Authentic Artifice?

Later this summer, Medieval Institute Publications and De Gruyter will publish Authenticity in Medieval and Early Modern Literature, a volume I have edited and which stems from a 2023 conference of similar scope. Eleven chapters in the volume ask what authenticity means, and how it is used, across select literary categories: forgeries, histories, translations, and continuations and canon-formation. I intended to write a blogpost about the volume, conceptions of authenticity, and the International Congress on Medieval Studies – after all, the ICMS is a conference where hundreds of academics, researchers, and enthusiasts attempt to get closer to the ‘authentic’ Middle Ages, whether by routes historical, creative, critical, archaeological, and so on.

But I have been struck, in the two years since the conference and in the writing and production of this volume, how much – and how rapidly – the conversation surrounding literary authenticity has shifted. The emergence of generative Artificial Intelligence (AI), in vastly more powerful forms than hitherto possible, has prompted both excited interest and intense apprehension, especially (it seems) in the academic Humanities sector. This edited volume is a firmly premodern exploration of authenticity, drawing a clear distinction between post-Enlightenment, post-Romantic authenticity (especially as linked with sincerity and individualism) and premodern authenticity, which emerges as something diverse, communal, authoritative. And yet my attention has been caught by the conversation surrounding authenticity and AI, and its resonances with the volume’s explorations of premodern literary authenticity.

Is AI authentic? That depends, in part, on your definition of authenticity. The current ‘usual sense’ listed in the Oxford English Dictionary is “genuine; not feigned or false,” and one sub-definition of this is “having the stated or reputed origin, provenance, or creator.”[1] Large language models (LLMs) are trained on other peoples’ work, and generative AI produces versions of that material; the obfuscation of origin, provenance, and creator immediately raises suspicions of inauthenticity. And this rejection of AI as authentic also evokes definitions of authentic as “original.” Until relatively recently, originality was an intrinsic aspect of authenticity (“an original document,” “original, primary, not a copy”), a definition strengthened by the advent of copyright in the eighteenth-century Statute of Anne.[2] That would exclude generative AI, by nature secondary, a copy, unoriginal – except that those particular definitions are now obsolete, replaced with another subsection of the “usual sense”:

Presenting the characteristics of the original; accurately reproducing a model or prototype; made or done in the original or traditional way.[3]

This definition has been attested since 1893, and contradicts the earlier sense that authenticity could only be found in an original; here is authenticity in something which presents the characteristics of or accurately reproduces an original. Generative AI does just that; it is not the original, but presents us with something like the original, a simulacrum. In this way authenticity has circled back to understandings of authenticity contributors explore in the volume. As I write in the Introduction, premodern writing could be considered authentic “if the addition, continuation, or even imitation was true to the style of the original author… authenticity could originate externally.”[4] Thus we might find an authentic medieval forgery or an authentic early modern continuation of classical works, as several in the volume do. So K. K. Ruthven calls MacPherson’s notorious Ossian forgery “an original and authentic fake.”[5] The meaning of authenticity has been in flux throughout the word’s history. Could authenticity now be redefined or its meaning adjusted in the light of AI – and will this be in ways which echo the premodern past or which push it somewhere new?

One way in which AI is testing authenticity in a particularly premodern sense is in its engagement with the question of authorship. A recent study from outside the realm of literary scholarship particularly illustrates the point, by Colleen P. Kirk and Julian Givi, who examine what they term the “AI-authorship effect” on the interpretation of authenticity in AI-generated marketing communications.[6] What is interesting to me (beyond their exploration of “moral disgust” responses to forms of AI communication) is the persistence of the link between authenticity and authorship. In the context of their research, the “AI-authorship effect” is “serially mediated by perceived inauthenticity”:[7] a consumer’s reactions to an AI-authored text depends on how authentic they find it, which prompts the question of how humans perceive something to be authentic at all. But that link is one which has persevered through the centuries, as Alastair Minnis explores in his chapter in this volume. So Hugutio of Pisa writes in his twelfth-century Derivationes:

AUGEO -ges auxi auctum, means to amplify, to make increase. Hence the word auctor, that is, “one who augments.” … And from autor, which signifies autentim, is derived autoritas, that is, “a wise saying worthy of imitation,” and autenticus -a -um [authentic].[8]

Author (autor), authority (autoritas), and authenticity (autenticus) are all linked, and Minnis traces this fundamental medieval connective tissue through a variety of other texts and contexts. While authenticity’s associative links have shifted, notably accruing connections with ideas of sincerity, it is fascinating to think about how authorship has remained central throughout various turbulent moments in authenticity’s history and future. What, then, does it mean for an AI to have ‘authored’ a text? If AI is the author of its new text, like a translator might be the author of a text configured in a new way (as chapters in the volume attest), does author still equal authority and authenticity?

Secondly: can artifice be authentic? The question was prompted by a recent paper by Megan Ward, “Can Chatbots Be Authentic?,” which makes the point that artifice is literally in the name of Artificial Intelligence.[9] Artifice is traditionally framed as an oppositional force to authenticity; artifice is “feigned or false.” Interestingly, however, this is a binary not supported in the medieval engagements with authenticity which the volume explores. Indeed, sometimes the opposite is proven true: artifice can lead to authenticity.

Ward also explores the history of “AI authenticity” and technologically mediated human communication, especially the simultaneous communication provided by the nineteenth-century technology of the telegraph. There were dangers, she argues, in “separating the body from language.”[10] This is precisely what happens in printed books, or parchment manuscripts, or papyrus scrolls – the body is separated from language (albeit without simultaneous communication). Has the debut of new technology (the codex, the printed book, the telegraph) and changes in written culture always threatened authenticity, and to what degree were those dangers realised? 

My particular field of interest is in forgery, particularly literary forgeries. This is perhaps the most straightforwardly ‘inauthentic’ literary form, and rather than asking whether generative AI is authentic, one might instead ask whether it is a forgery. But AI is changing the nature of forgeries in unexpected ways, too. I have often returned to Antony Grafton’s idea of the forger and the critic as inextricably entwined, one feeding the other, wound together like Laocoon and his serpent.[11] Two literary-historical examples: Robert F. Berkhofer III, in his excellent study of medieval documentary forgeries, makes the point that the developing reliance on documentation in the eleventh and twelfth centuries led to more forgeries.[12] Similarly, Walter Stephens and Earle A. Havens point to the early modern period, with all its focus on the new techniques of forgery-hunting humanist philology, as a “golden age” of forgery.[13]

In 2001 Ruthven extended Grafton’s idea to modern-day hackers who “[improve] one another’s performances until they become superstars.”[14] The cycle has only intensified, and humans are being removed from the equation altogether. Last year I updated Grafton and Ruthven’s analogies by citing AI-generated plagiarism and anti-plagiarism AI;[15] and at the time of writing, Springer Nature have announced the development of an AI tool which will spot AI writing in submitted publications.[16] Both the forger and the critic have become AI. What does that mean for authenticity, and for our field?

I am no special expert on generative AI; I avoid it as much as possible in my own research, and I do not believe it benefits my students to use it (especially in my particular field, teaching Old and Middle English literature). Perhaps this is a philistine approach: the scribe refusing to use paper instead of parchment, or parchment instead of papyrus. And some are making the alternative, positive case: Laurent Dubreuil’s 2025 Humanities in the Time of AI argues for “paradoxical optimism,” and that “AI offers a chance for the humanities to strengthen their relevance and their signification.”[17]

But I am interested in how ideas of literary authenticity will respond to the advent of this new kind of artifice. As I have said, the volume’s premodern focus links authenticity with the communal, the authoritative, and sometimes the accurate (and this is perhaps the least stable of all its links). Later in the millennium, ideas of authenticity were instead connected to the individual and ideas of sincerity. I open the volume by invoking forms of capitalist authenticity (“brand authenticity,” or “being true to oneself” styles of self-help guides), which push authenticity even further towards individualism, and which seem hollowed of sincerity. If I were to write that Introduction now, I would ask different questions. Is there a place for authenticity in AI-generated text, or in the new world of literature and AI? Its knowledge has not been “authentically” gained, having – in some cases, and allegedly – illegally used copyrighted material without permission (perversely returning to the communal, as though Queen Anne’s Copyright Statute was never enacted).[18] Will authenticity be de-coupled from ideas of authorship and authority, those links I suggested have endured, or will it take on new forms? The forms of forgeries, histories, translations, and continuations, the categories structuring the book, are already changing, and their relationship with authenticity with them. It is hard not to relate the unfolding technological advancements (or, some would say, crisis) to questions of authenticity which stretch back through our literary history.