A daughter in Manchester types a question into ChatGPT at eleven at night: which care homes near Chorlton are good for a mother with early dementia, and what do they cost? She gets three names, a sentence on each, and a suggestion to visit. Two of those homes will get a phone call this week. The third name was yours, or it was not.
This is now a normal way to start looking for care, and it is worth understanding what is happening underneath, because it is less mysterious than it looks.
The engine does not know your home
ChatGPT, Gemini, Claude and Perplexity do not keep a list of good care homes. When they are asked, they search, read what they find and summarise it. The names in the answer are the names that appeared, repeatedly and favourably, in the pages they read. The pages they read are the ones you would expect: carehome.co.uk, the CQC register, Google's own listings and reviews, local newspaper stories, directories, and your website if it says anything a machine can use.
So the question is not "how do I get ChatGPT to like me". It is "what does the public web say about my home, and is it enough to be worth quoting". That is a question with a checkable answer.
Check it yourself first
Before anyone sells you anything, spend twenty minutes asking the engines what a family would ask. Not "tell me about Fernbank Lodge", which is a question only you would type. The questions families type:
- best dementia care home near Lancaster
- respite care in Lancaster this month
- nursing home in Lancaster that takes couples
- care homes near Lancaster rated Good by CQC
Ask each engine each question. Write down who is named, in what order, and look at the sources it cites. You will see the pattern within ten questions: the same two or three homes keep appearing, and the same two or three sources keep being quoted. That is the shortlist, and that is the reading list.
What moves it
Three things, and none of them is a trick.
First, reviews on the sites the engines quote. If Gemini cites carehome.co.uk for every question about your town and you have 21 reviews there to a rival's 64, that gap is the answer. Ask the families who are happy to write one, at the right moment, and reply to every review you have, because the replies are read too.
Second, a website a machine can read. Not a beautiful one; a legible one. The care types in words, the town in words, the CQC rating stated, the fees stated or at least explained, pages that load and titles that say what the page is. Most care home sites fail this not through neglect but because they were built as brochures for humans with eyes, and an engine reads them as a machine with none.
Third, other people writing about you. A local paper story about the garden, a trade title piece about the manager's award, a listing in the council's directory. Each one is a page that says your name next to the words a family searches, written by someone who is not you. The engines weight that heavily, for the same reason a person would.
What does not move it
Paying. There is no advertising slot in a ChatGPT answer and no way to buy one, whatever the email in your inbox says. Stuffing your website with the phrase "best care home in Lancaster" forty times, which the engines read as the sign of a site with nothing else to say. Posting the same line on every directory. And a one-off effort: the answers change from week to week as the engines read new pages, so a push in March can be gone by June.
Check it on a schedule
Because the answers move, the useful thing is a record. The same questions, the same four engines, every week, with who was named and what was quoted. Over a quarter you can see whether the work is working, and when a rival starts appearing you can see which source put them there. That is a job for software, which is why bamboo does it, but the first twenty minutes are a job for you, and they are the twenty minutes that make the rest worth doing.