By
Stuart Kerr, Technology Correspondent,
LiveAIWire
Nearly one in three patients in the United States now uses
generative AI tools such as ChatGPT and Gemini when searching for doctors,
according to survey data reported by TechTarget.
That figure, drawn from research conducted across thousands of US adults,
represents a shift in how people begin their healthcare journey that is
happening faster than the healthcare system has prepared for. In a landscape
of rising costs, fragmented insurance networks, and opaque provider
directories, patients are turning to AI to cut through the noise and make
faster decisions about who to see.
The appeal is concrete. Patients use conversational prompts to
filter doctors by location, specialty, insurance coverage, and patient
reviews in a single exchange. AI can consolidate multiple sources in seconds,
compared to the manual sifting required by traditional search engines that
return a mix of paid placements, outdated directory listings, and review
aggregators of variable reliability. For anyone who has tried to find a new
GP in a new city, or a specialist in a narrow discipline accepted by a
specific insurer, the friction that AI can reduce is real and
familiar.
What This Means for You Right Now
If you are already using AI to find healthcare providers, the
practical upside is genuine but so are the risks you need to understand
before acting on what you find. AI tools can summarise publicly available
information about providers, surface common patient complaints from review
sites, explain what different specialties treat, and estimate waiting times
based on information in their training data. What they cannot reliably do is
verify that a doctor is currently accepting new patients, confirm that their
insurance affiliations are accurate as of today, or flag information that has
changed since their training data was compiled.
Healthcare directories are notoriously out of date. Doctors move
practices, retire, change insurance relationships, and shift clinical focus.
An AI system trained on web data inherits those inaccuracies without any
mechanism to signal when its information is stale. Patients who act on AI provider
recommendations without independent verification may find themselves calling
a practice that no longer exists at the listed address or discovering at the
appointment stage that their insurer is not accepted. The efficiency gains of
AI-assisted search are real; so is the failure mode when the underlying
information is wrong.
Trust and Reliability Concerns
An Ipsos
survey found that 31 percent of consumers already use generative AI
for healthcare queries, with many expressing concern about accuracy. That
concern is well-founded. Large language models can generate hallucinations,
responses that sound authoritative but are not grounded in verifiable fact.
In a domain where acting on wrong information can delay necessary care or
misplace confidence in an unsuitable provider, the stakes of that failure
mode are higher than in most other contexts where people use AI
casually.
A study from the Annenberg
Public Policy Center found that 75 percent of respondents
considered AI-generated health information useful, and 63 percent judged it
reliable. Yet a majority still double-checked with a doctor before making
decisions based on it. That pattern, finding AI a useful starting point while
retaining professional verification as a final step, reflects a pragmatic
approach to a technology still prone to confident error. It is also,
incidentally, the approach that produces the best outcomes. AI as a research
tool, human judgement as the decision point.
The Promise and the Pitfalls
Trust varies substantially by demographic. Younger patients, who
have grown up navigating digital health tools, telemedicine platforms, and
health apps, are considerably more comfortable acting on AI recommendations
than older patients or those managing complex chronic conditions who have
deeper experience of what goes wrong when health information is incomplete or
wrong. A BMJ report examining how patients integrate AI into primary care
decisions raised concerns about uneven adoption, noting that populations most
likely to benefit from improved provider discovery are also those least
likely to access AI tools confidently. The efficiency gains of AI-assisted
healthcare search may accrue primarily to the already digitally
confident.
Bias in AI health guidance is a related concern that researchers
have raised consistently. Models trained on web data that underrepresents
certain demographic groups, health conditions more prevalent in specific
communities, or specialists serving underserved populations will produce
recommendations that reflect those gaps. An AI system optimised for the
average user may systematically underserve the patients whose healthcare
navigation is most difficult and whose access to good information is most
limited.
Economic and Social Drivers
The shift toward AI-assisted provider search is tied to rising
economic pressure on patients as much as technological capability. The time
cost of navigating fragmented healthcare information is not trivial. Between
insurance portal interfaces, provider directory sites of variable accuracy,
review aggregators, and telephone calls to practices to verify what is listed
online, finding the right doctor through conventional means can take hours.
AI compresses that process in a way that patients recognise as genuinely
useful, not simply as a novelty.
For healthcare providers, the implication is that their digital
presence matters more than it once did. AI tools pull from publicly available
information, which means practices with well-maintained, accurate, and
detailed listings across directories and review platforms are more likely to
appear in AI-generated recommendations. Practices that have not audited and
updated their digital information in recent years may find themselves
underrepresented in AI responses regardless of the quality of care they
provide. This is a new competitive dimension in patient acquisition that most
practices have not yet recognised as requiring active management.
What Regulation Is and Is Not Doing
The regulatory framework governing AI use in healthcare is still
catching up with how patients are actually using these tools. The FDA has
cleared hundreds of AI-enabled medical devices for clinical applications, but
casual use of general-purpose AI tools by patients to navigate provider
selection sits outside most existing regulatory frameworks. There is no
requirement for an AI tool to disclose when its healthcare information may be
inaccurate, outdated, or drawn from sources of variable reliability. A
patient using ChatGPT to find a cardiologist has no equivalent of the
professional duty of care that would apply if a nurse or doctor were
providing the same guidance.
The practical consequence is that the accountability framework
around AI-assisted provider search is thin. The information may be useful,
but there is no mechanism for patients to know when it is wrong, no recourse
when it leads them astray, and no obligation on AI providers to ensure their
tools meet any standard of accuracy for health-related queries. As AI use in
healthcare navigation grows, that gap will need to close. Until it does, AI
remains a useful starting point for provider search, not a sufficient
one.
For a broader look at how AI is reshaping patient care, see our
coverage of what
your smartphone now knows about your health, the role of AI in
detecting
mental health conditions before doctors can, and the implications
of AI-driven
insurance decisions for patient access to care.
How Healthcare Providers Should Respond
Healthcare organisations that have not considered how AI tools
represent them to potential patients are operating with a blind spot that
will become more significant as AI-assisted provider search grows. Unlike
traditional search engine optimisation, which focuses on keyword ranking, AI
responses draw on the quality, accuracy, and consistency of information
across multiple sources. A practice with accurate, detailed, and consistent
listings on multiple platforms is more likely to appear in AI responses than
one whose information is inconsistently maintained across
directories.
The proactive steps are not technically complex. Auditing and
updating provider listings across major healthcare directories, ensuring that
insurance affiliations are accurate and current, maintaining active responses
to patient reviews, and keeping website information current are all practices
that improve AI representation as a direct consequence of improving
information quality for all discovery channels. The organisations most likely
to benefit from AI-assisted patient discovery are those already investing in
information quality rather than those making changes specifically in response
to AI.
Longer term, the question of how AI tools integrate with
healthcare systems will be shaped by decisions being made now at the
intersection of technology policy, healthcare regulation, and patient
advocacy. Whether patients using AI for provider search receive tools that
are accountable, accurate, and aligned with their interests rather than with
the commercial interests of the platforms providing those tools is not a
technical question. It is a governance question, and the answers will be
determined by who is at the table when they are made.
About the Author
Stuart Kerr is Technology Correspondent at LiveAIWire, covering
artificial intelligence, cybersecurity, and the social impact of emerging
technology. He publishes daily at LiveAIWire.com.