What AI actually finds on a funeral home's website: we measured 924 of them
Last week this site said no credible funeral-specific AI-visibility measurement existed, and that we'd rather say so than borrow a bad one. That's still true of everyone else's numbers. So we made one — and the method comes before the findings.
Last week, in our roundup of AI-visibility statistics, we wrote that no credible funeral-specific measurement existed — and that we would rather say so than borrow a bad one. We also wrote down the test any such measurement should have to pass: who measured it, on what population, with what operation, and when. And we promised that when a measurement with a published, checkable method appeared, it would be added there.
That is still the honest state of everyone else’s numbers. So we made our own. This post is the method before it is the finding, because putting the method first is the whole point.
Who measured it
We did — funeral.link. We build software for funeral homes, and we give away a free audit of exactly what this study measures. That is a conflict worth knowing before you read a number we produced: the finding sizes a problem our product sells against. It is the same disclosure we asked of others on the statistics page, and it applies to us with no discount.
What we can offer against that conflict is checkability. Every number below comes from a query we can re-run, on a method described in full on this page, with the parts we did not measure named as plainly as the parts we did.
On what population
924 American funeral homes across 40 states. They come from our own prospect directory, built by scouting Google Places — and the 924 are not a sample of that directory, they are all of it: every eligible funeral home in it with a working website, graded in a single sweep, none dropped after the fact. Of the 1,077 homes in the directory, 112 — about one in ten — have no website at all, and were skipped because there is nothing for any engine to read.
What this is not: a random draw from the 15,401 funeral homes operating in the United States (NFDA’s count). It reflects where we have scouted. Read it as a large, honest snapshot of the American industry, not a nationally weighted statistic.
With what operation
Each home’s live website and its Google Business Profile, read by the same scorer that runs our free audit. The crawler announces itself as FuneralLinkAudit/1.0, makes three polite fetches per site, and carries a link back to us; a site that turns it away simply doesn’t get read.
Six dimensions, five scored here: structured data, obituary pages, name-address-phone consistency, reviews, and site health.
When
All 924 homes on August 2, 2026, in one pass.
What we did not measure — and why the scale is out of 85
We did not ask ChatGPT, Gemini, or any other assistant about any of these homes. This study measures the foundations an AI answer is built from: the machine-readable layer, the obituary pages, the review signal, the consistency of the basic facts, whether the site responds at all. It does not measure whether an assistant actually names a given funeral home, and nothing below should be read as if it did.
You will notice the scores are out of 85 rather than 100. Our audit has a sixth dimension — the AI answer test, worth fifteen points — and it did not run in this sweep. So it scores zero out of zero and drops out of the total, instead of quietly counting as a failure. A measurement that scores what it never checked is precisely the thing we objected to on the statistics page. The odd ceiling is the receipt that we didn’t.
What we found
Out of 85, the average home scored 46.6 and the median exactly 50. The best home scored 82. The worst scored 2. Nobody scored the full 85.
| Grade | Homes | Share |
|---|---|---|
| A | 21 | 2.3% |
| B | 179 | 19.4% |
| C | 333 | 36.0% |
| D | 232 | 25.1% |
| F | 159 | 17.2% |
Two homes in ten are in good shape. Four in ten — 391 of the 924 — score a D or an F on the foundations AI answers are built from.
Before that reads as an indictment, it shouldn’t. These are working businesses whose websites were built for human visitors, at a time when human visitors were the only readers that mattered. Nothing here is negligence. The machines showed up later, and they read differently.
Where it actually breaks
| Dimension | Average | Out of | Homes scoring zero |
|---|---|---|---|
| Structured data | 6.9 | 20 | 330 (35.7%) |
| Obituary pages | 12.9 | 25 | 282 (30.5%) |
| Name / address / phone consistency | 5.8 | 10 | 194 (21.0%) |
| Reviews | 9.1 | 15 | 0 |
| Site health | 11.9 | 15 | 97 (10.5%) |
Structured data is the industry’s weakest dimension, and its most fixable. It is the block of machine-readable code that states, in a form software cannot misread, that this is a funeral home, at this address, with this phone number and these hours. More than a third of the homes we measured — 330 — have none at all, which leaves an engine to infer everything from prose. Our scorer’s own advice for it is one sentence: “Publish FuneralHome structured data (JSON-LD): name, address, phone, hours — schema.org/FuneralHome.” For most sites, that is an afternoon of work for whoever already maintains the site.
Reviews are the one signal the industry has already absorbed. Not a single home scored zero on reviews — the only dimension where that is true. Somebody told directors that Google reviews matter, and directors acted. Nobody told them about the layer underneath. That contrast is the most hopeful thing in the data: this industry moves on the machine layer when it knows the layer exists.
And the sites themselves mostly work. Site health — does the site respond, does it work on a phone — was the strongest dimension, at 79% of its available points. This is not a story about funeral homes having bad websites. It is a story about working websites that are hard for software to read. That is a different problem, with a cheaper fix, and it deserves the less insulting description.
The 116
116 homes — one in eight — have neither structured data nor a readable obituary page. No machine-readable identity, and none of the pages an engine would otherwise learn a funeral home from. The two absences compound; each makes the other more costly. A funeral home that recognizes itself in that description has found its starting point.
What a director does about this
Nothing here requires our product. Most of it is a request to whoever already runs your site, and the advice below is our scorer’s own language:
- “Publish FuneralHome structured data (JSON-LD): name, address, phone, hours” — the single change that moves the weakest dimension in the study.
- “Publish a crawlable obituaries page, linked from the homepage, with a stable URL for every obituary.” Then “keep the obituary page current — recency is a ranking signal families feel.”
- “Claim the Google Business Profile — it is the single strongest AI-visibility signal a local business has.”
- On the site itself: “Add a mobile viewport meta tag — most families arrive on phones,” and make sure the site answers requests at all. 97 of the homes we measured scored zero on site health; for them, everything else is moot until the site responds.
The limits, plainly
One point in time — August 2, 2026. A convenience census of where we have scouted, not a weighted national sample. Foundations, not citations: this study says what an assistant has to work with, not what it says back. We will re-run the sweep as the directory grows. And when we measure the answers themselves, we will publish that method too, and hold it to the same four questions.
Until then: the free audit runs this same scorer on any single funeral home — with the AI answer test included.
The FuneralHomeWebsites Team