Fake AI wikis for games spread invented facts and crowd out real fan sites

Eurogamer has documented a wave of AI-generated websites posing as fan wikis for video games, filling search results with invented characters, skills and lore while crowding out the volunteer-run encyclopedias players actually rely on.

The report grew out of a routine lookup. A Eurogamer writer searching for basic details about companion characters in the game Nighthawks landed on a site calling itself the Nighthawks Wiki. It was not editable the way wikis normally are, accepting corrections only by email, and it listed companions and skills that do not exist in the game. It also blended in unrelated material about the 1981 film and the Edward Hopper painting that share the name, citing both as sources. The writer traced the domain to Spaceship, a website building service that offers an AI website builder, and a quick search turned up many more sites of the same kind. Some were genuinely editable; others were wikis in name only. They clustered around indie games with modest audiences, precisely the games where a player might search for help and find no established community resource waiting.

A volume business built on thin margins

According to Eurogamer, the publisher Hooded Horse creates wikis for its own games in collaboration with fans, and CEO Tim Bender describes playing "whack-a-mole" with AI-generated competitors. One such site covered Hooded Horse's game Corsair Cove but belonged to a scattershot collection of more than 20 game wikis operated under the GameDB banner, whose creator admitted to building them with generative AI.

Bender's assessment, as reported by Eurogamer, is that generative AI lets commercial operators mass-produce sites and flood a game's search results to capture its traffic. The payoff per site can be tiny, but even a trickle of ad revenue justifies the effort when production costs almost nothing. He also suggests some sites may be after more than ads, such as data collected from visitors. Eurogamer found at least one AI wiki steering people toward downloading poorly reviewed subscription software, which the writer's malware protection flagged as "potentially unsafe". A Eurogamer editor also saw the Nighthawks wiki repeatedly redirect to spam URLs, though the article's author could not reproduce the behaviour. Emails sent to the sites for comment either bounced or went unanswered.

FlamingMojo, a member of the Independent Game Wiki Network, told Eurogamer that at least five AI wikis appeared for the 100-player shooter Wardogs within days of its September release. The harm is reputational as well as competitive: he says players who encounter false information blame the legitimate wikis that have not yet documented a topic, and then abandon wikis altogether.

Scrapers raise the cost of staying legitimate

Independent wikis are also absorbing heavy traffic from AI crawlers. Public internal data from the Dune: Awakening Community Wiki, which FlamingMojo runs and pays for out of his own pocket without advertising, shows 7 million crawler requests in a 30-day period, 5 million of them from Meta alone. He says Meta's crawler hits pages multiple times per second with no regard for sensible rate limits, while other scrapers can at least be steered by configuration files toward real pages. According to Eurogamer, his server costs have risen between threefold and tenfold because of generative AI, arriving at the same time as AI-generated sites compete with him for search visibility.

All of this plays out in an ecosystem already under strain. Fandom dominates fan-wiki hosting and has itself had to roll back AI features over their inaccuracies and unpopularity, while groups that leave such platforms often struggle to rank in Google results beneath the large umbrella sites. Volunteers are rarely paid either way, sacrificing ad revenue to platforms or covering their own hosting. The Dune wiki has responded by gamifying contributions, offering in-game rewards in affiliation with the developer.

Why it matters

The story is a compact example of a data-provenance problem. AI systems scrape volunteer-built references, cheap generated copies then outrank the originals in search, and the people who produced the underlying knowledge face higher costs and shrinking attention. Bender frames it to Eurogamer as one instance of a much broader struggle in which bad actors can easily crowd out information that dedicated people built over time. For anyone training AI tools on the open web, and for readers who click the first search result, telling the real wiki from the fake one is getting harder, and the consequences now include not just wrong facts but flagged downloads and spam redirects.