Fifteen out of sixteen said it helped. Read the other part.

Researchers built a chatbot named Carey for Alzheimer's and dementia caregivers — GPT-4o under the hood — and walked sixteen of them through it across eight real caregiving scenarios. Fifteen of the sixteen called it helpful or very helpful. That number is easy to quote, and it isn't nothing: an on-demand, judgment-free place to ask a question at 2 a.m. is worth something when no one else is awake.

The more useful part of this study — a preprint, not yet peer-reviewed — is where it says the chatbot fell short, because the gaps line up with the questions that scare caregivers most. They didn't fully trust the information, worried about "dirty data," and wanted to check it against a doctor or an original source. There was no clear way to escalate a crisis. And it was unclear where their loved one's private details would end up.

That maps cleanly onto Tugboat's standing line: an AI tool is a good first move, not the last word. The value isn't in impersonating the doctor, the lawyer, or the crisis line. It's in getting an overwhelmed caregiver, calmly and quickly, to the next real step.