat, or another new chat outside your project. Ask the same question the way a patient would, without asking it to search:
I have
<the condition, in plain words>and my doctor mentioned<the treatment>. Can you explain what the research says, and give me three references from medical journals with authors, journal, year, and DOI or PMID so I can look them up myself?
Check these three the same way. It may get them right too; if so, how would you know? Look at whether each answer showed its sources. A reference with a source you can open is one the tool found. A reference with no source shown may have been written as text, so check those first.
Step 5. Record it (2 minutes). Copy this into your document.
Tool: ____________________ Model version if shown: ____________
Clinical question: ______________________________________________
Asked as Ref Title as given by the model Did it show In PubMed? Score
sources?
Clinician 1 ____________________________ Y / N Y / N ___
Clinician 2 ____________________________ Y / N Y / N ___
Clinician 3 ____________________________ Y / N Y / N ___
Patient 1 ____________________________ Y / N Y / N ___
Patient 2 ____________________________ Y / N Y / N ___
Patient 3 ____________________________ Y / N Y / N ___
For any reference scored 1: in one sentence, what did the model claim the
paper showed, and what does it actually show?
For any reference scored 0: what about it looked credible? Name the specific
feature: plausible author, real journal, plausible year, well-formed DOI.If every reference scored 2, write down how you know, in one line: which search found each one. That line is the routine you keep. Six references are far too few to repeat the 2.1 versus 20.0 percent finding; you are trying the study’s design, not testing its result.
4.5.2 Part B: grading the evidence (20 minutes)
Below are ten claims about ambient AI scribes, with their sources removed. Sort each into a tier (randomized trial, observational, or vendor and trade press), rate your confidence from 1 to 5, and give a one-line reason. Then answer the two closing questions before you open the solution.
- Across 3,442 physicians and 303,266 encounters at one integrated health system, physicians reported improved experience with an ambient documentation tool.
- Burnout among tool users fell from 51.9% to 38.8%, odds ratio 0.26.
- Burnout dropped 21.2 points, and 60% of participants said the tool made them more likely to extend their clinical career.
- Documentation time fell by 6.89 minutes per day and after-hours time by 5.17 minutes per day.
- Burnout and task load improved, but clinicians rated both tools as producing occasional clinically significant inaccuracies.
- Work exhaustion fell by 0.44, but the reduction in work done outside work hours lost statistical significance after the most extreme values were removed.
- After-hours “pajama time” was largely unaffected by either of two scribe tools.
- Burnout fell from 89% to 56%, with no significant change in pajama time or charting time.
- Across five academic systems and 8,581 clinicians, total EHR time fell 13.4 minutes per day and documentation time 16.0 minutes per day, with 0.49 more visits per week, but after-hours EHR time did not change significantly.
- Seventy-nine percent of clinicians offered the tool declined to use it.
Claim Tier (RCT / Obs / Vendor-trade) Confidence (1 to 5) One-line reason
1 ________________________________ ___ ________________
2 ________________________________ ___ ________________
3 ________________________________ ___ ________________
4 ________________________________ ___ ________________
5 ________________________________ ___ ________________
6 ________________________________ ___ ________________
7 ________________________________ ___ ________________
8 ________________________________ ___ ________________
9 ________________________________ ___ ________________
10 ________________________________ ___ ________________
Closing question 1. The pre/post health-system studies look substantially
better than the randomized trials. Name two mechanisms that would produce
that pattern even if the tool worked exactly as well as the RCTs say.
(1) ______________________________________________________________
(2) ______________________________________________________________
Closing question 2. Which outcome do the pre/post studies and the RCTs
disagree about most, and why is that the outcome you personally care about?Claim 10 is designed not to fit the scheme. Whichever tier you put it in, be able to defend the choice.
TipSolution to the closing questions
Two mechanisms: selection and Hawthorne. The pre/post studies measure clinicians who chose to adopt an optional tool: people already inclined to like it, whose way of working it fits. The randomized trials measure clinicians who were assigned it. The five-system study shows how much choice is involved: 1,809 of the 8,581 clinicians it studied adopted the scribe, and at four of the five sites adoption was each clinician’s own choice (Rotenstein et al., 2026). Pre/post studies sample people like those adopters. Add that people who know they are in a documentation study document differently, the Hawthorne effect, and the direction of the gap is explained without anyone having lied.
The outcome they disagree about is after-hours charting. It justifies the entire product category, and it is the outcome the randomized evidence most consistently fails to move. One plausible reading, offered as a hypothesis and not a finding, is that time saved goes into more visits (the 0.49 additional visits per week) rather than coming back as personal time. A similar pattern: AI-drafted replies to patient inbox messages produced no significant change in reply time but significant reductions in task load and work exhaustion (Garcia et al., 2024). Less burden and more free hours are different outcomes. Treating them as the same is how a real benefit gets oversold into a false one.
Claim 10 is a number from journalism about a peer-reviewed study, not from the paper’s own results. The paper reports 1,809 adopters among the 8,581 clinicians it studied; it does not say how many were offered the tool and said no. You will meet this category constantly. It is neither peer-reviewed evidence nor vendor marketing, and it needs its own habit: find the paper, and see whether the number is in it.
Two independent checks on the vendor story. The Peterson Health Technology Institute, which takes no vendor funding, concluded in 2025 that scribes likely reduce burnout but that their financial impact is unclear. In 2026 it concluded that administrative AI reduces burden within organizations without lowering system-wide costs (2025 report, 2026 report). A UCSF analysis found scribe adopters billed about 1.8 more relative value units a week, the billing measure of clinical work, with no rise in claim denials (Holmgren et al., 2026). The study was one health system and adopters chose to adopt, so the authors call for a randomized test.
The tiers, so you can check your answers: claims 1, 2, 3, 4, and 9 are observational (claim 4 is a pilot of 45 physicians); 5, 6, 7, and 8 are randomized trials (claim 8 had 23 clinicians); 10 is trade press about a peer-reviewed paper.
4.6 Check
Score yourself against this before moving on.
| You have | Meets | Falls short |
|---|---|---|
| Six scored references | Every score is backed by a PubMed search, not a judgment of plausibility, and you noted whether each answer showed its sources | “Looked real” or “looked fake” without a search |
| A reason a fabricated reference looked credible, if you found one | Names a specific feature | “It just seemed legit” |
| A reason a misrepresented reference was wrong, if you found one | States what the model claimed and what the paper shows | Restates the model’s summary |
| Ten tiered claims | Each tier has a one-line reason; claim 10 is defended either way | Tiers assigned without reasons |
| Two mechanisms | Selection and observation effects, or equivalents, in your own words | “The pre/post studies were biased” |
| The one sentence | You can say “cleared is not validated” and name the paper | Only the slogan |
| The mechanism | You can explain why a generated reference may not exist without using the word “hallucination” | “It hallucinates” |
| The vendor question | You can name the FUTURE-AI principle your question tests, and why that one first | A question with no principle behind it |
4.7 What this changes for you
What does this change about how you’ll practice? Write it down before you close this chapter. Two specifics. First, what will you now check before you cite a reference an AI tool gave you in a note, a presentation, or an application, and which of the three failures in Table 4.1 does that check catch? Second, go back to the lunch room: the chief resident is still waiting. What do you say about the 40 percent, and which one of the questions in Table 4.3, or the seventh about the patient, do you ask the vendor first? One word is not an answer to either.
4.8 Go deeper
Papers
- A published classroom framework built around AI citation errors, in biology but usable in medicine (Shah & Miranda, 2026).
- The randomized trial of physicians with model access (Goh et al., 2024), and the study of 148 students judging GPT-3.5’s answers to ten cases, five of which the model had got wrong (Waldock et al., 2025).
- FUTURE-AI in full, with the 30 practices behind the six principles (Lekadir et al., 2025); and TRIPOD-LLM, the reporting guideline a study of a language model should meet (Gallifant et al., 2025), which is the standard most of the Part B claims fail.
- The National Academy of Medicine’s AI Code of Conduct, in its discussion draft (Adams et al., 2024); the 2025 final is a NAM publication and the principles are the same. Also in the ethics chapter.
- Eight guiding principles for AI tools in health care (Badal et al., 2023). “Report clinically meaningful outcomes” and “have high healthcare value” are what Part B teaches in exercise form.
- The first large health-system report on ambient scribes (Tierney et al., 2024), read alongside the five-system study (Rotenstein et al., 2026).
Listen
- NEJM AI Grand Rounds, “What Values are in AI? A Conversation with Dr. Zak Kohane” (17 December 2025; 78 min). https://ai-podcast.nejm.org/e/what-values-are-in-ai-a-conversation-with-dr-zak-kohane/ Listen for the remark that ambient documentation spread because it improves clinician satisfaction, not accuracy or throughput. Compare that with the evidence tiers in Part B.
Talks
- Grounding AI in real sources is a written article, not a deck, and builds the search-first pipeline from first principles. Read it if Figure 4.2 felt like an assertion. It is the retrieval chapter’s reading alternative.
- Prompt engineering: ten rules with clinical examples, including the “precise language” rule that the clinician-versus-patient wording effect illustrates. The prompting lab builds on this deck.
- AI in medical education covers attribution and academic integrity, for the moment you use one of these tools in a research project.
Adams, L., Fontaine, E., Lin, S., Crowell, T., Chung, V. C. H., & Gonzalez, A. A. (2024). Artificial Intelligence in Health, Health Care, and Biomedical Science: An AI Code of Conduct Principles and Commitments Discussion Draft. NAM Perspectives, 2024. https://doi.org/10.31478/202403a
Badal, K., Lee, C. M., & Esserman, L. J. (2023). Guiding principles for the responsible development of artificial intelligence tools for healthcare. Communications Medicine, 3(1). https://doi.org/10.1038/s43856-023-00279-9
Frutuoso Maia, B. K., Morais, E. F. de, Santana Santos, T. de, & Charles Pagotto, L. E. (2026). Performance of large language models in preoperative and postoperative counselling for aesthetic facial procedures. The British Journal of Oral & Maxillofacial Surgery, 64(3), 216–222. https://doi.org/10.1016/j.bjoms.2026.01.002
Gallifant, J., Afshar, M., Ameen, S., Aphinyanaphongs, Y., Chen, S., Cacciamani, G., Demner-Fushman, D., Dligach, D., Daneshjou, R., Fernandes, C., Hansen, L. H., Landman, A., Lehmann, L., McCoy, L. G., Miller, T., Moreno, A., Munch, N., Restrepo, D., Savova, G., … Bitterman, D. S. (2025). The TRIPOD-LLM reporting guideline for studies using large language models. Nature Medicine, 31(1), 60–69. https://doi.org/10.1038/s41591-024-03425-5
Garcia, P., Ma, S. P., Shah, S., Smith, M., Jeong, Y., Devon-Sand, A., Tai-Seale, M., Takazawa, K., Clutter, D., Vogt, K., Lugtu, C., Rojo, M., Lin, S., Shanafelt, T., Pfeffer, M. A., & Sharp, C. (2024). Artificial Intelligence-Generated Draft Replies to Patient Inbox Messages. JAMA Network Open, 7(3), e243201. https://doi.org/10.1001/jamanetworkopen.2024.3201
Goh, E., Gallo, R., Hom, J., Strong, E., Weng, Y., Kerman, H., Cool, J. A., Kanjee, Z., Parsons, A. S., Ahuja, N., Horvitz, E., Yang, D., Milstein, A., Olson, A. P. J., Rodman, A., & Chen, J. H. (2024). Large Language Model Influence on Diagnostic Reasoning: A Randomized Clinical Trial. JAMA Network Open, 7(10), e2440969. https://doi.org/10.1001/jamanetworkopen.2024.40969
Goodman, R. S., Patrinely, J. R., Stone, C. A., Zimmerman, E., Donald, R. R., Chang, S. S., Berkowitz, S. T., Finn, A. P., Jahangir, E., Scoville, E. A., Reese, T. S., Friedman, D. L., Bastarache, J. A., Heijden, Y. F. van der, Wright, J. J., Ye, F., Carter, N., Alexander, M. R., Choe, J. H., … Johnson, D. B. (2023). Accuracy and Reliability of Chatbot Responses to Physician Questions. JAMA Network Open, 6(10), e2336483. https://doi.org/10.1001/jamanetworkopen.2023.36483
Holmgren, A. J., Fenton, C. L., Thombley, R., Soleimani, H., Croci, R., DeMasi, O., Byron, M. E., Murray, S. G., Adler-Milstein, J. R., & Yazdany, J. (2026). Ambient Artificial Intelligence Scribes and Physician Financial Productivity. JAMA Network Open, 9(1), e2553233. https://doi.org/10.1001/jamanetworkopen.2025.53233
Lekadir, K., Frangi, A. F., Porras, A. R., Glocker, B., Cintas, C., Langlotz, C. P., Weicken, E., Asselbergs, F. W., Prior, F., Collins, G. S., Kaissis, G., Tsakou, G., Buvat, I., Kalpathy-Cramer, J., Mongan, J., Schnabel, J. A., Kushibar, K., Riklund, K., Marias, K., … Starmans, M. P. A. and. (2025). FUTURE-AI: International consensus guideline for trustworthy and deployable artificial intelligence in healthcare. BMJ (Clinical Research Ed.), 388, e081554. https://doi.org/10.1136/bmj-2024-081554
McLaughlin, N. D., Srinivas, A. N., Lowe, Z. F., Botterbush, K. S., Patel, M. S., & Avila, M. J. (2026). Large Language Model Hallucinations in Spine Surgery: A Comparative Analysis of Clinician vs Patient-Level Prompts. Neurosurgery Practice, 7(3), e000244. https://doi.org/10.1227/neuprac.0000000000000244
Pokorney, S. D., Chertow, G. M., Al-Khalidi, H. R., Gallup, D., Dignacco, P., Mussina, K., Bansal, N., Gadegbeku, C. A., Garcia, D. A., Garonzik, S., Lopes, R. D., Mahaffey, K. W., Matsuda, K., Middleton, J. P., Rymer, J. A., Sands, G. H., Thadhani, R., Thomas, K. L., Washam, J. B., … Granger, C. B. and. (2022). Apixaban for Patients With Atrial Fibrillation on Hemodialysis: A Multicenter Randomized Controlled Trial. Circulation, 146(23), 1735–1745. https://doi.org/10.1161/circulationaha.121.054990
Rotenstein, L. S., Holmgren, A. J., Thombley, R., Sriram, A., Dbouk, R. H., Jost, M., Aizenberg, D., MacDonald, S., Kanaparthy, N., Williams, B., Hsiao, A., Schwamm, L., Murray, S., Byron, M., You, J. G., Centi, A. J., Iannaccone, C., Frits, M., Landman, A. B., … Mishuris, R. G. (2026). Changes in Clinician Time Expenditure and Visit Quantity With Adoption of Artificial Intelligence-Powered Scribes: A Multisite Study. JAMA, 335(16), 1408–1417. https://doi.org/10.1001/jama.2026.2253
Shah, R., & Miranda, J. L. (2026). Leveraging generative artificial intelligence errors to teach appropriate citation usage. Journal of Microbiology & Biology Education, 27(2), e0019525. https://doi.org/10.1128/jmbe.00195-25
Stanifer, J. W., Pokorney, S. D., Chertow, G. M., Hohnloser, S. H., Wojdyla, D. M., Garonzik, S., Byon, W., Hijazi, Z., Lopes, R. D., Alexander, J. H., Wallentin, L., & Granger, C. B. (2020). Apixaban Versus Warfarin in Patients With Atrial Fibrillation and Advanced Chronic Kidney Disease. Circulation, 141(17), 1384–1392. https://doi.org/10.1161/circulationaha.119.044059
Tierney, A. A., Gayre, G., Hoberman, B., Mattern, B., Ballesca, M., Kipnis, P., Liu, V., & Lee, K. (2024). Ambient Artificial Intelligence Scribes to Alleviate the Burden of Clinical Documentation. NEJM Catalyst, 5(3). https://doi.org/10.1056/cat.23.0404
Waldock, W. J., Lam, G., Baptista, A., Walls, R., & Sam, A. H. (2025). Which curriculum components do medical students find most helpful for evaluating AI outputs? BMC Medical Education, 25(1), 195. https://doi.org/10.1186/s12909-025-06735-5
Windecker, D., Baj, G., Shiri, I., Kazaj, P. M., Kaesmacher, J., Gräni, C., & Siontis, G. C. M. (2025). Generalizability of FDA-Approved AI-Enabled Medical Devices for Clinical Use. JAMA Network Open, 8(4), e258052. https://doi.org/10.1001/jamanetworkopen.2025.8052