Well not quite. I decided that I’d try out ChatGPT and see if it was any better than me at writing a sermon. I dutifully asked it to write one for my congregation focussing on the sheep hearing the shepherd’s voice. It wasn’t bad, I thought. I added in a couple of one-liners to personalise it, so it sounded a bit more like me and off I went to church to preach it.
It felt weirdly OK preaching it. I was able to add my intonation, and after all I’d done some work deciding which angle I wanted it to take and I’d checked it through first, so I was pretty sure there were no obvious heresies.
But I would have felt bad just delivering it and moving on. So I came clean. After a moment’s quiet, I told everyone where I’d got it from. There was an audible gasp and a general air of polite horror.
And then the comments came, with the overwhelming sense that people felt ‘cheated’. I wondered why. Did people not realise that clergy get ideas from books on shelves, from websites, from podcasts? Is this any different? It could be argued that ChatGPT was just a more efficient form of information collecting which might even save me time to do other things. Setting aside the possibility that the Rector might just be a little lazy, using AI for this sort of task did however mean that I was at the mercy of whichever databases it had decided to mine. I didn’t know where it had gone to find out about 1st century Palestinian shepherds; it could have been anywhere and, importantly, it could have been wrong. Control of what I preached had shifted away from me to an anonymous, impersonal being. It’s a lot easier for me to judge the veracity of a source if it’s a book or an online article or a podcast and there are enough articles suggesting that ChatGPT might be biased; whilst perhaps less biased than some other forms of AI, it tends to select from white western sources. And it is, over a course of interactions, able to work out where I’m coming from on a particular issue. Unlike a book or journal article, which remains constant in its ‘opinion’, ChatGPT’s algorithms may pick up on my strong opinions and reinforce them providing what has become known as an individualised ‘echo chamber’.
It seemed too that my congregation felt the need of a human preacher to discern the value of what is available online. In a world where information is so readily available, how does anyone ever test the credibility of opinions, or the orthodoxy of religious statements? Certainly I have had complicated discussions with several people who have found some bizarre answers to questions from Google. Here the congregation and I had similar views. Part of the role of the preacher or faith leader is to provide some authority and discernment in order to guide and help a church. The best preacher will acknowledge their own bias but will also have had some good training and experience to provide a balanced and nuanced view on things. My experience of trying to get a sermon out of ChatGPT was that it lacked nuance, seemed unable to deal with uncertainty and open-ended questions and leapt too quicky to shallow or trite answers. The consumers of the sermon seemed to agree. Everything, they said, was a bit too certain, a bit too neat for them. Preaching this AI sermon helped me to see that the years of encouraging them to interrogate their faith, to engage with difficult topics and to ask questions had borne some fruit. I’m not convinced AI could do that yet. It certainly can’t replace hours of conversations and sharing of lives, chats over coffee and coffins, the holding of stories from newborns and centenarians that go into the formation of sermons in a parish church. If I plugged all that information into the prompt box, I might come closer, but I could never write it all down efficiently. I might as well pray and write my own sermon!
Others pointed out that the sermon lacked not local but national relevance. In a week where the Pope had died and the UK had celebrated 80 years since VE Day, ChatGPT failed to mention either. It can’t be blamed, of course; if I had done a better job of prompting it, ChatGPT could have included cultural references. Some criticisms were more about my competence as a user than the AI bot and as generative AI improves, as it is doing rapidly, and if users become more experienced in using it, then some of these criticisms will be answered. Deeper, more profound questions are whether AI can ever or should ever replace humans (or in my case, preachers) and can we stay in control of the process (and the ‘robots’) for ever?
This perhaps was behind the more emotional response of some in church: “It’s evil!”, “It has no place in church and you shouldn’t let it in!” which betrayed an underlying fear of what might happen, or a loss of control, or of questions about what it means to be human at all. Whether sentient AI is on the horizon or even possible is a matter of debate; scientists have yet to fully understand what consciousness is, let alone whether a machine can own it. But the idea of a robot that could think for itself, feel emotions and form relationships invites the question of what it means to be human, and what it means to believe that we alone are made in the image of God. Could we imagine a time when AI could replace some of the other things a priest does? There is something quite unique – sacred even – about the pastoral relationship of a priest with their parishioners. Certainly the backlash to the AI-generated “Father Justin” indicated that there remains something distinctively human in what is expected of a priest.[1] The same had been said of counselling, however, and AI is being increasingly successfully used in therapeutic situations so perhaps it is not impossible to see a future where many functions of a priest, not just preaching, are taken on by robots. AI, it seems, is on the brink of showing (or mimicking?) empathy but few people would like a computer rather than a doctor to decide when to turn off life support, even if the computer is likely to be more accurate in predicting the end of life. For the time being at least the truly human touch with its complexities, nuances and even illogicalities is something people continue to value.
What may be a longer way into the future is the creation of artificial superintelligence (ASI), raising another set of more complex questions. ASI could rule the world in a logical and possibly more efficient way than humans but where does that leave us a stewards of the planet, or beings with choices and free will and where does our own responsibility fit in? Those more favourably disposed to AI might suggest that its development is an outworking of our divinely inspired imagination, or even a participation in the creative imagination of the divine itself, but this will not excuse us from entering into the ethical debate around new technologies.
It’s complex and sometimes confusing to think through all the issues with AI. Perhaps a simpler solution would be ostrich-like, to stick our head in the sand and ignore it. But fear of technology or a disinclination to get involved, will not see us through the next decade. AI is here to stay and is developing at pace. We need to engage, to be part of the conversation, and even part of the experiment, if we are to have a voice in its future. The church cannot opt out of questions for practical reasons and should not opt out of the debate for ethical reasons. Professor Anil Seth from the Centre for Consciousness Science, in Pallab Ghosh’s excellent analysis of the debate sums it up well:
“We did not have these conversations enough with the rise of social media, much to our collective detriment. But with AI, it is not too late. We can decide what we want.”[2]
For now, my ChatGPT sermon is a long way from such worries. Its use to me is more as a stimulus for thought, providing a way into a sermon. Or perhaps for a neurodivergent preacher such as one colleague who uses it extensively, it may help organise original thoughts into a more comprehensible form for a congregation. If it were doing that my congregation would feel less cheated. And I don’t suppose they would mind if I asked it to help me fill in a grant application or write a risk assessment; less creative tasks which currently take me away too often from my role as their priest. But others have done more in-depth studies with more experienced AI users and found ChatGPT to have great potential in helping sermon writing, even analysing its methodology and showing it to be capable of sharing personal testimony, having sound Biblical analysis and admirable rhetorical devices.[3] If I were to practise its use, perhaps eventually no one would know the difference.
I’ll leave the final word to one person. “It was OK. Not awful but not as good as usual”. It’s a relief to know in the competition with a robot, on this occasion the preacher came out on top.
(This blog formed part of an assignment for an MA I’m completing but since the congregation’s responses were so much part of it I thought I’d pop it in here. I’ll spare readers my other essays!)
[1] Christopher Check, 24th April 2024 “Just ‘Justin” for now” https://www.catholic.com/news/just-justin-for-now accessed 4th June 2025
[2] Pallab Ghosh The people who think AI might become conscious 26 May 2026 https://www.bbc.co.uk/news/articles/c0k3700zljjo accessed 26th May 2025
[3] En Rui Chua. “ChatGPT’s Gospel Preaching Process: A Grounded Theory Study.” Advance March 22 2024 https://advance.sagepub.com/users/719258/articles/704271-chatgpt-s-gospel-preaching-process-a-grounded-theory-study?commit=2fd33bbe68d36a59a711b1e9093b68d9ccb4588c accessed 15th May 2025
Musings of a Maverick Priest 