This Wikipedia Clone Is Built Entirely on Things That Never Happened

What happens when an encyclopedia stops trying to reflect reality and instead begins generating one of its own, where every search opens not a page of knowledge but a new layer of invention, and where the boundary between reference, fiction, and hallucination is deliberately erased, yet still managed with enough internal consistency that it starts to feel like a world you could, at least in theory, study?

A new Wikipedia-style website is built entirely from AI hallucinations, offering users an endless stream of fabricated knowledge that nonetheless manages to feel strangely coherent.

The project, called Halupedia, presents itself as an “infinite” encyclopedia in which every search term or clicked link triggers a large language model on the backend to generate new content on the fly. The result is a sprawling system of invented facts, presented in the formal tone of a 19th-century scholarly archive.

“Every link leads to an entry that does not exist yet — until you click it,” reads the description posted on GitHub.

From the moment users land on the homepage, the project makes clear that it is an experiment in artificial fabrication. But once inside, the experience can resemble a functioning knowledge base, complete with hyperlinks, citations, and academic-style quotations. Some entries even include footnotes, which are themselves also invented.

The tone is deliberately authoritative. Articles are written as if they belong in a traditional encyclopedia, even as their contents drift into absurdity. One of the most frequently cited entries describes “The Great Pigeon Census of 1887,” an alleged nationwide effort in Britain to count every gold-crested rock dove.

The entry claims the exercise was led by a fictional “Royal Society for Avian Enumeration,” described as an organization dedicated to cataloguing bird populations across the United Kingdom of Great Britain and Ireland. It also introduces a supposed founder, Sir Reginald Featherton, who believed pigeon counts were essential for urban planning and equitable distribution of what the text calls “Parliamentary Crumbs.”

Much like Wikipedia, proper nouns within each article link to other entries, allowing readers to move deeper into the system. Clicking on a name such as the Royal Society for Avian Enumeration opens another fully fabricated page, which then links to more invented people, events, and institutions.

Users can also generate entirely new entries by typing queries into a search box. The system responds with a list of plausible sounding article titles, each loosely connected to the request. A search for a vulgar term, for example, reportedly returned suggestions such as “The Gnomish Mandate of Circular Reasoning.” Selecting any option triggers a message indicating that the system is “resolving a minor scholarly dispute,” before producing a freshly generated article in the same formal tone.

Behind the project is an attempt to make the hallucinations internally consistent. Rather than allowing contradictions to accumulate freely, the developers introduced a structure that treats the encyclopedia as a kind of shared fictional universe. A write forward feature allows future articles to be pre seeded with metadata that defines canonical facts such as dates and relationships.

“The LLM is instructed that the encyclopedia is hallucinated and absurd, but it must not contradict itself,” according to the project’s GitHub documentation.

Even so, consistency is not always maintained. One article on the Royal Society for Avian Enumeration states that it disbanded in 1927, while another entry linked within the system asserts that it dissolved in 1891, exposing the limits of enforcing coherence on generated text.

The developers appear to have embraced the idea that Halupedia is not just a database of fiction, but a self-contained narrative system that evolves through interaction. Each click becomes both a request for information and a prompt that reshapes the universe itself.

Yet, like many open experiments in generative AI, the project has also been vulnerable to misuse. Some user generated entries have included offensive or racist phrasing in their prompts. In most cases, however, the system appears to ignore the intent behind such inputs and instead generates unrelated, polished encyclopedia entries in its characteristic ornate style.

Still, the presence of toxic or adversarial prompting highlights a familiar challenge for generative systems operating without strong content moderation. Even when the output remains superficially neutral, the inputs reveal how easily such tools can be pressured toward harmful territory.

The project has drawn comparisons to other AI driven encyclopedic or knowledge style systems, including more controversial efforts that attempt to redefine what counts as information or authority online. In contrast to those projects, Halupedia does not claim to offer truth at all. Its premise is openly fictional, even playful, but it still raises questions about how easily authoritative language can simulate credibility.

At its core, Halupedia underscores a tension in modern artificial intelligence. Large language models are trained to produce plausible text, not verified fact. When placed inside a system that encourages continuity and citation, the output can resemble scholarship even when it is entirely invented.

That blend of structure and invention is part of the appeal. It produces something that feels like a functioning encyclopedia while quietly eroding the distinction between reference and fiction.

For some observers, the project is a form of commentary on how easily information systems can be mimicked. For others, it is simply an elaborate novelty, a way to explore the boundaries of generative storytelling.

Either way, Halupedia leaves users with a peculiar experience, one in which every fact is both confidently stated and fundamentally uncertain, and where the act of reading becomes indistinguishable from the act of generating.

As with many AI experiments, its significance may lie less in what it claims to know than in what it reveals about the systems producing it.

Stay ahead of the stories shaping our world. Subscribe to Impact Newswire for timely, curated insights on global tech, business, and innovation all in one place.

Dive deeper into the future with the Cause Effect 4.0 Podcast, where we explore the ideas, trends, and technologies driving the global AI conversation.

Got a story to share? Pitch it to us at info@impactnews-wire.com and reach the right audience worldwide