Why Patients Are Adopting Self-Serve Health Care: Convenience, Access Gaps, and the Hybrid Model

Patients now handle many parts of their care on their own, despite still seeing clinicians. They book appointments online, read test results before a clinician calls, track health data on wearable devices, order home tests, and ask artificial intelligence (AI) chatbots to explain health information.

Self-serve health care is a working term for patient-initiated tasks that once ran through staff, clinicians, or scheduled visits. A patient can start a task alone and still need a clinician to interpret a result, make a clinical decision, or provide follow-up.

The World Health Organization (WHO) self-care guideline lists convenience, cost, empowerment, and a better fit with daily life among the reasons people choose self-care. (32) It also notes that lack of access to quality care can drive self-care. This article examines why patients are self-serving more of their health care, where that approach falls short, and how practices can keep clinical support connected to it.

What patients now do for themselves, and how widely

Self-service now includes both routine administrative tasks and activities that involve health information or clinical judgment.

Six task types under one umbrella

WHO defines self-care interventions as drugs, devices, diagnostics, and digital tools that people can use fully or partly outside formal health services, with or without direct supervision. (32) Six types of patient self-service fit under that definition:

• Administrative transactions: booking, rescheduling, registration, intake, cost estimates, payment, records requests, and refill requests.
• Information access: viewing results and clinical notes, often before a clinician has interpreted them, and messaging the care team.
• Monitoring: smartwatches, smart rings, connected blood pressure cuffs, smart scales, and continuous glucose monitors used between visits.
• Self-testing and home collection: tests that differ in who processes the specimen, who interprets the result, and who supports the next decision.
• Patient-initiated virtual visits: the patient starts the encounter from home, and a clinician makes the clinical decision.
• AI-mediated navigation: conversational tools that explain terms, interpret results, and suggest whether to seek care, with unproven clinical accuracy for individual patients.

These types differ in clinical risk, meaning the chance that a patient acting alone misses, delays, or misjudges something that needs care. Rescheduling an appointment online carries little of that risk. Reading an abnormal result without an explanation, or deciding that a new symptom needs no evaluation, carries much more, because a wrong conclusion can delay diagnosis or treatment.

Where the task ends and the clinical decision begins

Each self-serve task sits next to a clinical decision that stays with the care team. A patient can request a refill online, but a clinician still authorizes it and decides whether reassessment is needed first. Booking an appointment leaves open whether the symptoms can safely wait until that date. A portal can display a result, but the patient may still need help understanding it and deciding what to do next.

Collecting a specimen at home usually leaves processing and interpretation to a laboratory, and even a test performed entirely at home points users back to care. The U.S. Food and Drug Administration (FDA) advises anyone with a positive result on the first fully at-home test for chlamydia, gonorrhea, and trichomoniasis to seek medical care. (29)

Adoption signals measure different behaviors

Portal access has more than doubled in a decade. In 2024, 65% of U.S. adults were offered and accessed a patient portal or online record, up from 25% in 2014. (21) Access meant at least one use in the past year. It ran higher among people managing a chronic condition (69%) or a recent cancer diagnosis (76%). Proxy access by caregivers rose from 24% in 2020 to 51% in 2024.

In the same national survey, 70% of patients with portal results viewed them there, and 58% viewed them before hearing from a clinician. (22) An August 2026 analysis using a narrower denominator put early viewing at 68.6%. (2)

Device and AI use are rising as well. Wearable use among U.S. adults rose from 30.2% in 2020 to 41.1% in 2024. (19) Estimates of AI use for health depend on the question asked:

• 32% of adults used AI for health information in the past year in a February–March 2026 poll. (14)
• 29% used it at least monthly in a May 2026 poll, up from 17% in June 2024. (26)
• 34% used chatbots for at least one of eight health purposes in a June 2026 survey. (12)
• 25% had ever used AI for health information in a late-2025 survey of 5,660 adults. (20)

Because these surveys focus on different health behaviors and distinct population groups, their findings cannot be combined into a single, comprehensive adoption rate.

Supplementation and substitution coexist

Many patients use AI around their visits, and a smaller group uses it in place of one. Among adults who used AI for health information in the past 30 days, 59% researched before a visit and 56% researched afterward. (20) The survey's funding partner reported that 84% of these users still saw a provider, and 14% skipped a visit they otherwise would have made because of AI-generated information. (31) The survey did not assess whether skipped visits were appropriate.

AI use can change different parts of care: preparation, access, interpretation, treatment decisions, or follow-up. A patient who arrives with written questions likely used AI to prepare. A patient who stopped a medication after a chatbot conversation used it to make a treatment decision, and that history needs review at the visit.

Why patients self-serve: five drivers on two tracks

Patients give five broad reasons for self-serving: speed, appointment scarcity, cost, early access to their own data, and privacy or past experiences of being dismissed. The same behavior can reflect a preference or a workaround for care the patient cannot reach.

Speed and care that fits daily life

Speed is the reason patients give most consistently. In a February–March 2026 poll, 65% of past-year AI health users cited quick or immediate information as a major reason. (14) In a late-2025 survey, 71% of recent AI health users wanted answers quickly, and 21% said they lacked time to make an appointment. (31)

Consumer industries like banking and retail have made convenience the norm. Phone queues and limited scheduling hours now feel needlessly clunky. Choosing an easier way to handle routine tasks does not mean patients value clinical expertise any less.

Appointment scarcity

New-patient waits are long and getting longer. An early-2025 survey of physician offices in 15 large U.S. metro areas found an average new-patient wait of 31 days across six specialties, the longest since the survey began in 2004. (1) Family medicine averaged 23.5 days, and metro averages ranged from 12 days in Atlanta to 65 days in Boston. Insurance acceptance can narrow the available options further. Of surveyed physicians, 82% accepted Medicare and 53% accepted Medicaid. These findings are specific to the surveyed specialties and metros.

Long waits push some patients toward AI. In the February–March 2026 poll, 18% of AI health users cited having no regular provider or being unable to get an appointment as a major reason, including 38% of users under 30 and 6% of users 50 and older. (14) When patients turn to AI because they cannot get an appointment, the alternative may be waiting for care rather than immediate access to a clinician.

Cost and financial exclusion

Cost drives a substantial share of self-service among lower-income patients. In the February–March 2026 poll, 19% of past-year AI health users cited being unable to afford a provider as a major reason. (14) The figure was 32% among users with household incomes below $40,000, and 25% of that group cited lacking a regular provider. In a late-2025 survey, 32% of recent AI users earning under $24,000 used AI because they could not pay for a visit, compared with 2% of those earning $180,000 or more. (20) Cost limits care more broadly as well. In 2026 survey data, 28% of U.S. adults skipped or postponed needed care in the past year because of cost. (11)

Self-service can also cost more than its entry price suggests. A low-cost test or app can lead to a visit, repeat testing, and treatment when a result is abnormal, and each step adds to the total. For patients who describe cost-driven self-care, practices can share self-pay prices, sliding-fee options, or community health center referrals.

Direct data access and the interpretation gap

Patients now see many results before anyone explains them. Under the 21st Century Cures Act final rule, many test results are released to patients through portals as soon as they are available. (2)

Patients may turn to AI for explanations after results are released and before they speak with a clinician. In the February–March 2026 poll, 28% of AI health users cited receiving results before discussing them with a provider as a major reason for using AI, and 41% wanted information before deciding whether to see a provider. (14) Of AI health users, 41% had entered test results or clinicians' notes into an AI tool, equal to 13% of all adults. Patients also use chatbots to make sense of symptoms, diagnoses, and lab results. In a June 2026 survey, U.S. adults reported using chatbots to identify the cause of symptoms (25%), learn more about a diagnosis (22%), and interpret lab results (20%). (12)

A fluent AI explanation can still be wrong. Asking at follow-up whether the patient used AI to read a result gives the clinician a chance to correct a misreading.

Privacy, dignity, and feeling heard

Some patients self-serve because asking a person feels uncomfortable or has gone badly before. In the February–March 2026 poll, 36% of AI health users said they felt more comfortable looking up health topics privately. (14) In the late-2025 survey, 21% of recent AI health users had felt dismissed or ignored by a provider, and 18% were too embarrassed to ask a person. (20)(31) Among adults who use social media for health information, 36% cited learning from people with similar experiences as a major reason. (26) Ultimately, many patients are just looking for a private, judgment-free space to ask questions and find comfort in others who truly understand what they are going through.

Privacy concerns are especially relevant for stigmatized health topics. The FDA has noted that fear or anxiety about sexual health testing can delay diagnosis or treatment. (29) A tool that feels private may still store and share data under the company's own terms.

Two tracks, and the patients in between

The same adoption figures include both convenience-driven and access-driven use. On the convenience track, digitally fluent patients choose self-service for speed, control, and personal metrics. On the access-workaround track, patients turn to AI, home testing, or remote channels because conventional care costs too much, takes too long, or feels uncomfortable. The same patient can move between the two. A patient may book online for convenience, then turn to a chatbot when the next appointment is weeks away or a visit costs more than they can pay. An overall adoption figure does not distinguish between these motivations.

In a June 2026 survey, 48% of upper-income adults used chatbots for health, compared with 33% of middle-income and 28% of lower-income adults, while cost-driven motives concentrate among lower-income users. (12) Uninsured adults were more likely than insured adults to use AI for mental health information (30% vs. 14%). (14) In pooled 2020–2024 national data, 74.9% of wearable users had household incomes of $50,000 or more. (19)

Patients managing chronic conditions or cancer, and caregivers with proxy access, use portals at high rates. (21) They are also more affected when records are fragmented or abnormal findings go unexplained. For an individual patient, the practice can ask directly what support they want.

Where self-service falls short of care

Self-service has clear limits when patients need to interpret results, act on AI output, decide what testing means, protect sensitive data, or connect information across different services.

Access without understanding

About two-thirds of patients who read results before talking to a clinician say they understood them. In 2024 national survey data, 66% of early viewers reported understanding their results well or very well, including 77% of patients with high digital literacy and 53% of those with low digital literacy. (22) A separate analysis of the same survey reported the full distribution of those ratings. Early viewers rated their understanding as poor in 6.6%, fairly well in 26.8%, well in 31.7%, and very well in 34.9%. (2) Together, 33.4% rated their understanding as poor or fairly well.

Encouragement from a clinician was also linked to better understanding. Patients whose provider encouraged them to use the portal understood their early results more often than patients who were not encouraged (67% vs. 54%). (22) The survey measured encouragement and understanding at the same point in time, so it cannot show that one caused the other. Given how much digital literacy affects understanding, patients who struggle with technology may need more follow-up, particularly if they have accessed abnormal results on their own.

AI that knows the answer but does not transfer it

Large language models correctly identified the relevant condition in 94.9% of test scenarios when researchers entered the scenarios directly. People using the same models did so less than 34.5% of the time. In this randomized, preregistered study, 1,298 adults in the United Kingdom (UK) assessed written scenarios using GPT-4o, Llama 3, or Command R+. A control group used standard resources, mainly search engines and the National Health Service (NHS) website. (3)

The models also chose the right level of care in 56.3% of scenarios when evaluated alone. Participants using AI chose it in fewer than 44.2%, about as often as the control group. The control group often named relevant conditions more often. The control group also named relevant conditions more often than participants using AI. Across all groups, participants tended to underestimate how urgent their scenario was. Strong model performance did not translate into better conclusions by the people using them.

Even with that performance gap, users often rate chatbots as convenient or helpful. In an October 2025 survey, 48% of adults who get health information from chatbots rated it highly convenient, and 18% rated it highly accurate. (18) A separate June 2026 survey found that 47% of chatbot health users rated the information extremely or very helpful. (12)

Missed follow-up and false reassurance

Many patients who use AI for health information skip clinical follow-up. In the February–March 2026 poll, 42% of adults who used AI for physical health information did not follow up with a doctor or other provider. (14) Among those who used it for mental health information, 58% did not follow up with a mental health professional. The poll did not capture the underlying needs, so it cannot show whether skipping follow-up was inappropriate.

A false-negative self-test or reassuring AI answer can delay assessment, and a false positive can cause anxiety and unnecessary care. Clinicians can ask at intake whether an app, test, or AI tool influenced the patient's decision to come in or to wait. When it did, the clinician can review what the tool said before it shapes diagnosis, treatment, or urgency.

Tests that generate more testing

Self-tests and wellness panels can lead to more testing and visits. A 2025 review of 30 self-tests bought from UK retailers in 2023, covering 20 biomarkers across 19 conditions, found that 27 recommended professional follow-up after a positive or abnormal result and 14 recommended it after a negative or normal result. (5) Many also lacked basic guidance. Eight stated who should use the test and when, and seven stated what to do after the result. For 11 of the 19 conditions, the marketed use conflicted with evidence-based guidance, a finding about appropriate use. A companion review rated 18 of the 30 tests high risk for equipment, sampling, or interpretation concerns, and a December 2024 repeat search found 63 tests on sale. (10)

Broad wellness panels carry a related risk. A systematic review of 484 direct-to-consumer tests advertised online in Australia classified 10.7% as having potential clinical utility. (25) Another 30.6% had limited clinical utility, 41.9% were commercial health checks without an evidence base, and 16.7% used methods or targeted conditions unrecognized by the general medical community. Healthy consumers were the main target market, and 56% of products did not state that they offered consultation before or after testing. The authors identified overdiagnosis, false positives, and further tests and clinical care prompted by the results as potential harms.

A borderline or unexpected result on a broad panel can lead to repeat testing and extra appointments. If a patient brings in one of these panels, the clinician can sort through which results are clinically meaningful based on the patient’s history and risk factors.

Private-feeling is not the same as protected

Patients worry about AI privacy and upload records anyway. In the February–March 2026 poll, 77% of adults were very or somewhat concerned about the privacy of medical information given to AI tools, including 65% of those who had entered their own medical information. (14) Thirteen percent of all adults had uploaded personal medical information to an AI tool.

The Health Insurance Portability and Accountability Act (HIPAA) applies to covered entities and their business associates, and many consumer apps, wearables, and connected devices fall outside it. (27) The Federal Trade Commission (FTC) Health Breach Notification Rule, amended effective July 29, 2024, covers most health apps and similar technologies that HIPAA does not. (9) Under the rule, a breach includes an unauthorized disclosure as well as a cyberattack.

Three FTC enforcement actions involved health companies accused of sharing users' health data for advertising or marketing purposes:

• GoodRx, February 2023: the first action under the rule, with a $1.5 million civil penalty, for sharing health data with advertising platforms after promising not to. (6)
• Easy Healthcare (Premom), May 2023: the second action, with a $100,000 civil penalty and a permanent ban on sharing users' health data with third parties for advertising. (7)
• BetterHelp, July 2023: a separate FTC Act Section 5 order requiring $7.8 million for consumer refunds. (8)

These cases show that how private a health app feels and how it actually handles data are separate questions.

Fragmentation and shifted burden

Self-service can spread a patient's health information across several systems. In 2024, 59% of individuals had multiple online medical records or portals, while 7% used an app that combined them. (21) Lab portals, wearable apps, telehealth services, pharmacies, and AI tools can create additional places where health information is stored.

The WHO guideline notes that self-care can shift costs from the health system to the individual and may worsen inequities in access. (32) Patients may also have to keep track of results, share information between clinicians, and arrange follow-up across services. Asking which outside portals, apps, tests, or devices a patient uses can help the care team identify information that may not appear in the main record.

What separates supported self-care from unsupported self-service

Evidence of benefit comes from specific self-care models that link patient-collected data to clinician feedback and a defined next step. These findings do not show that self-service in general improves outcomes.

Evidence from an integrated home monitoring model

Home blood pressure monitoring lowered blood pressure when readings fed back to both patients and clinicians. The HOME BP trial randomized 622 UK adults with treated but poorly controlled hypertension and internet access across 76 practices. (13) The intervention combined self-monitoring, guided digital self-management, and feedback to patients and clinicians. At one year, systolic pressure was 3.4 mmHg lower than with usual care (95% confidence interval [CI], −6.1 to −0.8). The diastolic difference was not significant, and adverse effects were similar. Participants were 94% white British, with average ages in the mid-60s, which limits generalizability.

A meta-analysis of 65 trials found that home monitoring lowered systolic pressure by 3.27 mmHg compared with usual care. (23) The effect was larger with telemonitoring or support from clinical staff, and it became uncertain after about 20 months. These findings support home monitoring with clinical feedback built in.

A home test built around what comes next

Even a test completed entirely at home may still require clinical follow-up. In March 2025, the FDA authorized the first nonprescription test for chlamydia, gonorrhea, and trichomoniasis performed entirely at home. (29) The test uses a self-collected vaginal swab, is intended for females with or without symptoms, and returns results in about 30 minutes. It correctly identified:

• 97.2% of positive and 98.8% of negative chlamydia samples
• 100% of positive and 99.1% of negative gonorrhea samples
• 97.8% of positive and 98.5% of negative trichomoniasis samples

The FDA advises people with positive results to seek medical care, and people with symptoms, recent exposure, or other concerns despite a negative result to contact a provider. It names false negatives, which delay treatment and allow spread, and false positives, which can lead to unnecessary treatment, as the main risks. The authorization covers this one test and its intended use.

Wellness wearables after FDA's 2026 guidance

Patients may bring readings from wearable products that were not reviewed as medical devices. On January 6, 2026, the FDA issued revised general wellness guidance that replaced the 2019 version. (30) Under it, the FDA may treat certain products that use noninvasive sensing to estimate blood pressure, oxygen saturation, blood glucose, or heart rate variability as general wellness products, outside premarket review. To qualify, a product must meet every condition:

• intended solely for wellness use
• noninvasive and not implanted
• free of interventions or technologies that could pose a safety risk to users or others without specific regulatory controls
• not intended to diagnose, cure, mitigate, prevent, or treat disease
• not intended to substitute for an FDA-authorized device
• no claims, functions, or outputs that prompt specific clinical action
• no values that mimic clinical ones unless those values are validated

Products that penetrate the skin, such as microneedle glucose sensors, fall outside the low-risk category. Clinically important readings should be confirmed with an appropriate validated measurement before they guide care.

Six questions that separate supported self-care

The WHO guideline emphasizes the need to connect independent self-care with access to quality health care. (32) Before a practice recommends a self-service tool or builds one into care, six questions help show where additional support may be needed.

Three concern the tool itself:

• Task: does it move logistics, or a clinical judgment the patient cannot safely make alone?
• Evidence: is the test, device, or tool validated for this person and this purpose?
• Guidance: does the patient know who should use it, what results mean, and what to do next?

Three concern the care around it:

• Escalation: is there a fast, named route to a clinician when results, symptoms, or worry escalate?
• Data: is it clear which law governs the data and who can see it?
• Cost and burden: does it reduce total cost and effort, or add confirmatory visits and tests?

A gap in any of these areas may require more support before the tool is recommended or its result is used in care. Clinical decisions may need clinician review. Uncertain or unvalidated results may need confirmation. Patients need clear instructions, a way to ask questions, and an explanation of any likely follow-up or additional cost. Privacy also needs to be reviewed separately from how convenient or private the tool feels.

How practices can respond: the hybrid model

Practices can move routine logistics online while keeping clinicians responsible for interpreting results, answering clinical questions, and arranging follow-up.

Make administrative self-service the default

Routine logistics carry the least clinical risk. Booking, rescheduling, registration, intake, cost estimates, payment, records requests, and straightforward refill requests can move online, with clinical authorization kept separate. An online refill request, for example, routes to a clinician review queue.

An online scheduler shows available appointments without adding any. With average new-patient waits of 31 days in large metro areas, a scheduler may simply make scarce capacity more visible. (1)

Staffed phone and in-person routes remain necessary for patients with low digital literacy, limited internet access, or a preference for help. A one-question survey after each online task can show which patients worked alone by choice and which did so because help was unavailable.

Pair immediate results with context

Federal rules limit delays in releasing results, and the practice controls the context around each release. The information-blocking regulations have applied since April 5, 2021, to health care providers, certified health information technology (IT) developers, and health information networks and exchanges. (17) Information blocking is a practice likely to interfere with access to, exchange of, or use of electronic health information, unless required by law or covered by an exception. (15)

On September 3, 2025, the U.S. Department of Health and Human Services (HHS) announced active enforcement led by the Assistant Secretary for Technology Policy/Office of the National Coordinator for Health Information Technology (ASTP/ONC) and the HHS Office of Inspector General. (28) Certified developers and health information networks and exchanges face civil money penalties of up to $1 million per violation. Providers face disincentives under a rule effective July 31, 2024, applied through Medicare programs for hospitals and clinicians and the Medicare Shared Savings Program. (4)

Some states require delays. Texas Senate Bill (SB) 922, effective September 1, 2025, bars electronic disclosure of certain sensitive results before the third day after they are finalized. (24) These include pathology or radiology reports reasonably likely to show malignancy and results that may reveal a genetic marker. Compliance counsel decides whether a specific delay fits an exception or a state requirement.

Within those rules, a practice controls the communication around each result:

• Before the test: explain its purpose, possible findings, and expected follow-up.
• At ordering: where workflows and the patient's request allow, record how the patient wants to receive results.
• At release: place plain-language context next to the result, covering what it is, any trend, known limitations, and the planned next step.
• After viewing: offer a clear route to ask a question or book a discussion.
• Across the team: make responsibility for reviewing results and closing unresolved follow-up visible.

Patients whose provider encouraged portal use were more likely to view results there. Seventy percent of encouraged patients viewed results in the portal, compared with 24% of those not encouraged. (22) Clinicians can mention the portal when they order a test and tell patients when results are likely to appear.

Evaluate patient-brought AI output, device data, and test results

Patients may bring device data and other outside health information into visits. In 2024, 73.4% of wearable users were willing to share device data with a clinician, and 19.2% had shared health information from a device or smartphone. (19) Clinicians have raised liability concerns about patient-generated health data, including acting on inaccurate data or not acting on data received. (16)

Before using that information to guide care:

• Wearable readings: confirm any clinically important reading with a validated measurement before acting on it.
• Direct-to-consumer results: review the indication, the testing laboratory, and whether the result changes management, and set a policy on whether and how outside results are documented.
• AI output: ask what the patient entered and what the tool advised, then reconcile it with the history and examination. Use the patient's question as a starting point for the visit.

Set expectations for AI use and data sharing

Patients often use consumer AI tools on their own, including to interpret symptoms, test results, or clinical notes. Practices can give patients basic privacy guidance before they share medical information with these tools. A portal message or after-visit summary can explain that many consumer apps fall outside HIPAA and that information uploaded to them is handled under the company's own privacy terms.

Practices also need to be clear when they use AI themselves. Seventy-two percent of U.S. adults say it is extremely or very important for a provider to tell them when AI is used in their care. (33) Practices using patient-facing AI can include that disclosure in consent materials and explain what the tool does with patient information.

Before offering an AI tool to patients, practices can ask vendors for evidence from testing with real users. In a 2026 randomized study, model performance alone did not predict how well people performed when using the model. (3)

Build escalation into every self-service front door

Routine administrative tasks can remain self-service. Results and interpretation need plain-language context and a clear route for questions. Symptoms or decisions about diagnosis, treatment, or urgency need clinician review.

The WHO guideline notes that some people use self-care interventions independently, while others need guidance and support. (32) Abnormal results, worsening symptoms, uncertainty, patient concern, or an AI answer that raises a clinical question are reasons to involve a clinician. A portal result, home-monitoring program, or patient-facing AI tool can show a clear contact option, such as a nurse line, message pool, or appointment route, on the same screen as the result or advice.

Measure whether the patient's need was resolved

Usage, satisfaction, and click-through rates show whether patients use a self-service option. They do not show whether patients understood what happened next, received appropriate follow-up, or had their problem resolved. Practices can also track:

• Access: completed tasks, abandoned attempts, and time to an appropriate appointment or response.
• Interpretation: comprehension of results, understanding of next steps, and unresolved questions.
• Follow-through: follow-up after abnormal results, completed assessments, and time to treatment.
• Burden: patient time and total cost, repeat or confirmatory testing, repeated data entry, and inbox volume.
• Equity: whether digital-only pathways leave out patients with low digital literacy or limited internet access.

Comprehension is one measure that can be collected directly. The national results survey asked patients how well they understood the results they viewed, and practices can ask a similar question after results are released. (22) Services that make clinical claims should also track clinical outcomes and safety events.

Frequently asked questions (FAQs)

Are patients using AI and other self-service tools instead of seeing a clinician, or alongside their visits?

In a late-2025 national survey, 84% of recent AI health users still saw a provider, and 14% skipped a visit they otherwise would have made.

Does growing patient self-service mean declining trust in clinicians?

Speed, privacy, cost, and difficulty accessing care are common reasons patients give for using self-service tools, and in one survey 84% of recent AI health users still saw a provider. These surveys do not measure trust directly, so self-service use alone should not be read as evidence that patients distrust clinicians.

How can a practice tell preferred independence from an access workaround?

Asking why the patient used the tool helps, since cost, long waits, or lack of a regular clinician point to a workaround.

Why can higher chatbot adoption among upper-income adults coexist with cost-driven use among lower-income patients?

Upper-income adults use chatbots more often, and lower-income users are far more likely to cite cost as their reason.

Does immediate portal access to results reduce the need for clinician communication?

In 2024, about one-third of patients who viewed results before hearing from a clinician did not report understanding them well.

Can a practice hold test results until a clinician has reviewed them?

Federal information-blocking rules restrict delays unless an exception or a legal requirement, such as a state delayed-release law, applies. Compliance counsel can confirm whether a specific delay qualifies.

How should a clinician respond when a patient brings an AI interpretation of their lab results?

The clinician can ask what the patient entered and what the tool said, then reconcile that output with the history, examination, and result.

Are AI chatbots accurate enough for patients to decide whether they need care?

In a 2026 randomized study, people using chatbots chose the correct level of care in fewer than 44.2% of scenarios, about as often as people using their usual sources.

Does HIPAA protect health information patients upload to consumer AI apps or wearables?

HIPAA applies to covered entities and their business associates, and many consumer apps and wearables fall outside it. The FTC Health Breach Notification Rule covers many of those apps.

Should a practice accept direct-to-consumer lab results or reorder the tests?

Reordering through a known laboratory makes sense when the result will drive a clinical decision, after reviewing the indication and the original laboratory.

Does FDA authorization of a home test mean results can be managed without professional follow-up?

The FDA advises users of the authorized home sexual health test to seek care after a positive result. Users with symptoms or recent exposure are advised to contact a provider even after a negative result.

Which patients most need a staffed alternative to digital self-service?

Patients with low digital literacy, limited internet access, complex conditions, or a preference for human help benefit most from staffed phone and in-person options.

The bottom line

Patients use self-serve health care for convenience and because traditional care can be slow, costly, uncomfortable, or difficult to reach. Clinical interpretation, escalation, and follow-up still need to remain available when a task goes beyond routine self-service. Patients can gain more control over their care without becoming solely responsible for interpreting findings, coordinating services, or getting follow-up.

Practices can review results-release workflows, patient-brought data policies, AI disclosure, and privacy guidance with clinical, compliance, and IT colleagues. Practices can also ask patients what led them to use a self-service option and look at what happened afterward, including follow-through and safety outcomes. Because the research on patient-facing AI is changing so fast, practices will need to revisit how they use these tools as the evidence evolves.