By
Stuart Kerr, Technology Correspondent,
LiveAIWire

Artificial
intelligence is being deployed across healthcare, social care, banking, and
public services at a pace that assumes users are comfortable navigating
digital interfaces and confident enough to engage with AI-mediated systems
without assistance. That assumption excludes a significant portion of the
population. Adults over 65 are the fastest-growing demographic in most
high-income countries, and they are also the group least likely to have
developed the digital self-efficacy that AI-first service delivery requires.
Research published in JMIR Aging
found that older adults who engage with AI-driven health tools frequently
experience anxiety rather than the efficiency gains the tools are designed to
deliver. The AI age gap is not a temporary transition problem. It is a
structural feature of how AI systems are being designed and deployed, and it
is widening.

A study published in Nature
Humanities and Social Sciences Communications identified digital
self-efficacy, the confidence to attempt, fail, and adapt in digital
environments, as significantly lower among older adults, particularly those
without ongoing family support or access to community technology literacy
programmes. These are not individual failings or generational reluctance.
They reflect a design culture that optimises for users who already inhabit
digital environments fluently, and that treats the resulting exclusion of
older users as an acceptable externality rather than a design failure. When
AI-first interfaces replace telephone or in-person access, users who cannot
navigate those interfaces do not receive a degraded service. In many cases,
they receive no service at all.

Where the Exclusion Is Felt

The most direct consequences of the AI age gap fall in domains
where older adults have the highest service needs and the least ability to
opt out of algorithmic systems. Healthcare booking systems that have replaced
telephone access with app-based interfaces and AI triage tools exclude
patients who cannot engage with those interfaces. Benefits systems requiring
digital identity verification create barriers for older adults who lack the
documentation trail that verification systems expect from users with
continuous digital histories. Telehealth platforms whose AI components assume
real-time text interaction disadvantage users with slower processing speeds
or limited familiarity with messaging conventions.

In each of these cases, the older adult who cannot navigate the
AI-mediated system is not simply inconvenienced. They may miss appointments,
fail to claim entitlements, or receive a lower standard of care than more
digitally fluent users. The efficiency gains that AI delivers to service
providers are partly achieved by reducing the cost of serving users who
require more support, and the users who require more support are
disproportionately older. The productivity argument for AI in public services
needs to account for this distributional effect explicitly, not only the
aggregate outcome.

Care Technology and the Surveillance Question

In residential and home care settings, AI systems marketed as
safety enhancements are being deployed at scale. Smart cameras, AI-powered
fall detection, emotion-sensing interfaces, and passive monitoring systems
that identify changes in daily routine are presented as tools that extend
independence and enable earlier intervention. These claims are sometimes
justified. Medway Council’s
deployment of an AI-driven in-home monitoring system demonstrated measurable reductions in care cost escalations by
enabling earlier intervention. The benefit is real. The governance question is whether the consent
frameworks and accountability structures surrounding these systems are
adequate for their capability and reach.

Research documented in open-access Frontiers proceedings raises
the risk that older users in care technology environments are treated as data
sources rather than agents, with algorithms making decisions about reminders,
medication queries, and daily scheduling without meaningful user control or
comprehension of what is being decided or why. When decision-making authority
shifts from the older adult to an AI system embedded in their living
environment, the line between assistive technology and surveillance is
determined by how the system is governed, not by its technical architecture.
That governance question is not always being asked before deployment
proceeds. As our analysis of AI’s
hidden infrastructure and its uneven social effects found, the
costs and benefits of AI deployment do not fall evenly, and older adults in
care settings are among the populations with the least power to contest
deployments that affect them most directly.

What Inclusive Design Actually Requires

The response that the evidence supports is not slower AI adoption
in sectors serving older adults. It is better design and more robust
inclusion requirements built into AI procurement and development from the
start. This means user research conducted with older adult populations before
deployment rather than after, interface design that provides genuinely
accessible alternatives to AI-first interaction pathways rather than nominal
accessibility features that fail in practice, and sustained investment in
community-based digital literacy support that enables older adults to engage
with AI tools on their own terms rather than being managed by
them.

Regulatory frameworks in the EU AI Act and UK AI standards are
developing requirements for AI systems in high-risk domains to demonstrate
accessibility and non-discrimination. The enforcement mechanisms are still
being constructed, and the assessment criteria for adequate accessibility
provision remain contested among regulators, developers, and disability and
ageing advocacy organisations. For the older adults currently navigating
AI-mediated service systems without adequate support, those frameworks cannot
develop quickly enough. The demographic pressure of ageing populations in
high-income countries means that the AI age gap is not a niche concern. It is
a mainstream governance failure with consequences that will compound with
every year that AI-first service delivery expands faster than accessibility
requirements can keep pace.

The Design Failure Underneath

The AI age gap is, at its root, a design failure rather than a
technology failure. The systems that exclude older adults are technically
capable of serving them. They exclude them because serving them well is more
expensive than optimising for users who are already fluent in digital
environments, and because the development processes that produced those
systems did not include older adult users as primary stakeholders in
meaningful ways. User research conducted primarily with younger, digitally
fluent participants produces systems that serve that population well. The
exclusion of older users is a predictable consequence that is neither
invisible to designers nor adequately addressed in most development
processes.

Procurement frameworks for public sector AI services are beginning
to include accessibility requirements that go beyond the minimum web
accessibility standards that have governed public digital services since the
early 2000s. Whether those requirements are written with sufficient
specificity to drive genuine change in how systems are designed, rather than
nominal compliance, is the critical question. As our analysis of how
AI affects learning and development differently across populations
found, the distributional effects of AI deployment are not accidental
byproducts of technology that is otherwise neutral. They are the predictable
consequences of design choices, procurement decisions, and deployment
priorities that can be made differently if the political will and
institutional capacity to do so exist. For older adults, the question of
whether that will develops before their exclusion from AI-mediated services
becomes a public health issue is one that policy frameworks are currently
failing to answer with adequate urgency. Our coverage of how
AI affects different communities in unequal ways shows this pattern
extending across multiple domains.

About the Author

Stuart Kerr is the Technology Correspondent for LiveAIWire. He
writes about artificial intelligence, emerging technology, and the forces
reshaping work, business, and society.