AI vs human: should celebrants worry about AI writing ceremonies?

Every few months a new AI tool launches that can write a wedding ceremony from a paragraph of inputs. Then the question shows up in celebrant Facebook groups: are we about to be replaced?

I'm a celebrant-in-training, and I'm also the person who built one of these tools. So I think about this question most days. Here is the honest answer.

What AI is actually good at

Modern language models are remarkably good at a specific kind of work: producing a competent first draft of a structured piece of writing, in seconds, from a handful of inputs. For ceremonies, that's real. Tell it the couple's names, the style of ceremony, the must-includes, the length — and it will produce something with the right shape: welcome, story, address, vows, rings, declaration, send-off, in the right order with sensible word counts.

That's genuinely useful. The blank page is the part of ceremony writing that most celebrants find hardest, and AI obliterates the blank page. It also has an enormous vocabulary and decent rhythm, which means it produces prose that sounds like prose rather than bullet points.

What AI cannot do

It cannot meet the family. It cannot sit at the kitchen table with the bride's father and watch his eyes when he talks about her mother who isn't here anymore. It cannot hear the exact tone in which the groom says "she's mental" about the dog and know whether to lean into the joke or steer around it. It cannot pick up that the daughter of the deceased doesn't want a single religious line in the funeral, even though the mother technically attended church.

All of that is the celebrant's actual job — the part the family is paying for. The writing is the tip of an iceberg whose body is listening, care, and judgement. AI doesn't do any of those things.

AI also cannot deliver the ceremony. It can't hold the room. It can't pause when a baby cries and pick the thread back up gracefully. It can't adjust the third sentence on the fly because the bride is more nervous than anyone expected. Delivery is half of celebrancy and AI has no access to it.

The relationship that works

The right mental model is: AI is the typing pool. The celebrant is the writer. The typing pool can produce a draft from your notes in seconds. The writer reads it, throws out half, restructures, adds the specific lines that only come from having been in the room, and brings it to the family.

This is the same relationship that exists between a ghostwriter and an author, a junior associate and a barrister, an assistant and a chef. The work that scales — first drafts, structure, vocabulary — can be delegated. The work that matters — judgement, relationship, presence — cannot be.

In practice, what this looks like for a working celebrant: you do the family meeting first, take detailed notes, then feed the bones into a tool and let it produce a draft in 30 seconds. You spend the next half-hour cutting, rewriting, personalising — but you started from page seven instead of page zero. That's the win.

What this means for fees

Some celebrants worry that if AI can produce a draft in 30 seconds, families will start expecting cheaper ceremonies. So far I see no evidence of this. Families don't pay for the words on the page. They pay for the trusted person who will stand at the front of the room and say their child's name with care. That value is unchanged by AI.

If anything, the celebrants who use AI well will be faster per ceremony, which means they can take on more ceremonies, spend longer with each family, or take longer evenings off. The tool changes the economics on the input side, not the output side.

The two real risks

That said, there are two genuine risks worth being honest about.

First: a generation of celebrants who don't learn to write.If new celebrants rely on AI from day one and never struggle with the structure themselves, their judgement about what's good and what isn't will be shallower. The craft has to be in your head before you can tell whether the draft in front of you is any good.

Second: families who try to skip the celebrant altogether. Some couples will, occasionally, try to write their own ceremony with ChatGPT and have a friend read it. They will mostly regret it. The friend will not handle the room. The script will be generically polished and specifically dead. The ceremonies that go badly in this way will become the case studies that send the next ten couples back to a proper celebrant. The market self-corrects.

Where this leaves us

AI will write more of the words on more of the pages, over time. The celebrant's job — listen, judge, write the irreplaceable specific lines, hold the room — gets harder to replace, not easier. The middle of the work shrinks; the ends grow.

The celebrants I expect to thrive in the next five years are the ones who treat AI the way a good chef treats a sous chef: useful for the prep work, indispensable for the volume, absolutely not allowed near the actual plating. The ones who either refuse to use it on principle, or who use it lazily without editing, will both lose ground — for different reasons.

Trying one

If you've never used one of these tools and want to see what a 30-second draft actually looks like, the ceremony writer on this site is free for one script a month. Run it on a hypothetical ceremony you already know well — a friend's wedding, a relative's memorial. See what comes back. Decide for yourself what it's good for and what you'd never trust it with. That's the only honest way to form a view.