What ChatGPT tells families about your funeral home

Run the audit yourself

Open ChatGPT and ask, plainly: “Tell me about [your funeral home] in [your town].” Then ask the harder ones — “Is it well regarded?” “What do families say?” “Who else should I consider?” Read the answers as a stranger would, because that is who will read them next.

Earlier posts had you check who gets recommended when a family asks an assistant which funeral home to call. This is the sharper, more personal audit: what the machine believes about you, by name. The two can diverge — a funeral home can be recommended on location and still be described from a thin, stale record — and the description is what a family reads in the minute before they decide whether the recommendation was right.

Treat it like a file you are opening, because it is one. Put the same questions to Gemini and Perplexity too, copy each answer into a document with the date, and resist the urge to argue with any of it while you read. The exercise is not to feel judged; it is to see your record the way every future family will: without your context, without your history, with only what the machines could find.

What you are looking for

Three kinds of finding matter. Absences: you are not named for questions you should own — that is a visibility gap. Errors: a wrong address, a closed location still listed, a predecessor’s name — that is a consistency gap, and machines inherit it from stale directories. And framing: if the answer leads with the consolidator’s marketing language while describing you in a shrug, your public record is thin where theirs is thick.

Each one looks different on the page. An absence reads as silence where your name belongs — asked for cremation providers in your own town, the assistant lists two competitors and a crematory. An error reads as confident wrongness: “located on Oak Street” when you moved in 2021, “founded by Harold Beck” when the Becks sold it a decade ago — each one inherited from some directory nobody has touched in years. And errors are the norm, not the exception: when the analytics firm Searchable queried ChatGPT, Perplexity, and Gemini more than 13,000 times about real companies, small businesses came back with at least one fabricated fact half the time — 50%, against 32% for big brands. Framing is the subtlest: the corporate-owned funeral home across town gets “known for its modern facilities and comprehensive pre-planning services” — language lifted straight from its own busy website — while you get “a funeral home in Fairview.” Both descriptions are technically accurate. Only one would reassure a stranger.

Corrections that actually propagate

You cannot email a model. You correct the sources it reads: your own site’s plain-language pages, your Google Business Profile, the big directories, your obituary page’s steady pulse. And corrections do propagate, because the answer is not stored — ChatGPT builds a local answer by searching Bing’s index and the open web live, so once the crawlers re-read a fixed source, the fixed fact is what the next family gets.

Match the fix to the finding. An absence is corrected by giving the machines something to find — pages that answer the questions you were absent from, in full sentences, on your own domain. An error is corrected at its source: hunt down the stale directory or old listing repeating it, fix it there, and make sure your site and profile state the current fact plainly enough to outweigh the residue. Framing is corrected by writing the description you want borrowed — a history page with real dates and names, a staff page with real people, services described in the words a neighbor would use. Machines assemble descriptions from whatever text exists; a home that has never written its own story in plain language has, in effect, asked to be described in a shrug.

Make it a quarterly habit

The answer changes as models update. A quarterly read — same questions, noted answers — tells you whether your record is improving, and it costs nothing but the asking. Few competitors are doing even that.

Keep the transcripts in one folder, dated, and compare each new read against the last: errors gone, absences filled, framing warmed. That trend line is the truest marketing dashboard a small funeral home can own — it measures the exact thing a family sees, and every improvement in it was work you did once that keeps paying.