Read the websites of six senior living communities in the same market and you will start to notice something. They use the same words. Nestled in a vibrant community. Where every day brings new possibilities. A place to call home. Compassionate care tailored to your loved one’s unique needs.
Some of that predates the current tools. Our industry has always leaned on a shared vocabulary. But the volume has changed, and the reason is not a mystery. When a marketing coordinator asks a chatbot to write an assisted living page, it produces a confident, grammatical, entirely average paragraph, because average is what the model was built to produce. Do that across six communities and you get six pages that could be swapped without anyone noticing.
We use these tools too
This is not an argument against AI. We use it daily, and we have written about where it belongs. It is genuinely good at volume work: first drafts, subject line variants, reformatting, scheduling, summarizing a month of performance data into something a regional director can read on a phone.
What it cannot do is know your building. It has never met your executive director, never eaten in your dining room, and has no idea that the thing families actually remember about your community is the courtyard where residents sit out at four in the afternoon. Ask it to fill that gap and it will fill it with something plausible, which is the whole problem. Plausible is not specific, and specific is the only thing that separates you from the community down the road.
Sameness is a pricing problem
When a family cannot tell two communities apart on anything but price and location, they will decide on price and location. That is not a branding inconvenience. It is margin. Every generic sentence on your site quietly pushes the conversation toward the one comparison you cannot win at scale.
The communities holding their rates in a competitive market are usually the ones a family can describe to a sibling in one sentence. Not the ones with the most polished copy. The ones with a reason to exist that survives being repeated secondhand.
Families can tell, even when they cannot name it
You do not need a reader to consciously identify text as machine-written for it to cost you. Copy that says nothing specific reads as a place with nothing specific to say. A daughter comparing three websites at eleven at night is not running a detection tool. She is deciding which one feels like a real place with real people in it, and she is doing it fast.
The tell is usually not a word. It is the absence of anything only your community could have written.
What we do instead
The fix is not to ban the tools. It is to stop asking them to supply the part only you have. Before anything gets drafted, we collect the material a model cannot invent: what your executive director says in the first five minutes of a tour, the objection your sales team hears most, the three things residents' families mention in reviews without being asked, the story of how the community got its name, what the neighborhood was before it was this.
Then the tools do what they are good at. Shape it, tighten it, produce the five variants we need for testing. The judgment stays with people who have walked the building, because that is the part that was never the model’s to give.
AI did not make marketing generic. Skipping the part that required knowing something did.