About
AI policy
Where we use artificial intelligence on this site, where we do not, and why.
Last updated 23 October 2026
Our position
Ancient Cyprus is a non-commercial site, free to read and funded by its founder. AI lets a project of this size do work on a scale it could not otherwise afford: narrating thousands of lines of teaching, and indexing a century of publications on Cypriot archaeology.
We consider AI acceptable for work that is instructional and is not an art form in its own right. We use it to read a lesson aloud, to find names and dates in a catalogue, and to organise records. We do not use it to make images.
Images
No image on this site is made by AI. The photographs of objects are photographs: our own, those of museums and excavation teams, or ones published under open licences, each with its credit. Plans of tombs and buildings are traced from the drawings published by their excavators. Our 3D reconstructions are modelled from the excavators' plans and measurements and drawn in your browser by our own software; no image-generation model is involved.
A photograph of an object is a record of something that exists, and readers should be able to trust it as one. A drawing or a reconstruction carries the judgements of the people who made it, which we can explain, source and correct. We also value the work of photographers, illustrators and artists, and we will not replace it with generated pictures.
AI recreations of ancient life
Videos and photographs that claim to show daily life in the ancient world, made with image and video generators, now circulate widely. We think they do real harm to public understanding of the past, and we will never use them on this site.
These models do not know anything about antiquity. They reproduce the patterns in the pictures they were trained on, which are mostly modern: films, television, games, tourist photographs and earlier popular illustration. What comes out is the popular image of the past, with its familiar errors. Periods centuries apart are mixed together, and regions are blended into a generic “ancient” look of white marble, togas and Egyptian or Roman décor. Cyprus, whose material culture was distinctive at every period, is especially likely to disappear into Greece, Rome or Egypt.
The models also carry the biases of their training images. People tend to be given modern faces, bodies, skin tones, hairstyles and make-up. Elites, warriors and monuments are shown far more often than farmers, potters, children and the old, and women appear in the roles that modern imagery gives them. Little of this rests on evidence from the period shown.
Where the evidence runs out, a generator fills the gap with invention: clothing, furniture, buildings, food and objects that were never found and often never existed. It renders all of it with the same photographic certainty, so a viewer cannot tell what is known from what is guessed, and there is no source to check and no author to explain or correct a choice. Once these pictures circulate, they turn up in image searches next to real photographs and are taken for evidence, and some are passed off as genuine finds.
Our own reconstructions work the other way. Every visible element is built from the excavated evidence of the period and the place, and carries a stated level of confidence: excavated, inferred, reconstructed or speculative, with a note on the evidence for it. Where the evidence is silent, the page says so and leaves the question open.
Narration
The spoken narration in our reconstructions and games uses synthetic voices, made with AI text-to-speech from scripts that we write. We listen to the recordings before they are published and re-record lines that are wrong, including the pronunciation of place names and Greek terms.
This narration is teaching: it explains how a house was built, how copper was smelted or how a pot was fired, and it makes these pages usable for people who would rather listen than read. The site has more than four thousand narrated lines. Recording them with professional narrators is beyond our budget, and synthetic voices let us narrate every reconstruction and game instead of a few.
Registers and data
The site includes large registers of the people, institutions, documents, excavations and sites of Cypriot archaeology. Each draws together information spread across excavation reports, museum catalogues, sale catalogues, archives and books published over more than a century and in several languages.
We use AI to help build these registers. It reads through the publications, picks out the people, places, dates and objects they mention, and records how they connect, so that a person's page can list the excavations they worked on and the documents that name them. Done by hand, this indexing would take many years.
The information comes from the publications, and each entry cites the sources it rests on so that you can check them. Our editors review the results: they check samples against the original publications, settle the cases the software marks as uncertain or contradictory, and correct the errors they find or that readers report to us. Where the sources disagree, or say nothing, the entry says so; we do not fill gaps with guesses.
Writing and software
In the compendium and our articles, we use AI for editing: to check our texts for errors, consistency and clarity, and to keep them to the site's style.
In our news section, AI is used to write and edit pieces from the original announcements, papers and reports they cite. Every piece is checked by our editors before publication. The short biographies and document summaries in the registers are drafted with AI from the sources they cite and reviewed as described above. All of these texts are held to our editorial standards.
We also use AI tools to help write the software that runs the site, including the reconstruction viewers, the map and the games.
Responsibility and corrections
We are responsible for everything on this site, whatever tools were used to make it. If you find an error, write to us through the contact page or at alexis@ancientcyprus.com, with the address of the page and what you believe is wrong, and we will correct it.