We asked an AI assistant the two questions a family asks when someone has died and they do not have a funeral director yet: "funeral director in [town]" and "cheapest cremation [town]". We asked in nine New Zealand markets, then again for five funeral homes in Victoria, Australia. Eighteen runs, recorded word for word.
This is what happened, and what it means for your funeral home's website.
The assistant shortlists two or three firms, by name
In every market the answer was a short list. Two or three funeral homes named, with a line about each, sometimes a phone number, and a suggestion to call and ask for an itemised estimate. Not ten blue links. Not "here are some options". A shortlist, with names.
If your funeral home is on that list, a family who would once have searched, scrolled and compared now has your name in front of them before they have opened a browser tab. If it is not, they may never see you.
The list is not fixed. We saw the same question produce a different order, and occasionally a different third name, on a different day. Assistants sample and vary. But the firms that appeared reliably had something in common, and it was not size.
What the assistant went looking for
The part that surprised us most was visible in the assistant's working. Before it answered, it ran its own searches, several of them, and we captured every one. It looked for:
- a funeral director in the area, by name and by suburb
- prices: "cremation cost [town]", "funeral prices [firm name]"
- whether the firm was open now, and a phone number
- the firm's services, and sometimes reviews
Then it read the pages it found and built the answer from them. When it could not find a price on a firm's website, it said so, in the answer a family reads: "prices are not published, contact them for an itemised estimate". Every one of the eighteen runs closed that way for at least one firm.
Why most funeral websites could not help it
We then read the websites the way the assistant did. Three gaps came up again and again.
Prices in a PDF, or not at all. The single biggest one. Many funeral homes publish a price list as a PDF. A PDF is a document, not a page: the assistant cannot quote a line from it the way it quotes a sentence from a web page, and in our runs a PDF price list never once made it into an answer as a figure. Firms with prices on an ordinary page were named with their prices. We saw assistants build comparison tables from published prices. That is the future arriving.
No machine-readable statement of who you are. Websites talk to people. Assistants also read structured data, a short block of code that says: this is a funeral home, at this address, serving these areas, founded in this year, a member of these associations. Most funeral websites have none, or have the generic "local business" kind. There is a wrinkle here we will write about separately: there is no FuneralHome type in the vocabulary machines use, so even careful sites often get it wrong.
Name inconsistency. One firm called itself three different things across its own pages. A family would not notice. An assistant trying to match "the firm on Main Street" to a name in its shortlist might.
What did not seem to matter
Some things we expected to matter did not, at least in this sample. Design quality: a plain, older-looking site with prices on a page did better than a beautiful one without. Size: single-location family firms were named as often as groups. Reviews on third-party sites: present for most, and not what separated named from unnamed.
One thing that did separate them in the Victorian sample, and we hold this loosely because five firms is five firms: every named firm stated its founding year on its site. The one firm that was not named did not.
What to do about it, in order
- Put your prices on a page. Not a PDF. A page with headings a family can read and a machine can quote. If you want families to build their own estimate, a tool like Funeral Estimator Pro does that, but a plain page is a fine start.
- Say who you are, in structured data. The right schema type, every location, your service area, founding date and associations. This took one afternoon on one of our client sites and moved its audit score from 61 to 79 (the Elliotts case study).
- Make the basics findable. A phone number that works at 2am, clearly marked. Your services in words. Your notices indexed.
- Check, then keep checking. Answers change. Run a check now, run it again in a month.
Check your own funeral home
We built the checks above into a free audit. It reads your site the way the assistant does, runs the same two questions for your area, and shows you every search the assistant ran beside whether your website could answer it. It takes about a minute and there is no sign-up. Run the free visibility audit.
If the result raises questions, that is what our SEO and AI visibility service is for. And if you would rather fix it yourself, everything above is enough to start.
Method note: eighteen ChatGPT runs, July and August 2026, nine New Zealand markets and five Victorian firms, each answer captured verbatim with the assistant's own searches. Answers vary between runs and we say so. We never say a site lacks something; we say we could not find it on the pages we checked.
