1. We verify the basic facts
We start with official documentation for current features, availability, service requirements and pricing. These sources confirm what the supplier says the system does. They do not independently prove benefits.
2. We search for independent evidence
Where health, wellbeing or care outcomes are claimed, we look for peer-reviewed studies, independent evaluations, public-sector reports and credible service data. We distinguish a study of the exact product from evidence about a broader category.
3. We grade the evidence
Relatively strong
Independent controlled evidence, replicated findings or a substantial credible literature.
Promising or established
Useful independent studies, established care practice or strong evaluations with limitations.
Limited
Small studies, indirect evidence, case material or observational reports.
Mainly claims
Manufacturer evidence, testimonials or plausible mechanisms without useful independent confirmation.
4. We include failure modes
A product description explains normal operation. A care decision also needs to consider batteries, broadband, mobile coverage, false alarms, missed detections, ignored prompts, subscription changes and the availability of whoever receives an alert.
5. We assess privacy by feature
We report cameras, microphones, location tracking, cloud dependence and family dashboards separately. We do not compress these into one decorative privacy score, because a camera used deliberately for a video call is different from passive surveillance.
6. Commercial relationships stay separate
Affiliate links or sponsored visibility may eventually support the site. They cannot purchase a higher evidence grade, remove a limitation or convert overseas availability into UK availability through sheer optimism.
7. Corrections and dates
Every directory entry carries a last-checked date and source. Corrections can be sent to hello@robotcarer.co.uk. Substantive changes should be recorded in a public update log as the database grows.
Use of AI
Software may assist with structuring data, checking internal consistency and drafting plain-language summaries. Factual claims must remain traceable to sources, and automated text does not create evidence. A language model sounding confident is not a new category of clinical trial.