SafeHome AI is a self-guided home safety assessment built to close part of that gap. A family photographs the rooms, answers a short questionnaire about health and mobility, and receives a prioritized report: findings ranked by severity, specific recommendations with cost ranges, and a short list of what to address first.
It was built by a board-certified geriatric physical therapist, and the clinical logic is mine.
Where this sits relative to what you already do
Capacity and environment are two separate levers, and they operate on different timelines.
Strength training is the intervention that changes what a person can tolerate. Nothing here substitutes for it, and under-challenging older adults does more damage than any throw rug ever has.
But strength takes months, and the patient goes home Thursday. Most older adults never see a physical therapist at all, and the ones who do see us for a handful of visits. The house is there every day either way.
This is the lever families can act on immediately, without us, while the training does its slower work.
What the evidence actually supports
Worth knowing precisely, because the headline is more specific than most summaries suggest.
Cochrane defines higher risk concretely: a fall in the past year, recent hospitalization, or needing support with daily activities.
That distinction is a targeting instruction, not a footnote. This is worth recommending for the patients in front of you who meet those criteria. It is not a blanket intervention, and I don't present it as one.
The same review found little or no difference in fall-related fractures, fall-related hospitalizations, or health-related quality of life. I don't claim those outcomes.
The other finding worth your attention: the OTIS trial (Cockayne et al., HTA 2021) put occupational therapists into 1,331 homes and found no benefit — largely because the recommendations were never carried out. Identification was never the bottleneck. Adherence is. That is the specific problem the prioritized, costed, "start here" report format is built to address.
Where it fits, by setting
Inpatient rehab and skilled nursing
The family photographs the home while the patient is still admitted. The team reviews the environment before discharge instead of hearing about it after the fall. This is the setting where the tool does the most work, because it answers the question we could never staff: what does the house actually look like?
Acute care and discharge planning
The patient is leaving Thursday and nobody is going to the house. The report goes in the discharge packet as something the family can act on this weekend — an ordered, costed list instead of a vague warning about the stairs. Case managers and social workers: this is for you as much as for therapy.
Outpatient
Ask about the nights. Anyone with a fall in the past year, a recent hospitalization, or help with daily activities meets the criteria where the evidence lives. Assign it as homework, then spend five minutes of a later session reviewing the report — clinical reasoning applied to the environment they actually live in.
What it can't do
- Photos can't measure. Bed and seat heights, wall blocking behind a tub surround, handrail diameter, true nighttime light levels — cameras auto-brighten and flatten depth. Every report includes an "Items to Verify In Person" section for exactly this reason.
- There is no validation study yet. The clinical logic has been stress-tested internally across complex profiles, but agreement against in-person professional assessment has not been formally measured. A validation study is the obvious next step, and I'd welcome input on the design.
- It is not a substitute for an in-person evaluation. Reports flag when an OT, PT, or CAPS-certified professional should be involved, and say so explicitly. If anything, it routes people toward you.
Recommending it takes one sentence
Tell the family to go to safehome-ai.com and photograph the rooms. That's the whole workflow.
You're not sharing patient information with anyone. The family uses the tool directly; nothing passes through you, and nothing about the patient is retained after their report is generated.
For practices, agencies, and programs
If your organization would use this at scale — reports under your own name, built into a discharge workflow, or offered to your clients as part of a home-safety program — a professional version is in development.
I'd rather design it with clinicians than guess at what you need. Tell me what would make it usable in your setting: justin@safehome-ai.com
Try it
The free assessment covers two rooms, requires no signup, and takes no payment information.
I have no affiliate arrangement, and I'm not asking you to sell anything. What I'd genuinely value is your criticism — run one on your own home or a patient's, and tell me where the reasoning is wrong, what it missed, or what it flagged that didn't matter.
Certified Aging-in-Place Specialist
13 years across outpatient, hospital, skilled nursing, and long-term care
Recently covered in Laurie Orlov's Aging and Health Technology Watch.