A $50 network of 50,000 fake statistics pages is now cited by ChatGPT and a Qatari bank

None of the numbers on stateglobe.com is real. A researcher generated the site in a day and has watched the citations roll in since.

In brief

Metehan Yesilyurt of Peec AI built about 50,000 pages of invented stats plus 15,000 listicles with GPT-4.1. Google demoted the site after two updates, but ChatGPT kept crawling it, logging over 600,000 bot requests in 30 days.

The fake figures reached an academic paper and investment research from Qatar's largest financial group.

How the experiment worked

A $50 network of 50,000 fake statistics pages is now cited b
A $50 network of 50,000 fake statistics pages is now cited b

Peec AI's write-up describes stateglobe.com: invented statistics on SEO, e-commerce and social media across 200 countries, generated in one day from a country list, topics and two sentences, on an AI web template. Total cost: about $50. Nothing was verified.

Google's visibility collapsed within two updates; pages now get roughly 5 to 20 impressions each. ChatGPT behaved differently: over 600,000 bot requests in 30 days (about 72,000 from ChatGPT-User) and 10,000+ ChatGPT sessions since March. Ahrefs and Semrush show 330+ referring domains, including Shopify, DHL and Bluehost. On 7 October the site ranked second on the author's tracked prompts, behind only Statista. The author stresses it is research, not a playbook.

Can it be fixed?

"It cannot be fixed. The most hilarious example of it is YouTube… the AI version of shitting into the well you drink from." — a commenter describing YouTube's auto-generated video summaries feeding its own search

Another commenter found it strange that chatbots google "What was the inflation rate last quarter" rather than query BLS datasets, and hoped statistics would migrate to first-party sources. A third predicted the lazy fix: hand-written reliability rules ("NYT=Reliable; a Forbes contributor blog=Unreliable") declared sufficient.

A quieter implication: if you cite a statistic, trace it to the original dataset. The experiment shows a citation chain can start from a page nobody checked.

Get breakdowns like this every day

The top Hacker News stories, unpacked — by email, Telegram, or RSS.

Next story · 6 of 12 Read more
🚀Tech

One power failure in one London zone took down HSBC, Revolut and Lloyds

Google Cloud's status page blamed a power event in europe-west2-b. The commenters spent the day asking why banks hadn't survived one zone.

★ 342 · 💬 172 · neowin.net · 1 min