One of the topics of conversation around the Matthias Media table in recent months has, unsurprisingly, been artificial intelligence. More specifically, we’ve been drafting and discussing our policy on the use of AI in the production of resources. This article is not an official announcement of where we’ve landed (it’s still a work in progress); it’s more of a peek behind the curtain at some of the ideas we’ve discussed, mixed with a couple of my own musings on the topic. And while this is well short of a final policy, I hope these reflections might help those of us charged with teaching the Bible in any setting—including the vital ministry of Christian writing—to think about what our task might look like in the days to come.

It seems everyone is having the same discussion in one form or another. Just this week, the White House hosted tech leaders from across the AI world to sign the ‘Accord on Super Intelligence’. Much closer to home for us, a colleague here at Matthias recently shared this paragraph, which was posted online as an excerpt from the policy that Eerdmans (a US-based Christian publisher) recently sent to their contracted authors:

The scholar who wrestles with a sentence for an hour, who discards three drafts before finding the right word, who sits with a difficult source until understanding comes, is not merely producing content. The scholar is being formed. The practice of writing shapes the writer: it demands honesty before the blank page, patience with complexity, and accountability to readers who deserve a human mind behind the words they trust. When that process is bypassed or automated, the product may appear unchanged, but the person who would have been formed by the labour is diminished.*

I’m sure many writers—and many readers—will be moved and encouraged by those words. It’s a beautiful picture, but also a challenging one. Bible teachers have long faced the temptation to take ‘short cuts’ in their work by relying on ‘super intelligence’ that surpasses our own. Rather than spending deep and sustained time in the word as we prepare to teach it (in whatever form)—allowing it to form us before we form our teaching—we can rely too much on the commentaries, or we can “jump online and see what Don Carson said in his talk”. In this sense, AI is not a new temptation for the Bible teacher; it simply brings a fuller and more confronting version of the temptation.

Where AI may be genuinely new is in its ability to do much of the work of presenting a final, polished version of the work. Sure, commentaries have been around for a long time; but after we read a commentary or two, we still need to do the work of packaging a talk or a chapter. AI can potentially do much of this for us. And yes, a preacher has long been able to steal chunks of a sermon from Carson or Keller or Spurgeon or whomever. AI does not remove the danger of plagiarism, given what we know about how LLMs work (our draft policy talks about “unintended plagiarism”), but it makes the danger feel smaller, more distant.

But this is where the Eerdmans policy is so insightful. The careful, sometimes painstaking elements of Christian writing—pondering which word to use, choosing between a full stop and a semi-colon, agonising over when to start a new paragraph—is not wasted time or mere pedantry. Okay, yes, at some point it’s mere pedantry. But most authors get better by becoming more interested in the details, not less. The gospel is in the details. Labouring over our choice of words (whether delivered from the pulpit, the lounge chair, or the written page) is part of letting God’s word do its work in us.

John Piper addressed this in an important book called Seeing Beauty and Saying Beautifully. Piper writes about George Herbert, George Whitefield and CS Lewis as three examples of people who “made poetic effort in their Christ-exalting communications”:

They made poetic effort to see and savor and show the glories of Christ. This effort was the God-dependent intention and exertion to find striking, penetrating, imaginative, and awakening ways of expressing the excellencies they saw. My thesis is that this effort to say beautifully is, perhaps surprisingly, a way of seeing and savoring beauty.

When you spend a little longer polishing Sunday’s talk, or you take some extra time over that weekly email to the church or your Bible study group, or you give that sample chapter one more read before sending it to Matthias Media instead of dumping it into ChatGPT, what are you doing? Yes, you’re spending more time dwelling on God’s word so it can change you, which is always good. But if Piper and the Eerdmans statement are right, you’re also doing something more specific: the very act of polishing that talk, improving that email, refining that chapter, becomes part of how God forms you. Proofreading becomes praise. Editing becomes edification.

But so far, most of what we have said (and most of the Eerdmans statement) have been about the author. What about the recipient—whether reader, hearer or Bible-study member?

The Eerdmans statement says readers “deserve a human mind behind the words they trust”. What might this mean?

As part of writing this article, I tried a little experiment: I made AI write me a song. I told an LLM to write a prompt that I could feed into an AI music generator. The core instruction was to write a song called Where, O Death? that celebrated Christ’s resurrection, not with a congregational song of praise but with a rock song. You can listen to it here.

If you’ve not listened to an AI song before, as I hadn’t, it might be an odd experience. But the one thing I promise it will not be is a personal experience.

It will not be an experience shared between you and a fellow image-bearer who laboured to create the art. The machine that created that song has never sat at the bedside of a dying loved one. It has never attended a funeral and been made to face its own mortality. It hasn’t laboured to create an artistic expression of grief mingled with hope, weeping and smiling while it laboured. All it’s done is follow the path of least resistance and aggregated 30 years’ worth of arena rock into an audible smoothie.

This is why I’m becoming more optimistic about the future of creative endeavours in an AI world. I’m not claiming there will be no impact. But I think humans will continue to seek out and respond to art produced by other humans because human creativity is the essence of art. It’s a window into how one person—or perhaps one group of people—apprehends reality. This is part of why the term ‘AI slop’ has caught on: we recognise the value of a human being producing something by wrestling long and hard with their subject matter—bringing with them all their unique gifts, all their experiences of life and sin and grace and grief and hope—as opposed to a machine turning the crank and producing, well, slop. Familiar and polished slop it may be, but slop it remains.

After all, humans didn’t exactly lose interest in chess after ‘Deep Blue’ beat Garry Kasparov in 1997. Americans didn’t stop watching Jeopardy! after ‘Watson’ beat the game’s best human players in 2011. What’s more fun: asking Siri to tell you who’s won the most Oscars, or going to a quiz night with friends and falling about laughing because you realise you’re not as smart as you thought you were?

I digress, but only slightly. Because I think this gets to the heart of what Eerdmans meant when they said readers “deserve a human mind behind the words they trust”. Bible teaching in all its forms is hard work. But the hard work is worth it, and not only because it forms the teacher. It’s part of the relationship between teacher and audience.

I don’t remember where I saw these words recently, but I remember them jumping off the screen: “The writer struggles so the reader doesn’t have to”. What a glorious description of the Bible teacher’s calling—and of the difference between human creative labour and AI slop. Almost by definition, there’s no struggle in anything an AI generates; it never takes the road less travelled, and by nature will only ever be adequate or good. The task of any Bible teacher (including an author) is to grapple with God’s word until they see marvellous things there, then struggle to make those marvellous things as clear and compelling as possible.

When I’m listening to a sermon, I don’t really want to know what some commentator thinks about this week’s Bible passage. I want to know what grabbed the heart and mind of you, human preacher standing in front me, as you grappled with God’s word this week, saw marvellous things, then thought long and hard about how to say them so that I may see them too. And I don’t really want to know if you had a cool idea for a book and then used AI to figure out what human culture at large says on the topic. I want to know what insight or experience made you think this was a cool idea. I want to know that you searched the Scriptures on your topic, wrote your outline, asked your family about it over dinner, wrote an introduction then shredded it, rewrote your outline, found two key passages you missed the first time, and ended up writing a great book.

That’s what your readers deserve.

* This was posted online as part of an internal email sent from Eerdmans to contracted authors, and so has not been verified as authentic. But there is no reason to doubt its veracity; here is the Facebook post.