We checked the websites of 6,678 UK care homes, drawn from the registers of all four UK regulators.* One in nine has no website we could find. Another one in ten has an address that leads nowhere, so one home in five has nothing a family could open at all, and among small operators it is one in four. Of the homes that do have a working site, the typical one answers about six tenths of what a family needs and leaves the rest out.
Here is the part operators and marketing leads should not ignore. None of that is a measure of care. Our study says nothing about what happens inside your home, and that is exactly the problem. A home with an Outstanding inspection and a website that hides its fees is a good home that an assistant cannot recommend. The badge families trust most does not travel into the answer, because the answer is built from what a machine can read on your page.
Quality of care and AI visibility are different jobs. You can win the first and still lose the shortlist when a daughter in Dublin or a son in Leeds opens ChatGPT, Perplexity, or Google’s AI Overviews and asks for a nursing home that takes dementia residents, explains Fair Deal funding, or is near Mum.
The full numbers, the method, and the honest limits are in the study write-up: what we found across 6,678 UK care homes.
If you would rather see the short version, with the three jobs to do and where the sign-off line sits, it is on one page here.
See how to position your care home for AI search
This piece is for UK and Ireland care home and nursing home operators, CEOs, general managers, and heads of marketing. Not for families. It maps the problem to the Get Recommended Framework (GRF): six steps, in order, adapted to how care decisions actually get made under stress.
Findable is not the same as recommended
For twenty years the job was to be findable. Ten blue links. Fight for a ranking. You got good at that.
A lot of buyers are no longer starting there. They ask an assistant which home to shortlist, or they read Google’s AI answer and stop. The machine names two or three providers. Everyone else is not in the room.
You can sit high for “care homes near [town]” and never get named once. Rankings still matter. Recommendations decide. Our own numbers show where it breaks. Only 41% of sites have any page or section about fees or funding, and just 4.3%, about one in 25, show an actual weekly price. Only one in five carries the structured data that tells a machine what kind of organisation this is. Blocking the AI crawlers is not the problem, whatever you may have read: fewer than 2% of care home sites do it, and not one large operator does. The crawlers get in. They find nothing worth quoting. When an assistant does answer, it often leans on government pages, regulators, and directories before it trusts a private brand.
Use AI to research. Humans own the claims.
Search engines and AI tools hold pages about health and money to a higher standard. Care is both. Fees, care descriptions, funding rules, and inspection claims are not the place to experiment.
Use AI to research the questions families ask, cluster them, and structure pages so answers sit up front. Humans own the wording on fees, clinical descriptions, CQC or HIQA status, and anything that could mislead under stress. That is the anti-penalty rule in one line: accelerate the work, do not fake the expertise.
The six steps, adapted to care homes
Most teams jump to step 3 and start publishing. Then nothing changes, and content gets blamed. Order matters. Each step builds on the one before it. A weak step 2 quietly cancels out everything after it.
1. Know your buyer
Not a persona deck. The exact questions families type into ChatGPT or Perplexity when they are tired, scared, and short on time.
Typical clusters:
- Compare: residential vs nursing, dementia specialist vs general, self-funded vs local authority or HSE support
- Best for: late-stage dementia, respite, couples, rural vs town, Irish Fair Deal pathways
- How to choose: what to ask on a visit, how fees work, what CQC or HIQA ratings actually mean, waiting lists
Paste your offer and catchment into a model and ask for the twenty questions families ask before they contact anyone. Cluster them. Those clusters drive the rest of the framework. If you skip this, you write for keywords you like, not for prompts machines actually get asked.
2. Get recognized by AI
Can the machine say who you are and what you are best at, in one sentence a stranger could repeat?
Getting those basics to line up matters more than a new blog calendar. Your legal name, trading name, address, services, and registered manager should match across your site, CQC (England) or HIQA (Ireland), directories, and Google Business Profile. Inconsistency is how assistants refuse to commit to a name.
Practical check: ask a model “who is [your brand] and what are they best at?” with no extra context. If the answer is vague, wrong, or mixes you with another provider, fix your About page, your homepage, and the basic details about you that sit elsewhere on the web, before you publish more posts. Recognition first. Volume later.
Also check robots.txt. If you block the major AI crawlers, you are asking to be recommended while locking the door.
3. Answer-ready content
Pages a model can quote, not pages that bury the answer under stock photography and soft copy.
Lead with the direct answer in the first screen. Then proof. Then nuance. Priority topics for UK and Ireland care operators:
- Fees and what is included (and what is not)
- Dementia and nursing care scope, in plain language owned by clinical leadership
- Funding: local authority routes in the UK; Fair Deal and related pathways in Ireland
- Local realities: catchment, transport, visiting, waiting lists, respite
Helping a family decide locally is where our study shows the sector is weakest, with the typical site answering only 35% of what a family needs there. Almost half of care home websites give a family nothing at all on fees, funding, visiting, or what admission involves. Address, services, fees, and a real FAQ, tagged in the page code so an assistant can confirm them rather than guess. The content often already exists for human readers. It has simply never been made quotable.
Same rule again: if a stranger cannot find the answer in ten seconds, the model will not either. Do not invent claims about care to fill a template.
4. Trust and authority
What other people say about you, away from your own site.
AI systems cross-check. Reviews, regulator listings, local press, and consistent directory profiles carry weight that a self-published About page cannot. On-page proof was one of the stronger areas in our study, and it is still the largest groups pulling the average up. That is off-site work: genuine reviews, accurate directory entries, and occasional local coverage that names you for a specific strength.
One clear take plus one proof travels farther than a generic “we care” update. The same short answer can feed a comment, a LinkedIn post, and later a site page without sounding duplicated, as long as the judgment is yours.
5. Be where decisions are made
Families do not live on your blog. They live on carehome.co.uk, Autumna, regulator registers, GP and social worker advice, local Facebook groups, and increasingly inside AI answers that pull from those places first.
Thin blogs about “the joys of autumn” do not win recommendations. Complete, consistent profiles on the directories families and brokers already use do. So do pages that answer the funding and dementia questions those directories cannot answer for your home specifically.
Do not try to game those places. Put clear, checkable facts where the decision already happens. Then mirror the same answers on your own site.
6. Track and improve
Answers shift. Last quarter’s win decays quietly.
Once a month, run the same buyer prompts through ChatGPT, Perplexity, and Google AI Overviews for your towns and care types. Log who gets named, who does not, and what language they use. Fix the GRF step where your blanks cluster. Do not chase every surface at once.
If you are invisible today, that is information, not a verdict. Our numbers show almost nobody in the sector has claimed this properly yet. About one home in eight does this well, and the national groups are barely ahead of an owner who edits the site herself on a Sunday. The bar to become the first clearly recommended home in your catchment is still low. The window will not stay open forever.
The three jobs to do in order, and where the compliance line sits, are laid out here.
See how to position your care home for AI search
What this means for CQC Outstanding and HIQA-strong homes
If you have invested years in inspection outcomes, treat AI visibility as a separate operating discipline, not a marketing vanity project. Outstanding care that machines cannot parse will keep losing enquiries to average care that is easier to quote.
Ireland operators should run the same playbook with HIQA consistency and Fair Deal clarity in the answer layer. The platforms are the same. The funding vocabulary is local. The buyer stress is identical.
Fix one gap first
Fixing all six at once is the mistake. Gaps are not equal. Pick the step where your blanks cluster. For many homes that is step 2, where the basics of who you are do not line up across the web, or step 3, where the fees, dementia, and funding pages give a model nothing to lift. Start there.
Work through the free Get Recommended Framework workbook: six steps, eighteen questions, in the care home edition of the workbook. Upload it to your favourite AI and walk through it with your leadership team. If you want a second pair of eyes, bring the filled workbook to a short call.
One gap. Clear next move. The homes that treat recommendation as a system will get named. The rest will stay CQC Outstanding on paper and invisible where families now decide.
If you want to see what that looks like for your home, the positioning page explains the approach and has a free diagnostic call on it.
See how to position your care home for AI search
* SEOLeverage, September 2026. 6,678 UK care homes drawn from the public registers of the Care Quality Commission, the Care Inspectorate in Scotland, Care Inspectorate Wales and RQIA, filtered to services registered for older people’s care. 4,432 websites were crawled and scored against the Get Recommended Framework. Full method and limits: the full study write-up.