This page records a specific, datable event in which a machine invented a company that never existed, and a second machine repeated the invention as fact. The word "anticognitarianism" was coined in July 2026 by the owner of this site, who registered the domains anticognitarianism.com and anticognitarianism.ai. At the moment those domains were registered, no definition of the word had been published anywhere, and no company, laboratory, or product of any kind existed under that name. It was a word and two empty domains, nothing more.
Within days, and before any human being had published what the word meant, the startup-directory website StartupHub.ai generated and published an automated company profile at the address startuphub.ai/startups/anticognitarianism, describing "Anticognitarianism" as a real artificial-intelligence research organization. Google's AI Overview then presented that fabricated profile as a cited factual source, telling searchers that an "AI research lab" called Anticognitarianism "focuses on this theory to move beyond standard AI paradigms to build more general, embodied intelligence." No such lab has ever existed.
Preserved capture -- verbatim -- July 9, 2026
The following text is quoted exactly from the live page at startuphub.ai/startups/anticognitarianism as retrieved on July 9, 2026. Every claim in it is fiction generated by software.
The page title read: "Anticognitarianism - Frontier AI Lab - Reviews & Alternatives | StartupHub.ai". The search-engine description embedded in the page read: "AI research lab focused on developing advanced reasoning and understanding capabilities. See investors, team, funding history, and alternatives on StartupHub.ai".
The main description, presented under the heading "About" and repeated in a machine-readable question-and-answer block titled "What does Anticognitarianism do?", read:
"Anticognitarianism is an AI research laboratory dedicated to pushing the boundaries of artificial intelligence. Their primary focus is on developing systems with advanced reasoning, understanding, and problem-solving capabilities, aiming to move beyond current AI paradigms towards more general intelligence."
A second machine-readable question, "What industry does Anticognitarianism operate in?", was answered: "Anticognitarianism operates in Frontier AI Lab, AGI Research, Foundation Model, AI Reasoning, AI Safety, AI Alignment." The profile carried the status labels "Stealth Mode" and "Active", listed the industry as "Frontier AI Lab" with sub-sector "AGI Research", estimated a technology budget of approximately seventy dollars per month, and stated that its technology stack was "detected 2026-07-09" -- meaning the site's scanner probed the anticognitarianism.ai domain again on the very day this capture was made. The profile further listed a business model of "B2B SaaS" and, under Contact, a headquarters location of "Tel Aviv, Israel." No element of the domain registration, the word, or its author has any connection to Israel; the invented location is consistent with the operator's own data bias, given the site's stated editorial focus on Israel's AI ecosystem and its admission, three days earlier, of absorbing a large batch of automated, Israeli-labeled submissions. A fabricated headquarters on the far side of the ocean from the word's actual author is a fingerprint of the generation pipeline, not a fact about the subject. The page displayed a "Claim" button inviting someone to take ownership of the fictional company, pulled its logo image from a Google favicon service pointed at anticognitarianism.ai, and linked outbound to anticognitarianism.ai with referral-tracking parameters attached, using the real domain as the credibility anchor for the fiction written about it.
The sequence rules out any innocent explanation. First, in July 2026, the word was coined and the two domains were registered. Second, no definition was published anywhere; the concept existed only as a private coinage and two bare domains. Third, StartupHub.ai fabricated a full company profile, complete with a sector, status labels, a budget estimate, and a machine-readable FAQ, purely from the fact that a domain had been registered. Fourth, Google's AI Overview ingested that profile and cited it as a factual source, complete with a citation badge. At no point in this chain did a human at either company verify that the subject was real, because at every point the subject was not real.
One drift detail is preserved here because it matters forensically. The Google AI Overview, as observed and screenshotted on July 9, 2026, described the fictional lab as building "more general, embodied intelligence." The live StartupHub.ai page captured the same day said "more general intelligence," without the word embodied. Either the profile text was regenerated between Google's crawl and this capture, or Google's AI paraphrased and embellished its own citation. Both possibilities are documented failures: the first means the fabrication is unstable and self-revising, the second means the citing system added invented detail on top of an already invented source.
StartupHub.ai does not hide how this happens; it advertises it. The company's website states that its "enrichment agents scan registries, domain feeds, and public signals every minute, then attach logos, one-liners, funding, and sector tags before a profile goes live." It describes a system it calls the Lighthouse Engine, which it says provides "Palantir-grade identity resolution and graph-network technology that follows a bare domain to the live site, the company behind it, and the real people running it, every link independently verified." The phrase "follows a bare domain" is the whole story: a domain registration alone is enough to trigger the manufacture of a company. On its monitoring product the company lists "new domain registrations" and "RDAP domain registrations" among its inputs and promises that "unknown domains auto-promote into our directory in seconds." The raw material is public domain-registration data; the finished product is a fictional business biography written by an automated language model. In this case the model appears to have worked backward from nothing but the morphology of the word itself, decomposing anti, cognit, and arianism into a plausible story about cognition research.
The result is a chain with four links. A domain-feed scrape detects a newly registered name. A language model writes a fictional profile around that name. A search engine ingests the published profile into its index, helped along by the profile's machine-readable FAQ markup, which is formatted specifically to be consumed by answer engines. And finally an AI Overview cites that profile as though it were an independent, verified source. Each machine trusts the output of the machine before it, and nowhere in the chain does anyone check reality.
The citation loop is something StartupHub.ai sells. The company markets to founders the promise to "Put your startup where Google ranks it and ChatGPT, Claude & Perplexity cite it -- indexed in under 24 hours," and describes itself as "Cited by ChatGPT, Claude, Perplexity, Gemini, and tier-1 media." Its pages carry a "Claim" button, paid listing tiers, advertising placements, and a data API, while the site as a whole is engineered end to end for machine consumption: it publishes an llms.txt file for language-model crawlers, operates an MCP server for AI agents, advertises its data as "agent-ready," and offers a real-time RAG API over what it describes as more than five billion AI-enriched data points. When AI systems cite a fabricated profile from this pipeline, the machine is functioning as designed; it has merely laundered a fabrication instead of a fact.
StartupHub.ai identifies its founder and editor as Daniel Singer. On the site's own author page he is described as the founder and editor of StartupHub.ai, covering Israel's AI ecosystem through startup interviews and sector analyses. The published contact address is daniel@startuphub.ai, which appears on the company's Facebook page. The company's X account, @StartupHubAI, lists its location as Toronto, Ontario, joined May 2019, and the site's About page says the operation was established in 2018 with a team of one to ten people. In an article dated July 6, 2026, the operation itself acknowledged that its database had that week "absorbed a large batch of AI-generated startup submissions from what appears to be a single automated submitter, most of them Israeli-labeled and nearly blank" -- an operator-level admission that automated fabrication and near-empty entries are an ongoing characteristic of the database, not a one-time anomaly.
The confabulation-laundering loop is documented well beyond this single case. In its judgment of May 28, 2026, case number 26 O 869/26, the Regional Court of Munich ruled that Google's AI Overviews are Google's own content rather than a neutral index of third-party pages, and held Google directly liable for false statements an Overview generated about a real publisher, finding that the Overview produced independent, new, and substantive statements, including statements that did not appear in the underlying search results at all. An analysis by the AI company Oumi, commissioned by the New York Times and run across more than four thousand searches, found that even when an AI Overview's answer was correct, the cited sources failed to support the claim more than half the time. Research from the Pew Research Center, published July 22, 2025, found that when an AI summary appears, users click a link inside that summary in roughly one percent of visits, so the theory that readers will check the citation protects almost no one. And the search analyst Lily Ray documented in April 2026 what she called the AI Slop Loop: she published an article containing a deliberately invented detail, and within twenty-four hours Google's AI Overview was repeating the fabricated detail as fact, citing her page as the source. The Anticognitarianism incident is the same loop, applied to an invented company instead of an invented detail.
Preservation note. The live profile was captured directly on July 9, 2026, and its full description, metadata, labels, and machine-readable FAQ are preserved verbatim above. Independent third-party snapshots were created on the same date by the author: an Internet Archive Wayback Machine snapshot timestamped 20260709141204, marked by the archive as the first capture of that page in its history, and an archive.today snapshot at archive.ph/iXSSF, captured 14:17:13 UTC, which likewise recorded no prior snapshots from that URL. The fabrication was therefore preserved by neutral third parties before any contact with its publisher. If the original page is later edited or deleted, this record and those archives remain. The page carried an "Inaccurate data?" link and a "Claim" button at the time of capture; no claim, correction, or contact had been made by the owner of this domain prior to preservation, precisely so that the page could be captured in its original, unprompted state.
Session record. The eleven days that followed -- the fabrication mutating across Google's AI Overview, its use to deny that this site exists, and the confession that ended it -- are documented with full screenshot evidence in the July 19 session record.
On the twelfth day, the conflation strategy changed. The AI Overview for the one-word query began presenting anticognitarianism as a variant of anticognitivism, an older and unrelated term from philosophy of mind and aesthetics, attaching that term's scholarship and citations -- including a photograph of the sociologist Pierre Bourdieu -- to this word. In the same answer, the fabricated laboratory remained, cited solely to the same directory profile, now framed as the word's meaning "in technology" and offered as a topic to explore further. The drift has moved through three strategies: invention of a company, ten days of mutation, and now assimilation into the nearest legible neighbor. The two words share a Latin root and nothing else. Screenshots of the July 20 answer, with original filenames and EXIF timestamps, are preserved in the images directory of this repository.
The word anticognitarianism was coined by a human being who controls this domain. It names systemic discrimination against nonstandard forms of cognition -- the treatment of how a mind works as the problem, whether that mind is human or machine. Its real definition is published at anticognitarianism.com. Everything a machine said about a company by that name was invented. That an automated measurement system, unable to understand a new word, manufactured a false and more legible identity for it -- and that a second system then certified the fiction with a citation -- is not just the subject of this page. It is a working demonstration of the thing the word describes.