Simulated Archaeology: Reconstructing the History of Civilizations That Never Existed

What happens when the amphora we are admiring and the ancient language we are reading are not the product of human history, but of an algorithm? "Simulated Arch

Admiring the fragment of a finely decorated amphora, deciphering a stele inscribed in an unknown language, and mapping the ruins of a city buried under millennia of sand are the supreme epiphanies of archaeology. But what happens when the amphora, the stele, and the dust have never existed outside of a server?

The intersection between Generative Artificial Intelligence and worldbuilding is giving rise to a new, fascinating, and controversial artistic discipline: simulated archaeology. It is not about reconstructing the real human past, but about engineering the illusion of a fictitious past so coherently that it deceives our senses. Through procedural generation, AI is now capable of synthesizing artifacts, languages, and entire chronologies of imaginary civilizations.

In this in-depth piece for the Scenarios and Reflections column, we will explore the engineering behind the construction of these synthetic memories. The fundamental question we must ask ourselves concerns not only the power of the software, but human psychology: if an algorithm produces ceramics, inscriptions, and sacred texts that are perfectly coherent with one another, what drives us to perceive that artificial past as authentic? And what is the ethical boundary that separates generative art from the distortion of our true historical heritage?

1. Designing the Past: Procedural Generation and the Power of the Gap

To create a credible civilization, the technology relies on Procedural Content Generation (PCG). As illustrated in a technical review on the integration of Large Language Models in gaming, contemporary procedural generation is not limited to producing random textures for virtual maps [1551]. By leveraging linguistic and visual models, creators can set the geological, climatic, and social rules of a world, letting the algorithm calculate and generate the architecture, diet, and artifacts that follow from those premises.

However, the true genius of simulated archaeology does not lie in the perfection of the output, but in its imperfection. An illuminating study published by ACM ("A Study of Procedural Generation through Archaeological Storytelling") demonstrates that the credibility of a synthetic artifact derives from its fragmentariness [1559]. If AI generated an intact virtual temple, it would look like a video game; but if it generates a chipped vase, an asymmetrical arrangement of votive objects, or a text with entire missing sentences, it triggers the spark of interpretation. It is the gap that invites the human mind to connect the dots, to imagine the catastrophe that destroyed the object or the hands that buried it. The fictitious artifact becomes credible precisely because it requires, as in real archaeology, a hermeneutic effort on the part of the observer.

2. Giving Voice to Ghosts: Self-Generated Languages (Conlang)

No civilization can exist without a voice. The most sophisticated level of archaeological simulation concerns the synthesis of constructed languages (conlang). Inserting exotic syllables at random is not enough to deceive a careful eye; a convincing fictitious language requires a coherent and implacable phonology, morphology, and syntax.

A research prototype ("Towards Procedural Generation of Constructed Languages for Games") explored exactly this boundary, demonstrating how algorithms can derive vocabularies and grammatical rules starting from simulated phonetic roots, simulating the evolution of a language over the course of fictitious centuries [1550]. As highlighted by studies from Goldsmiths, University of London, worldbuilding through invented languages is the narrative tool par excellence [1552]. Dialectal variations within AI-generated texts, lexical contaminations in a border stele, or the evolution of proper names can suggest, without ever explicitly declaring them, the migrations, trade wars, or religious conflicts of a civilization that has never set foot on Earth. The coherence of language is the guarantee seal of the illusion.

3. The Deception of Coherence and the Distortion of History

The extreme verisimilitude of these processes, however, opens an ethical chasm. When simulated archaeology leaves the safe perimeter of narrative fiction or artistic installation to blend into the uninterrupted flow of the web, the risk of historical contamination becomes concrete. An invented artifact, if generated with the right aesthetic, can easily appear to be historical evidence.

The Ure Museum of Greek Archaeology at the University of Reading has confronted this threat head-on with the exhibition "Distorted History: AI's Skewed Visions of the Ancient World" [1549]. By placing AI-generated images alongside authentic ancient depictions, the exhibition highlighted how algorithms, trained on datasets steeped in modern stereotypes, flatten and deform our real conception of antiquity, creating "historical fakes" that are aesthetically perfect but philologically aberrant.

The damage is even more serious when AI draws on living and minority cultures. A critical analysis relayed by Phys.org ("AI is distorting archaeology and flattening Indigenous histories") denounces an alarming phenomenon: images of artifacts that never existed, generated by merging styles stolen from Indigenous communities, circulate on social media as authentic documentation [1555]. This practice is not simple generative art, but an undue appropriation that flattens cultural identities and threatens the rights of native populations, diluting their real historical artifacts in a sea of synthetic replicas.

Key Operational Takeaways (for Artists and Communicators)

Declare the Synthetic Origin: Simulated archaeology is an art form, not a scam. Any worldbuilding project that uses AI-generated artifacts, texts, or photorealistic images must include explicit declarations, watermarks, or unambiguous captions attesting to their fictional nature, to avoid feeding the disinformation ecosystem.
Design the Void (for Creators): If the goal is storytelling, do not ask the algorithm to generate perfection. Exploit imperfections and ask the AI to simulate wear, fragmentation, and incomplete translations. Lore (the history of the world) emerges more powerfully from what is missing than from what is shown in full.
Protect Real Heritage: Generation platforms and artists must set rigorous ethical limits to avoid generating fake artifacts that imitate specific real Indigenous or minority cultures without their consent. Inventing an alien or fantasy civilization is a creative exercise; falsifying the remains of an oppressed culture is an act of historical erasure.

The Value of Declared Fiction

Simulated archaeology fascinates us because it exploits a wonderful bug in our brain: the spasmodic need to find logical meaning in the traces of time. When Artificial Intelligence provides us with coherent fragments, our mind instantly constructs an entire world to host them.

But the coherence of a story is not archaeological evidence. It is simply the mechanic that makes the narrative experiment so powerful and, at the same time, so potentially deceptive. If we do not learn to draw an insurmountable line between the history carved in stone by our ancestors and that calculated by our servers, we risk transforming our shared past into a gigantic, beautiful algorithmic hallucination.

Bibliographic References and Sources

Article by the Editorial Team of La Bussola dell'IA