Eurogamer says LLM-written "wikis" for video games are now showing up at the top of Google, often stuffed with made-up details, and are crowding out the volunteer-run wikis players actually rely on. The report, published October 27, 2026 and picked up on Hacker News on November 2, quotes Hooded Horse CEO Tim Bender and Independent Game Wiki Network member FlamingMojo. It is a small, concrete case study of a much bigger problem: when generating a plausible website costs almost nothing, the search results people trust fill with plausible nonsense. It also fits a pattern we have followed in how AI is eating the web's collective memory and in manufactured sources that AI recommendation engines cite.
This post separates what Eurogamer reports from what we can and cannot confirm, adds the context around Fandom, Wikipedia and AI crawlers, and ends with practical checks for readers and for anyone running a community wiki.
TL;DR: what is reported and how solid it is
| Question | Answer | Confidence |
|---|---|---|
| What happened? | AI-built sites styled as wikis rank first for indie game queries and carry false information | Reported by one outlet, with named sources |
| Which sites? | A "Nighthawks Wiki" and a 20-plus game collection called GameDB; five or more sites for the game Wardogs | Reporter's own examples |
| Who says it is AI? | One creator admitted generative AI use; the reporter inferred it for others and says so | Mixed: one admission, others inferred |
| Is there a safety angle? | One site pushed a paid-subscription download that the reporter's malware tool flagged | Single reported case |
| Are real wikis hurt? | Yes: lost search traffic, reputation damage, crawler load, hosting costs up three- to ten-fold | Maintainers own figures |
| Is there a fix? | Community ownership, publisher-run wikis, and better search ranking signals | Opinion, not a shipped solution |
What Eurogamer actually found
The author set out to check basic facts about companion characters in the game Nighthawks and clicked the first Google result, a site titled the "Nighthawks Wiki". Per the Eurogamer report, it was not a wiki at all: nobody could edit it, and the only way to submit a "correction" was by email. It listed companions and skills that do not exist, and it mixed in details about the 1981 film and the Edward Hopper painting that share the name, even citing them as "sources" on pages where they had no relevance.
The reporter is careful about attribution. She writes that she does not like to accuse anyone of using generative AI because its hallmarks are hard to pin down, but traced the domain to a website-building service that also sells an "AI Website builder". The email address on the site bounced. A quick search then turned up many similar sites, some genuinely editable and some only calling themselves wikis. They tend to target indie games without huge communities, the kind where players search for help and find no thriving forum waiting.
That targeting is the economic insight. A big franchise already has a well-defended wiki and a loud community. A game with a few thousand players has neither, so the first page that answers "who are the companions" wins by default, true or not.
The publisher view: whack-a-mole
Hooded Horse is unusual because it builds wikis for its own games with fans, rather than leaving the space to hosts. Its Corsair Cove official wiki is an example of the model. CEO Tim Bender told Eurogamer the company now plays "whack-a-mole" with AI-generated competitors.
One of those, according to Eurogamer, appeared for Corsair Cove but belonged to a collection of more than 20 games run by a site called GameDB, whose creator admitted to using generative AI. Bender's description of the incentive is blunt: generative AI "enables a run of basically commercial enterprises to mass produce and try to flood the search engine results of games to soak up their traffic". He adds that players who stumble onto these pages end up confused, and that the company hears about it.
Eurogamer says its attempts to reach GameDB through the public contact address failed with a "remote server is misconfigured" error. We could not find independent coverage of GameDB or this specific Corsair Cove listing in our own searches, so treat the specifics as the reporter's findings with named, on-record sources rather than as broadly confirmed. That is a normal standard for a single-outlet story, and it is why we cite the original directly.
Wardogs and the speed of the flood
The Independent Game Wiki Network's FlamingMojo told Eurogamer that at least five AI-made sites appeared in the days after the 100-player shooter Wardogs launched in September. Speed matters here. A human community needs time to write pages, and in that gap a generated site with confident, wrong answers can lock in the top search spot.
FlamingMojo says the effect is reputational for real wikis: players are misinformed by a fake, then "bounce off" wikis entirely when the legitimate page has not been written yet. The harm lands on volunteers who were not the cause.
The safety angle: tracking and software downloads
Bender raised a second concern: "Some of them, you don't know what data they're collecting." In at least one case Eurogamer found, a generative-AI site directed visitors to download additional software with a paid subscription. The reporter says that software was poorly reviewed online and that opening the site made her malware software warn that it was flagged "potentially unsafe".
An editor's note adds a wrinkle worth repeating: when a Eurogamer editor tried the Nighthawks wiki, it repeatedly redirected to spam URLs, though the author could not reproduce this. That is one inconsistent observation, so we do not claim the sites are malware distributors. What it does show is that the same cheap-site economics that produce low-quality text also suit scam operators. If you need a refresher on staying safe from sites like this, see our guide to protecting yourself and your family online.
The other squeeze: AI crawlers on real wikis
The report also describes a double burden. Independent wikis now absorb traffic from AI scrapers. Eurogamer cites public internal data from the Dune: Awakening Community Wiki showing 7 million requests from AI crawlers in a 30-day period, 5 million of them from Meta. FlamingMojo says he can tame most scrapers with files that steer them to real pages, but that Meta "hammers every page multiple times per second" with no sensible rate limits. He puts the rise in server costs at between three- and ten-fold.
We have not independently verified those numbers, and Meta was not quoted in the excerpt we read. They are the maintainer's figures from the wiki's own public statistics, which is exactly how such claims should be labeled. The broader question of whether AI companies should honor sites' crawl preferences is live; we looked at one side of it in OpenAI's stance on robots.txt-style training signals, and publishers have threatened action over lost traffic in our coverage of Google Zero and crawler blocking.
A magnifying glass resolving scattered shapes into an orderly grid, representing checking AI-generated wiki claims against reliable sources
Why fan wikis were already fragile
Eurogamer notes that fan wikis were a contested part of games culture before generative AI. Hosting sites such as Fandom dominate the space, and groups that leave struggle to rank in Google against the umbrella sites. Fans rarely make money from the work: they either give up ad revenue to the big platforms or pay for their own hosting. FlamingMojo, for example, pays for the Dune: Awakening Community Wiki out of his own pocket and runs no ads.
Eurogamer also says Fandom introduced AI features that had to be rolled back for inaccuracy and unpopularity. We could not find a primary Fandom statement confirming a formal rollback; community accounts from 2024 describe backlash over inaccuracies in its AI "Quick Answers" feature, so we treat the rollback wording as Eurogamer's characterization. Separately, a July 2026 Plagiarism Today account of the GTA Wiki dispute shows how fraught the relationship between Fandom and its volunteer editors has become. That article does not allege AI was used to copy the GTA content, so it is background on trust, not evidence for the AI-wiki claim.
The same dynamic played out at a larger scale on Wikipedia. In June 2025 the Wikimedia Foundation paused a trial of AI-generated article summaries after editors objected strongly, with one warning of "immediate and irreversible harm" to readers and the site's reputation. More recently Wikipedia's volunteers have built defenses of their own, which we covered in how WikiProject AI Cleanup flags AI-written articles. Fan wikis have no such organized immune system.
Why the economics favor the fakes
Bender's argument is that low profitability makes things worse, not better: "There's not that much ad revenue, but there's going to be a little bit and that's all it takes." When publishing a page costs fractions of a cent, a trickle of ad income is enough to justify thousands of pages. A human wiki that earns nothing has to be funded by love.
This is a general pattern, not a games quirk. Bender himself frames it as "one example of a much broader struggle": tools have made it easy for bad actors to flood the zone and crowd out information that people spent real time making. Open-source maintainers have reacted with rules about AI contributions, such as QGIS banning AI-assisted contributions. The defensive instinct is the same: raise the cost of low-effort content.
What developers are saying
The Hacker News thread is small (33 points and 11 comments when we read it) but the comments show the range of reactions. One commenter, skhr0680, argued Fandom had been a poor host long before AI. jacquesm said AI is degrading carefully curated material across the board, from codebases to literature. xlayn described grabbing copies of Wikipedia and OpenStreetMap from before the AI boom as a hedge against contaminated sources. And jonathanberger asked what the real solution is, suggesting spaces where all contributed content is human-written, even if the only enforcement is friction that makes pasting LLM output hard. These are individual opinions, not findings.
What it means for people who build with AI
This matters beyond games for three reasons.
Retrieval quality. If you build agents or RAG systems that browse the web, these sites are exactly what your retriever will find. A fluent, confident, wrong page can beat the correct one on rank. Source allow-lists, publisher-first retrieval and citation checks help; our explainer on AI hallucination and how to catch it covers the verification habits.
Training loops. Text produced by one model and published as a "wiki" becomes input for the next crawl. That is the contamination loop xlayn alludes to, and it is one reason curated, human-edited sources keep their value.
Cost externalities. Crawler load is a cost the scraped party pays. Teams shipping crawlers should honor robots rules, rate-limit, and identify themselves.
How to check a wiki before you trust it
- Look for an edit button, page history and recent changes. Real wikis have them; many AI-built sites offer only an email address.
- Read the "sources". Irrelevant citations, such as a film for a game, are a red flag, as in the Nighthawks example.
- Start from the publisher. Steam pages, official Discords and publisher sites usually link the community wiki that is actually maintained.
- Be wary of any wiki that asks you to download software or redirects you elsewhere.
- If you run a community wiki, publish clear crawler rules, rate-limit by user agent, and consider reward systems. The Dune: Awakening wiki, per Eurogamer, gamified contributions with the developer giving out in-game rewards.
For teams running AI agents that browse unfamiliar sites, an agent that follows a fake wiki's "download this" prompt is a genuine risk. AgentBeam, the agent security platform from the explainx.ai team, is built to stop agents before they take dangerous actions like that.
What we do not know
We do not know how many of these sites exist, who runs them, or how much traffic they actually take. Eurogamer did not get comment from any of the operators it contacted. The Meta crawler figures come from a maintainer, and the reporter could not reproduce the spam redirect an editor saw. If any operator or platform responds, we will add an update.
FlamingMojo's closing line is the most hopeful in the piece: "The way we survive the AI pollution is with community." It is a modest answer to a structural problem, but it is also the one that has worked for Wikipedia so far.
Details reflect Eurogamer's October 27, 2026 report and the sources linked above as of November 2, 2026.