Building the science behind accessible care.
Our work sits at the intersection of computer vision, public health, and dentistry. We're rigorous about what AI can and cannot do — and transparent about both.
Our approach
SmileSafe AI uses computer vision models trained on dental imagery to provide preliminary screening signals — not diagnoses. Every claim we make about accuracy is bounded by what photo-based screening can actually accomplish: it can flag visible patterns, but it cannot replace clinical examination, X-rays, or a dentist's professional judgment.
We use Anthropic Claude as the foundation model behind our scan analysis. We supplement it with our own structured prompts and post-processing to produce consistent, plain-language reports across English, Spanish, Ukrainian, and Russian.
Important. SmileSafe AI is not currently FDA-cleared as a medical device. We position our scan as a health and wellness screening tool, similar to many consumer apps that flag possible concerns and recommend professional follow-up. See our Medical Disclaimer for full context.
What we research
Our research focus is shaped by the populations we serve — immigrant families, elderly patients, low-income communities — and the gaps they encounter in dental care access.
Screening accuracy
How well does photo-based AI screening identify early signs of dental concerns compared to in-person examination? We're collecting paired data (kiosk scan + dentist exam) to measure sensitivity and specificity.
Access & outcomes
Does kiosk screening actually move people from "I should probably see someone" to "I have an appointment"? We track conversion rates and time-to-care across host locations.
Multilingual equity
How does AI screening perform across different languages and cultural contexts? Where do reports succeed at reducing barriers, and where do they fall short?
Workflow for dentists
Do AI-pre-screened patients arrive at appointments better prepared? Do dentists save time? We measure efficiency gains and patient satisfaction in partner practices.
Validation status
We're early in our public health research. As of 2026:
- Pilot deployments are underway in select pharmacies and community spaces in the Austin, TX area
- Internal validation studies compare AI scan flags against dentist findings on a curated dataset
- Partner studies with academic institutions are in active discussion
- Peer-reviewed publications are not yet available — we'll list them here as they appear
Open questions we'd love to study
If you're a researcher in dentistry, public health, or AI evaluation, we'd love to hear from you. Some questions we're particularly interested in:
- How does kiosk-based screening compare to traditional outreach methods in reaching underserved populations?
- Does the visual feedback in a scan report change patient self-care behavior over 6–12 months?
- What's the right calibration between AI sensitivity and false positive rate for a screening tool that funnels to professional care?
- How do multilingual reports affect family decision-making in households where dental care decisions are made jointly?
Data and ethics
All research is conducted with informed consent and in compliance with HIPAA and applicable data protection laws. We do not sell or share identifiable data. De-identified, aggregated data may be used for product improvement and research publications, with appropriate IRB review when required.
Our research practices are reviewed regularly by external advisors with backgrounds in dental medicine, AI ethics, and public health.
Working in this space?
If you're a researcher, public health practitioner, or institution interested in collaborating, reach out at viktoriia.f@smilesafeai.com. We're especially interested in partners who serve immigrant or low-income communities.