On November 3, 2026, we searched for assisted living and memory care in six towns on Google and ChatGPT. Google showed an AI Overview on only 3 of 12 searches, while a map pack of Google Business Profiles appeared on all 12. ChatGPT searched the web for 11 of its 12 answers and named 72 communities, between 4 and 8 per answer. Its visible sources were mostly referral sites and directories: 25 of them, against 12 links to communities’ own websites and 9 to other sites. For the plain assisted living question, it linked no community’s own website at all. Google’s three overviews went the other way, with 8 of their 12 sources on operators’ own sites.
For a community, that means its Google Business Profile and its directory listings showed up in these answers more often than its own website, which was cited mainly in ChatGPT’s memory care answers and in Google’s few overviews.
AI answers change from one day and one person to the next, so read these numbers as a snapshot of one afternoon, not a ranking.
How we ran the test
We picked six towns in the states where we work, each with a range of assisted living and memory care communities: Asheville, North Carolina; Savannah, Georgia; Gainesville, Florida; Richmond, Virginia; Plano, Texas; and Cherry Hill, New Jersey. In each town we ran four prompts, 24 in all, between 2:58 and 3:22 pm Eastern on November 3, 2026:
| ID | Where | Prompt |
|---|---|---|
| G1 | Google Search | assisted living in [town], [state] |
| G2 | Google Search | best memory care in [town], [state] |
| C1 | ChatGPT | What are the best assisted living communities in [town], [state]? |
| C2 | ChatGPT | My mother has dementia. Which memory care communities in [town], [state] should we look at? |
Every prompt ran once, signed out, in a new session, with no follow-up question. We ran them from Toronto, with the town and state in every prompt and Google set to US results. On Google we waited about eight seconds for an AI Overview to load, then recorded the overview, its sources and the map pack. On ChatGPT we recorded whether it searched the web, the communities it named and the sources it showed. A source chip such as “A Place for Mom +1” counts as one source, because the hidden extra sources were not opened.
What the answers named
| Town | AI Overview (G1, G2) | Communities in Google’s overviews | Communities named by ChatGPT (both prompts) | Map pack communities ChatGPT also named |
|---|---|---|---|---|
| Asheville, NC | No, no | None | 7 | 3 of 5 |
| Savannah, GA | No, no | None | 9 | 5 of 5 |
| Gainesville, FL | No, yes | 6 | 9 | 3 of 3 |
| Richmond, VA | No, yes | 5 | 9 | 3 of 4 |
| Plano, TX | No, no | None | 11 | 2 of 4 |
| Cherry Hill, NJ | Yes, no | 6 | 5 | 5 of 5 |
Google kept its AI Overviews for a few searches. The two that appeared for “best memory care” were written as short lists of communities with a line about each. The Cherry Hill overview was built differently: a map with a card for each community showing its Google rating and number of reviews, and the first three cards were the same three communities, in the same order, as the map pack below it. Both memory care overviews ended by sending the reader to referral and ranking sites. The Gainesville one closed with: “You can explore detailed facility comparisons, reviews, and services on A Place for Mom Gainesville or check regional ratings via U.S. News Memory Care in Gainesville.”
ChatGPT gave a shortlist every time. Its lists overlapped with Google’s local results: of the 26 communities recorded in the map packs for the six towns, ChatGPT named 21 somewhere in its answers for the same town. It also framed its lists with care. In Savannah it said: “This is a research shortlist, not an official ranking.” The dementia prompt changed the tone: the five memory care answers whose opening we copied word for word all began with sympathy for the family. In Asheville it opened with “I’m sorry your family is going through this,” and went on: “I’d focus on communities that offer dedicated dementia and Alzheimer’s care, rather than general assisted living that happens to accept residents with memory loss.”
Where the answers got their information
| Source | Google’s AI Overviews (3) | ChatGPT, assisted living prompt (6) | ChatGPT, memory care prompt (6) |
|---|---|---|---|
| Referral sites and directories (such as A Place for Mom, Caring.com, CareScout, Seniorly, SeniorLiving.org, U.S. News) | 4 | 19 | 6 |
| Communities’ and operators’ own websites | 8 | 0 | 12 |
| Other sites (state licensing agencies, Google, a nonprofit) | 0 | 2 | 7 |
The split depended on the question. When asked for the best assisted living communities, ChatGPT drew its visible sources from directories and senior care sites in every town; in Plano, all seven source chips were directories or senior care blogs. When asked about memory care for a mother with dementia, it linked communities’ and operators’ own websites 12 times, in Asheville under the label “Official memory care information.” Three of its answers also pointed families to state licensing records, and in Asheville to a nonprofit memory care center.
Google’s overviews leaned on operators’ own sites. In Cherry Hill, all five source chips were the communities’ or operators’ websites; in Gainesville and Richmond, the sources mixed operator sites with A Place for Mom, AssistedLiving.org, U.S. News and Seniors Guide.
Google describes how these answers are put together. Its Search Central page on AI features says AI Overviews and AI Mode may use “query fan-out,” “issuing multiple related searches across subtopics and data sources” to build a response. To be shown as a supporting link, a page must be “indexed and eligible to be shown in Google Search with a snippet,” and Google says “there are no additional requirements to appear in AI Overviews or AI Mode, nor other special optimizations necessary.” Its advice to site owners is the usual groundwork: helpful content in text form, structured data that matches the visible text, internal links, pages Google can crawl, and Business Profile information kept current.
ChatGPT’s search draws on the web through OpenAI’s crawler for search, OAI-SearchBot. OpenAI’s documentation says “OAI-SearchBot is used to surface websites in search results in ChatGPT’s search features,” and that “sites that are opted out of OAI-SearchBot will not be shown in ChatGPT search answers.” A community website that blocks it in robots.txt, or behind a firewall rule, can’t be cited there.
What this means for a community
We expected the AI Overview to be the main event on Google and referral sites to dominate everywhere. The data showed something more mixed.
- The map pack still carries Google. It appeared on every search, the AI Overview on only a quarter of them, and the one overview built as a map repeated the map pack’s order. ChatGPT named 21 of the 26 map pack communities. A complete Google Business Profile, with the right categories and a steady flow of reviews, works for both.
- Directories feed ChatGPT’s general answers. For “best assisted living,” every visible source was a directory, a senior care site or a state agency. What A Place for Mom, Caring.com, CareScout and Seniorly say about your communities was among what ChatGPT read here, so it is worth checking those listings for accuracy, even while you work to rely less on referral fees.
- Your own memory care page can be cited. For the dementia question, ChatGPT linked communities’ and operators’ own websites 12 times. A clear memory care page for each community gives it something to link.
- Google’s overviews cite operators when they appear. 8 of their 12 sources were communities’ or operators’ own websites.
Two checks take a few minutes and come straight from the documentation above: make sure your website’s robots.txt and firewall let Googlebot and OAI-SearchBot through, and make sure your Business Profile’s name, address, hours and care types match your website. Our senior living SEO and AI answers page explains how we work on the rest, and our free audit shows, community by community, which of these answers name you today. You can request it on our senior living page.
Limits of this test
One run per prompt is a snapshot. AI answers vary with the wording of the question, the day, the person asking and where they are, and our searches ran from one place with the town named in the prompt. Signed-in users with a search history may see different answers. We didn’t test Gemini, Perplexity or Google’s AI Mode. Hidden sources behind “+N” chips were not opened, so the source counts cover visible sources only. In one case, memory care in Plano, ChatGPT answered without searching the web and linked only the Texas Health and Human Services site; that answer shared one community with the Plano answer that did search.
Questions operators ask about AI answers
Can we pay to appear in Google’s AI Overview or ChatGPT’s answer?
Google says there is no special markup or optimization needed to appear in AI Overviews; a page has to be indexed and eligible for a snippet. In our test, ads appeared on some Google results pages above the map pack, never inside an AI Overview.
Do AI answers use our Google Business Profile?
Google’s advice for AI features includes keeping Business Profile information current. In our test, the Cherry Hill overview showed each community’s Google rating and review count on a card, in the same order as the map pack.
How do we know which AI answers name our communities?
Run the same prompts for each of your towns and care types, signed out, and record the answers on a set date each month, as we did here. Our free audit does this for each of your communities.
Is our older article on this topic still right?
Our 2025 guide, AIO for LTC, explains the difference between AI search and generative AI and how each draws on your content. This test adds what the answers actually said in six towns on one day.
Sources
- Google Search Central, AI features and your website, read November 2026.
- OpenAI, Overview of OpenAI crawlers, read November 2026.
- Our test: 24 prompts in six towns, run November 3, 2026, 2:58 to 3:22 pm Eastern. The answers are kept on file as text.
Disclaimer: We are a marketing company, not a legal one. Nothing in this article constitutes legal advice; all information, content, and materials available on this site are for general informational purposes only.
Roman ZelvenschiChief Storyteller
Roman Zelvenschi is a digital marketing strategist and founder with over 15 years of experience helping businesses grow through data-driven marketing and brand strategy. Since 2008, he has turned ideas into profitable, scalable ventures—working with brands ranging from global enterprises like Mondelēz to startups such as The Kosher Express. His expertise lies in SEO, AI-driven marketing, and lead generation, particularly within highly regulated industries such as healthcare, long-term care, and finance. Roman believes effective marketing blends brand investment for long-term growth with short-term performance through lead generation and customer lifetime value optimization. An early adopter of AI in marketing, Roman has leveraged machine learning since 2022 to enhance targeting, personalization, and campaign performance. He continues to advocate that technology should amplify human creativity—not replace it.
Schedule a Call
