The average oncology appointment runs about 18 to 20 minutes. In that Talking to an AI doctor window, a patient hears their diagnosis, treatment options, possible side effects, timelines, follow-up schedules, and what happens if things go sideways. Then they go home.
Most of them remember about 20% of it. according to research presented at the Congress of the European Society for Radiotherapy and Oncology (ESTRO 2026).
That number comes from decades of patient recall research, and it doesn’t surprise anyone who works in oncology. It’s just accepted as a condition of the environment, like how you can’t hear well in a loud restaurant. The information is technically delivered. Whether it lands is a separate question entirely.
AI pre-consultation tools are being studied as one answer to this. The idea: let patients talk to an AI doctor before they ever sit across from the real one. Ask every question they’re afraid to ask. Get oriented. Show up prepared. The early findings are interesting, and not in the way most tech announcements are.
Why the first oncology appointment fails most patients
The appointment nobody prepares you for
You’ve just been told you might have cancer, or you’ve just been confirmed. Now you have 3 weeks until your first oncology appointment, a folder of PDFs you don’t understand, and a browser history full of PubMed abstracts you’re only half reading.
You write down questions. Then you lose the paper. Then you wonder if you’re even asking the right things.
This is the standard experience. Oncologists know this. Nurses know this. Patient advocates have been documenting it since the 1980s. The first real consultation is often a cognitive disaster, through no fault of anyone. The patient is frightened. The doctor is running behind. The clinic is understaffed.
So patients nod, collect their discharge papers, and walk out with a fraction of what they needed.
What happens inside a frightened brain?
This needs a physiological explanation, because it’s not obvious why smart, capable people walk out of these appointments so poorly informed.
When you’re scared, your working memory shrinks. This is documented biology. Cortisol, the stress hormone that spikes under threat, directly impairs the hippocampus. The hippocampus is what helps you consolidate new information into long-term memory. So the worse the news, the harder it is to retain what you’re told.
An oncology appointment stacks up against this problem. You’ve been anxious for weeks already. You’re in an unfamiliar environment with instruments you don’t recognize. The doctor is using terminology that means nothing to you or means something slightly different from what you assume. You’re aware that decisions are being made about your body that you don’t fully understand.
Then the doctor asks if you have questions.
Most patients say no or ask something peripheral because they don’t have a framework yet for the questions that matter. You have to know enough to ask. The first consultation is too often the first place that foundation gets built, and it’s the worst possible setting for it.
The 80% that disappears
Research on patient recall in oncology is remarkably consistent. Studies from the 1970s through recent years find that patients retain between 20% and 40% of what they hear in a consultation, with the lower end being more common when the news is distressing.
What gets lost isn’t random. Patients tend to retain the diagnostic labels, the first thing said, and the last thing said. Everything in the middle, often where the real clinical detail lives, fades fast. They remember “chemotherapy” but not which one. They remember “surgery” but not what kind or when.
This has real consequences. Patients who misunderstand their treatment plan are less likely to follow it correctly. The patients who don’t know what side effects to expect are more likely to stop treatment when those side effects hit. Patients who don’t understand why a protocol exists are less likely to adhere to it when it’s difficult.
Understanding isn’t a comfort measure. It’s a clinical one.
What an AI pre-consultation actually is
The tool, explained plainly
The tools being studied here aren’t chatbots that point you to WebMD articles. They’re purpose-built conversational systems trained on clinical oncology content. Some are institution-specific platforms built by cancer centers. Others are standalone products developed by health tech companies working with oncologists.
The experience typically works like this: a patient logs in 1 to 2 weeks before their real appointment. They enter their diagnosis or upload their pathology report. The AI walks them through what the diagnosis means, what treatment categories are usually discussed for their case, what questions are worth preparing, and what to expect from the appointment itself.
It is not a diagnostic tool. It doesn’t prescribe. It doesn’t replace the physician’s judgment about an individual patient’s case. The best analogy is a well-informed friend who happens to have read every clinical guideline on your cancer type and has unlimited time to talk it through with you.
Text, voice, and the avatar question
The format varies across tools. Some are text-only, more like an advanced FAQ system with follow-up capability. Some are voice-based, which lowers the reading literacy barrier considerably. A few use avatar interfaces, a digital figure with a face that responds conversationally.
The avatar versions matter for a specific reason. There’s something about a face, even a synthetic one, that makes hard conversations feel slightly more manageable. Patients report feeling less alone. Whether this is meaningful or just aesthetics is genuinely unclear, but the patient satisfaction scores for avatar-based tools tend to run higher than text-only equivalents.
More practically: patients can come back to these tools as many times as they want. They can ask the same question 5 times in a row until they actually get it. They can ask something, go away and think, and come back an hour later to ask a follow-up. That kind of iterative understanding is simply not possible inside an 18-minute appointment.
What it is not
It’s worth being direct about what these tools can’t do, because the hype often outruns the reality.
AI pre-consultation tools are not therapy. They’re not crisis support. They don’t replace the human relationship with a physician. They can’t read your face, notice when you’ve gone quiet, or adjust tone based on how you’re holding up in real time. A good oncologist does all of those things instinctively. The AI does none of them.
And they don’t know you specifically. They know your cancer type, generally. Your individual biology, your comorbidities, your particular pathology—those come from your care team. The AI is working with population-level clinical knowledge, not personal data. That distinction matters, and the tools that communicate it clearly perform better than the ones that don’t.
The research: what the evidence actually shows
Knowledge retention improvements
Multiple studies across different cancer types have measured what happens to knowledge scores when patients use AI pre-consultation tools. The consistent finding: preparation improves retention.
Research on breast cancer patients comparing those who used AI pre-consultation tools with those who went through the standard pathway shows knowledge score improvements in the range of 25% to 40% at the point immediately after the real consultation. The control groups show the typical pattern: modest retention right after and a significant drop-off at the one-week mark.
The mechanism is straightforward. When a patient already has a framework for what’s being discussed, new information has somewhere to attach. The oncologist says “FOLFOX” and the patient has context for what that means, rather than just writing the word down phonetically and planning to Google it later.
Anxiety: where the evidence is strong and where it gets complicated
Studies measuring anxiety in oncology patients using standard instruments (the State-Trait Anxiety Inventory is the most common) find lower pre-appointment anxiety in patients who used AI pre-consultation tools compared to controls. The reduction is most measurable in the 48-hour window before the appointment. That’s when anxiety in unprepared patients tends to spike hard.
The relationship works through comprehension. Patients who understand what an appointment will involve are less surprised by it. Less surprise means less reactive cortisol. Less cortisol during the appointment means better retention of what gets said.
The anxiety findings hold strongly in some studies and more modestly in others. The pattern across the literature suggests the tools help most when anxiety is at a manageable level to start. Patients in acute distress respond less predictably to an AI interface. Some find it genuinely helpful. Others find the simulation of care alienating in a way that increases distress.
So: AI pre-consultation reliably helps patients in the moderate-anxiety range. It’s less reliable as a support tool for patients in crisis.
What happens to the appointment itself?
Research tracking appointment dynamics rather than just patient recall shows something interesting: appointments with prepared patients run slightly longer and get rated by both patients and physicians as more productive.
The length increase in several studies is about 3 to 5 minutes. That sounds small. In oncology, where appointment time is a genuinely scarce resource, 5 minutes is not small. And the extra time goes toward specific discussion rather than orientation. The physician isn’t spending the first 8 minutes explaining what staging means. That part’s already handled.
Nursing teams at institutions piloting these tools report a reduction in the follow-up calls where patients phone back after going home and realizing they didn’t understand what they’d agreed to. That’s a measurable workflow change. It also means patients aren’t sitting at home confused and frightened for 3 days while waiting for someone to call them back.
What patients actually ask when no one is judging them
The questions they don’t ask the doctor
One of the more revealing findings from AI pre-consultation research comes from analyzing the actual questions patients ask when they know they’re not being timed and there’s no social pressure.
They ask about prognosis far more directly than they do in real appointments. “What are my odds?” in many different phrasings. In real appointments, patients often avoid this question because they’re afraid of the answer or because they don’t want to make the physician uncomfortable. With an AI, those social calculations disappear.
They ask what treatment will feel like rather than what it is. “Can I work during chemotherapy?” “Will I lose my hair with this specific regimen, or is it a different type?” “Will the nausea be constant, or does it come and go?” These are practical, daily-life questions. They’re not in the clinical literature the oncologist is drawing from, but they matter enormously to the patient deciding whether to agree to a treatment.
They ask the same question multiple times. They backtrack. They say, “Wait, explain that again from the beginning.” This is normal human learning behavior, and it gets suppressed almost entirely in a real appointment because patients don’t want to waste the doctor’s time or look slow.
The AI doesn’t check a clock. It doesn’t have a waiting room.
The self-advocacy effect
Patients who arrive at consultations having already articulated what they don’t know tend to advocate for themselves more effectively. They ask follow-up questions rather than just nodding. They’re more likely to say “I’m not sure I understood that” to a human physician because they’ve already practiced saying it.
Research on patient self-advocacy in oncology links it to better treatment decisions and better quality of life during treatment. It correlates with adherence. Patients who feel like active participants in their care, rather than recipients of it, do better across multiple outcome measures.
AI pre-consultation is, among other things, a low-stakes rehearsal for speaking up.
The limits worth taking seriously
Health literacy gaps
The tools perform best with patients who have higher health literacy and comfort with digital interfaces. Patients who struggle with reading, or who aren’t comfortable navigating apps, get measurably less from them. Some studies document this gap explicitly. Others don’t mention it, which is itself worth noting.
This is a design problem with a design solution. Tools built at a plain-language reading level, with audio alternatives and intuitive navigation, close the gap considerably. The tools that assume literacy and digital fluency serve a narrower population than their developers usually advertise.
Language coverage is thin.
Most AI pre-consultation tools are available in English and, in some cases, Spanish. Oncology serves enormously diverse populations. A tool that works well for a fluent English speaker doesn’t automatically translate to someone whose primary language is Mandarin, Arabic, or Tamil. Machine translation in these tools is improving, but the linguistic nuance in explaining oncology concepts accurately across languages is not a solved problem.
The wrong-information risk
A concern physicians raise consistently is the patient who arrives confident about something the AI got subtly wrong.
These tools are trained on population-level clinical data. The patient sitting across from the oncologist is not the population average. There are edge cases. Individual biology varies. Comorbidities change recommendations. A tool that gives a general explanation of “how chemotherapy for your cancer type usually works” might inadvertently anchor a patient on expectations that don’t match their specific situation.
The better-designed tools flag this explicitly: “This information describes general approaches. Your doctor will tailor everything to your specific case.” The weaker ones don’t, and the confident patient with subtly incorrect information is harder to course-correct than the uncertain patient who comes in with questions.
The emotional gap
AI can approximate empathy. It can respond to distress cues with appropriate language. It cannot actually be with someone in the way a skilled physician or nurse can.
For patients in acute psychological distress, who are not simply anxious but genuinely overwhelmed by their diagnosis, an AI interface can feel cold or insufficient in ways that worsen rather than improve their state. The tools that integrate referrals to human support, a nurse navigator, a social worker, and a patient advocate handle this better than tools that treat every patient interaction as a knowledge-delivery problem.
What physicians actually think
The case for it
Oncologists who’ve worked with institutions piloting AI pre-consultation tools generally describe better appointments. Patients who follow the conversation rather than nodding blankly. Patients who ask questions that indicate real understanding. Patients who catch something and say, “Wait, I thought you said earlier that…” rather than storing their confusion for later.
The reduction in follow-up calls matters to physicians too. Every call back from a patient who didn’t understand their discharge instructions takes time, often outside scheduled hours, often from a physician or nurse who’s already managing a full day. Reducing that particular friction has value beyond the patient experience.
The case against
The physician skepticism that isn’t purely turf protection centers on a few real concerns.
Liability is murky. If a patient makes a treatment decision partly based on information from an AI pre-consultation tool, and that decision leads to harm, the accountability structure is unclear. This is a legitimate problem for institutional adoption. The tools exist in a space that medical malpractice law hasn’t fully caught up to.
Some physicians worry about depersonalization. There’s a version of this technology that gets used by systems to reduce physician time with patients, framing AI pre-consultation as a substitute for adequate appointment length rather than a supplement to it. That concern is worth taking seriously. The evidence shows these tools improve outcomes when they’re added to adequate care. What happens when they’re used to replace it is a different question and a less studied one.
Who gets the most out of this?
First-time oncology patients
Patients facing their first oncology appointment after a new diagnosis show the largest knowledge gains. The baseline is lowest, the content is entirely unfamiliar, and the framework-building effect of pre-consultation is most pronounced. There’s the most room to improve, and the improvement is measurable.
Clinical trial candidates
Patients being considered for clinical trials are navigating genuinely complicated consent and eligibility processes. Understanding what a Phase 2 trial means, what randomization is, and what the difference between standard of care and investigational treatment is: these are specific concepts that take real explanation. Pre-consultation tools designed around clinical trial literacy show strong results in this population.
Patients with lower baseline health literacy
Counter-intuitively, patients with lower health literacy at baseline show large knowledge gains when the tool is well-designed. The key phrase is “when the tool is well-designed.” Plain language, audio support, and simple navigation: when these are built in from the ground up rather than added later, the gap between high-literacy and low-literacy patients narrows considerably.
What’s coming in the next few years?
Personalization from actual records
Current tools work from a patient’s self-reported diagnosis. The next generation integrates with electronic health records, so the AI is working from the actual pathology report, the specific genetic markers, and the treatment history. That changes the quality of the conversation significantly. General explanations give way to specific ones.
Real-time integration with appointments
A few institutions are already piloting systems where the oncologist, before entering the appointment, sees a summary of what their patient asked the AI, what they understood, and what they’re still confused about. That summary replaces the first 5 minutes of orientation at the start of the appointment and points the physician directly to where clarification is needed.
Post-consultation follow-up
Early work is being done on using these tools after the appointment, not just before it. Patients forget the majority of what they heard. A follow-up AI session 24 to 48 hours later, reinforcing the key decisions and catching what slipped, could extend retention in ways that matter for treatment adherence. The pre-consultation and post-consultation use cases are probably more powerful in combination than either is alone.
Multimodal interfaces
Voice-first interfaces with more sophisticated natural language understanding are already replacing text-only versions in newer tools. Visual aids that let patients point to a diagram and ask questions in plain language are being tested. These reduce the literacy barrier in ways that text interfaces cannot.
The actual stakes
Cancer patients make serious decisions under enormous psychological pressure, usually without adequate preparation. That’s been documented for 40 years. AI pre-consultation tools don’t fix the whole problem. They address a specific, measurable part of it: the information gap between diagnosis and first consultation.
The knowledge improvements in the research are real. The anxiety reduction in the pre-appointment window is real. The more productive appointments are real. Patients who understand what’s happening to them make better decisions about their own care. They adhere to treatment protocols better. They report higher quality of life during treatment. These are meaningful outcomes for people facing some of the hardest decisions of their lives.
The tools aren’t finished. The health literacy gaps are real. The language coverage is thin. The liability frameworks don’t exist yet. But none of those limitations change the direction.
Knowing what to expect. Knowing what to ask. Knowing that your questions are worth asking. Those things change what patients do in the appointment and what they do after it. For someone walking into a cancer consultation scared and underprepared, that’s not a small thing.
It’s the whole thing.
