It is the most reasonable question anyone asks us: "I already have ChatGPT on my phone. Why would I need a separate AI for caregiving?" Here is the honest answer.
General chatbots are genuinely useful, and under the hood Aging Sidekick runs on the same class of AI — Claude, one of the leading models, running through Amazon Bedrock, an enterprise AI platform. The difference is not the engine. It is everything built around the engine: what the assistant knows about your situation, where it is allowed to get its facts, how its behavior is tested, and what happens to the information you share.
TL;DR: A general chatbot starts every conversation from a blank page and draws on the open internet. Aging Sidekick starts from a detailed profile of the person you care for, researches only vetted sources like Medicare.gov and the National Institute on Aging, cites every recommendation, is tested against a 118-scenario safety suite, and builds a working record — assessments, notes, action items — that makes each answer more specific than the last.
A general chatbot starts from zero every time
Ask a general chatbot about your mother and it knows nothing — not her mobility, not her medications routine, not that the bathroom is upstairs or that you live 500 miles away. So you type the backstory again. You forget a detail, and the advice quietly bends around the gap. Tomorrow you start over.
Aging Sidekick keeps a structured profile of the person you care for: mobility, hygiene, daily living, the home environment, social connection, and independence and support. When you ask a question, the assistant reads that whole picture first. "Should Mom be living alone?" is a different question when the assistant already knows she uses a walker, skips meals when tired, and has a neighbor who checks in on Tuesdays — and the answer comes back different, too.
The open internet is not a care plan
A general chatbot learned from the open web — government sites, but also forums, outdated articles, and marketing pages. It can also simply make things up: an agency that does not exist, a phone number that rings nowhere, a Medicare rule that was never real. The output sounds confident either way.
Our assistant does its research differently. When it needs a fact, it searches a closed list of 27 vetted sources — Medicare.gov, the National Institute on Aging, the CDC, the Eldercare Locator, AARP, and others — and cites what it found so you can click through and check. Its standing instructions forbid inventing agency names, phone numbers, or certifications. The full list and how it works are in why the assistant only cites sources you could look up yourself.
A general chatbot
Starts from a blank page, draws on the open internet, answers everything with equal confidence, and forgets you when the chat ends.
Aging Sidekick
Starts from your person's real profile, researches only vetted sources with citations, declines the questions it should not answer, and adds every assessment and plan to a record that sharpens the next answer.
It is tested for this job specifically
General chatbots are tested broadly — coding, essays, trivia. Ours is tested narrowly and deeply on caregiving. Before changes ship, the assistant runs a 118-scenario evaluation suite: everyday questions, messy and incomplete records, boundary-pushing medical questions it must decline, and emergencies it must route to a human. In the most recent full run, it caught every scenario that called for a safety escalation, with zero misses. The methodology is public in how we test the AI before it talks to your family.
It builds a record, not a chat log
Every assessment you complete, note you save, and action item you check off becomes part of your working record. The assistant reads that record on every question, which is why families tell us the answers feel less generic in week three than in week one. To be precise about what that means: your information is used as context for your own answers. It is never used to train the AI itself — nothing you share makes any model smarter for anyone else.
Where your information lives, what the AI sees, and what we will never do with it are spelled out in where your family's information actually goes.
When a general chatbot is the right tool
We are not going to pretend otherwise: for a quick definition, a draft of an awkward email to a sibling, or a recipe Mom might actually eat, a general chatbot is great. Many of our users keep both. The difference shows up when the question is about your person and the answer needs to be right — who to call, what Medicare actually covers, whether tonight is safe.
Ask it something real
Create a free profile for the person you care about and ask the question that has been sitting in your chest. Check the citations yourself.
Start free