Mere Orthodoxy November 10, 2026 14 min signal 2026-11-10

How should Christians approach AI translation?

Archived text mirror, captured for personal reading when the Signal ran. Formatting is approximate — read the original at the source.

Photo by Yen Vu / Unsplash

As new technologies arise, Christians need to go deeper into the Scriptures, deeper into their theology to ground themselves in yesterday’s truth for today’s ambiguities.

The digital repository at “The Faith Received” by Mere Orthodoxy provides both a means for Christians to dig deeper into their received tradition for modern issues but also a prompt for further debate surrounding how to respond to and implement AI. The main source of negative feedback surrounding “The Faith Received” has been in relation to its AI-reliant translations.

Relying in-part on the texts found at “The Faith Received,” I want to build a theological case for the rationale behind AI translation that avoids potential dangers, preserves the good of human translation and editorial work, and safeguards against the humanization of AI. The theological case for AI translation involves three moves: understanding human communication, understanding the task of translation, and understanding how AI can play a proper role in the translation process.

What Is Human Communication?

Texts written by humans are more than predictive words on a page. Something special happens when humans write and speak. We use these external signs to point to a deeper reality. Augustine points to this reality in On Christian Doctrine (1.2):

But there are other signs whose whole use is in signifying, such as words. For no one uses words except for the sake of signifying something.

Augustine expands further on this concept in the second book (2.1):

For a sign is a thing which, beyond the appearance it presents to the senses, causes something else from itself to come into thought: just as when we see a track, we think that the animal whose track it is has passed by; and when we see smoke, we know that fire is underneath; and when we hear the voice of a living creature, we notice the affection of its mind; and when a trumpet sounds, soldiers know that they must either advance or retreat, or do whatever else the battle requires.

For Augustine, human language takes something from deep inside us, symbolizes it, and allows others to sense those symbols, grasp the internal intent of the author, and even respond appropriately to the message. It’s the plausibility of this process that Derrida challenged as he argued against the notion that communication entails a stable, determinate, transferrable meaning from communicator and recipient. He held that “essential” to the process of communication was the inherent reality of failure to convey intent from one person to another via signs (see “Signature Event Context,” p. 15). Derrida’s default was suspicion of rather than confidence in the plausibility of signs to symbolize our thoughts in a reliable manner.

Contrary to the postmodern outlook of Derrida, Christians have largely held to the more confident Augustinian notion that communication through signs is—although not without its challenges—a reliable way for deep internal words to be expressed by an author/speaker and understood by the recipient. And this view is rooted deeply into Christian understanding of revelation and the Trinity. The very plausibility of the written Word of God being accessible to humans relies on the plausibility of humans being used to convey God-breathed internal words in proper written words, and for other humans to be able to access the external words and thus understand the interior word of the Author. And in a similar way as the written Word of God extends to us, the incarnate Word of God proceeds to us. As Augustine writes in On the Trinity (15.11.20):

For that which is brought forth by the mouth of the flesh is the voice of the word: and it is also called a word, because of that for which it was assumed so that it might appear outside. For thus our word becomes in a certain way the voice of the body, by assuming that in which it may be manifested to the senses of men; just as the Word of God was made flesh, by assuming that in which it also might be manifested to the senses of men. And just as our word becomes voice, and is not changed into voice; so the Word of God was indeed made flesh, but far be it that it should be changed into flesh.

This claim falls in a larger argument (15.10–11), where Augustine is making the point that knowledge (scientia) progresses from deep in the mind where we may not even consciously think about it into an inward word (verbum cordis) as we form thoughts apart from any particular language.

For a thought formed by the thing which we know is a word which we say in the heart: which is neither Greek, nor Latin, nor of any other language; but when it is necessary to convey it to the knowledge of those to whom we speak, some sign by which it may be signified is assumed. (On the Trinity, 15.10.19)

From the verbum cordis, our language proceeds in outward expression (signum) through speech, gesture, and writing. And those expressions or signs allow for a corresponding (tale…quale) idea to be formed (fiat) within the hearer:

But when we speak to others, we apply the ministry of the voice, or of some corporeal sign, to the word remaining within, so that through a certain sensible commemoration such a thing may be made also in the mind of the hearer, as does not depart from the mind of the speaker. (On the Trinity, 9.7.12)

Aquinas also picks up this thread and expands on it somewhat in his Summa. The process of human communication from an interior word of the communicator to an interior word in the recipient through intermediary signs lays an important foundation for then understanding what translation is.

What Is the Task of Translation?

The task of translation is the transfer of signs from one medium to another in a manner that retains fidelity to the interior word of the original communicator. Let’s unpack that claim.

First, translation is a type of medium transfer. A variety of transfers in medium can occur to an external sign. A spoken word can be transferred into a written word, a written word into spoken word, a written word into digital text, a spoken word into an amplified word, a word in one language into a word in another language. The transfer of external signs from one medium to another does not inherently disrupt the flow of thought from the encoder to the decoder. As Bavinck writes in his Philosophy of Revelation:

The distinction between the cause and the condition, between the mediation and the object of the perception, for all this, retains its full validity. Just as writing and reading, telegraphy and telephony avail themselves of all sorts of mechanical movements of hand and tongue or of all kinds of visible signs and audible sounds, and nevertheless presuppose at each end of the process a thinking subject which by means of the signs understands the thought . . . (Lecture 3)

A sermon serves as a great example of this concept. A preacher who wrestles with the Word and then proclaims it to the gathered congregation speaks as if God were speaking through him (1 Pet 4:11). That word may be amplified by a sound system, carried by radio waves, or recorded in an audio file. The audio file could be transcribed into text. That text could be translated into another language. Someone can pick up that translation and read and be convicted and challenged by the interior word of the speaker, as shaped by the internal word of the Spirit. The possibility of this reality is truly incredible.

Second, translation is beholden to the standard of fidelity to the internal word of original communicator. This raises a host of challenges. Ludwig Wittgenstein would press us to think about the fact that our language is doing far more than conveying simple objects.

Augustine, we might say, does describe a system of communication; only not everything that we call language is this system. (Philosophical Investigations §3)

He suggests a series of “language-games” we play as we present results, tell stories, sing songs, tell jokes, etc. (§ 23). All of these are more complex than mere symbol-for-symbol, token-for-token transfer. They call for something more complex in the act of translation. Ferdinand de Saussure also noted these challenges (Course in General Linguistics, see especially 115–16). English and French—for example—function differently as linguistic systems, and a word-for-word translation may not be a faithful translation. Here Jerome is helpful to us in his “Letter to Pammachius.” Defending his translation approach, Jerome writes, “apart from the holy Scriptures, where even the order of words is a mystery, I express not word for word, but sense from sense.” Although sometimes the translation of words in order would cause confusion, we need to proceed in a manner that makes broader consideration of the likely interior word behind the exterior signs. Translators must do more than exchange one sign for another.

Augustine also reinforces the importance of conveying the sense of the text as he surveys the translation of the Septuagint in the City of God (18.43):

…the same Spirit would be saying both; and this very thing otherwise, so that if not the same words (non eadem verba), yet the same sense (idem sensus) might become clear to those who understand well…

The goal of the translator is to achieve the sense of the author even when the words can’t be translated word-for-word.

Beyond this challenge is the reality that the mind of the author isn’t accessible to translators of ancient works. Augustine isn’t available for us to query as to his meaning. So is the task of translation impossible? This problem is similar to the issue inherent to language. As we’ve seen, Augustine calls us up to a greater level of confidence in the possibility of interior words conveying themselves to recipients by fiat. Paul Ricoeur in some ways supports the approach of Augustine specifically when it comes to translation. Ricoeur suggests that texts when written gain “semantic autonomy” (Interpretation Theory, 29). Although Ricoeur will go on to prioritize meaning to the recipients over the original meaning of the author (30), his point is still worth noting. The text, the symbols encoded by the author, and any accessible context provide us with the closest means to understand the interior word of the author. So the job of the translator is to faithfully transfer the text of the artifact from one language medium to another and thereby most nearly convey the interior word of the author to those who have access to the new medium.

Third, the standard by which a translation is judged “faithful” presents unique challenges. Who gets to determine the faithfulness of a translation? Epistemologically, faithfulness is grounded in the source, not in outside translators, editors, peer reviewers, publishers, etc. Similar to how Scripture is understood as having a quality of inspiration and canonicity inherent within itself (i.e., “manifest themselves to be the word of God,” WLCF 1.4). The church recognizes rather than creates that authority. In a similar, though not identical, way, the fidelity of a translation isn't conferred by a translator, editor, publisher, or reviewer. Fidelity is a property of the relationship between the translated text and its source. Human judgments can discover, affirm, challenge, or correct that fidelity, but they don't create it simply by pronouncing a translation faithful.

This conviction drove the earliest translators such as Tyndale, whose 1526 Bible was prefaced with the recognition that he still needed to do more work to “seeth it better.” But even that need for improvement didn’t need to hold the translation up for formal review or authorization. Unfaithful aspects could be weeded out in time as they were recognized.

In our current digital era, the ability to open source the process of ensuring or achieving faithfulness in translations becomes easier and easier. Accessibility of digital texts and ancient manuscripts provides the average person a method to confirm or contradict the faithfulness of the translations we have, without relying solely on publishers or credentialed authorities.

Human artistry and error are the blessing and the curse of translation work. At times, translators beautify the text and shape the course of human language in their wake (as did Tyndale and Luther). But at times, human translators debase the text intentionally or unintentionally. In Wulfila’s Gothic translation of Philippians 2:6 or the translation of John 1:1 by Jehovah’s Witnesses, we find key orthodox theological statements that appear skewed in translation. And so all translation must and can be held up to the standard of fidelity to its source, the exterior word of the author as an artifact of his or her interior word, as difficult as that process may be.

How Can AI Play a Role in Translation?

If human communication is the expression of interior words with exterior words, and if translation is the transfer of signs from one medium to another in a manner that retains fidelity to the interior word of the original communicator, then what happens when AI is involved in the process of translation? Do we impart to AI a soul with an interior word? Do we give AI an inappropriate role with personified characteristics reserved for human beings and the imago Dei?

I believe John Searle’s Chinese Room analogy helps us answer these questions. Back in 1980, he wrote “Mind, Brains, and Programs,” suggesting how we ought to understand what’s happening when a computer provides words that look like those of a fluent speaker of a language. His thought experiment imagines a person who can’t understand Chinese but can produce Chinese responses by following an English rulebook for what Chinese characters to write in response to the Chinese characters he received. Although the responses appear meaningful, they are simply a product of symbol manipulation, not necessarily the product of an interior word.

Now connect Searle’s thought experiment to AI. Symbols come in one window in Latin. According to a massive rulebook and in full context (not just symbol-by-symbol), the LLM sends back a response in English. No interior word. Simply external signs transmuted from one medium to the next. Modern LLMs can receive prompting, system instructions, and skills that will enable them to provide editorial notes surrounding their responses, qualifying alternatives, annotating ambiguities, or highlighting parallels. All of this is the process of transfer from one window to the next without interior understanding.

One staunch critic of modern AI tools, Emily Bender, draws attention to the fact that LLMs function as stochastic parrots, manipulating signs without understanding what they signify. To impart understanding to the LLM is to impart something uniquely human to AI. And we don’t need to impart that understanding in order to allow LLMs to do what they do effectively: rearrange symbols according to human-provided rules.

As with human translation, LLMs run into challenges of artistry and error. It can replicate art, lacking true originality as an artist. As a result, without significant human guidance, its translations will lack the beauty and detail of normal human translations. And AI is prone to hallucinate, resulting in problematic translations that can be caught via multiple blind runs. That said, LLMs tend to translate with less bias in contrast to human translators. In my experience, I’ve never achieved skewed translations without specific instructions in a particular direction. An LLM will translate a work on mathematics in Latin according to the same grammatical rules as a work on theology. Both of these factors—just like the corresponding challenges with human translations—are a real blessing and a curse. While humans are prone toward artistry that may break the confines of fidelity, LLMs are prone toward fidelity that skews away from something resembling artistry.

Conclusion

Christian translators don’t need to feel compelled to use AI nor to reject the use of AI. They should follow the guidance of their conscience and use their God-given gift in the way they are led. Christian translators who avoid AI should do so with a passion to exercise their gifts and artistry in translation within the appropriate bounds of fidelity. They should strive to faithfully understand the internal word of the author and convey that word through a new medium into the hearts of new readers. They should beware not to upstage the author or build a name for themselves on the legacy of the author. Instead, the goal is to elevate the author and his or her internal words as conveyed by the written artifact.

Christian translators who leverage AI should do so with confidence that their work aligns with how Christian theologians have reckoned with communication and translation theory over the centuries. New technology may force us to reckon with our ancient theology in new ways, but those historic understandings help clarify rather than muddy the waters on complex issues related to the ethics of AI use. We can leverage AI to arrange signs without imputing an internal voice behind the words.

Much remains for humans to do when it comes to the translation of texts, whether we use AI or not. Developing stronger rules for translation, editing and reviewing translations, and lending our voices to the credibility or lack thereof in the torrent of translations now being released. And in so doing we can testify to what AI is and is not, what humans are and are not, all the while supporting literacy and extending access to our Christian heritage to the next generation.

The Weekly Digest

Premier Thought Every Thursday.

All of our recent essays and podcasts, delivered to you. Free.

First Name

Last Name

Free. Delivered Thursday mornings.

Technology

The Author

Technology

More essays by Technology →

The Faith Received

The Author

The Faith Received

More essays by The Faith Received →

The Author

Phil Thompson

Phil Thompson (PhD, Columbia Biblical Seminary) serves as director of digital strategy for the North American Mission Board and pastor of Christ Fellowship Eastside in Taylors, South Carolina. He previously served for seven years with The Gospel Coalition, building and launching, among other resources, The Carson Center for Theological Renewal. He has been married to Laurel since 2008 and has four daughters: Lane, Kately, Harper, and Darcy.

More essays by Phil Thompson →

Mere Orthodoxy

The Author

Mere Orthodoxy

More essays by Mere Orthodoxy →

Give a membership

One month, three, six, or a full year of Mere Orthodoxy for a friend, a pastor, or a seminary student. From $10. Give a gift →

For a small team

A church staff, a session, or a reading group. $80 per seat, five to ten seats. See group pricing →

III

For your whole organization

Churches, seminaries, colleges, and denominations. $2,500 a year, as many people as you like. Get organizational access →

For students

A discounted membership for students with a school email. Digital, or with the print journal. Join as a student →

← latest Signal