Stateglobe.com, the 50,000‑stat site, continues to be cited by ChatGPT months after Google deindexed it.
FAKE STATISTICS SITE GAINING TRACTION IN CHATGPT DESPITE GOOGLE BAN
*Metehan Yesilyurt’s 50,000‑stat experiment shows AI can amplify fabricated data long after search engines erase it. The fallout reshapes how brands guard against misinformation in LLM outputs.*
Metehan Yesilyurt, a GEO researcher, launched stateglobe.com in March 2026 with a single purpose: flood AI with fabricated data. He generated 50,000 fake statistics in one day using GPT‑4.1, spent about $50 on hosting and content, and let the site sit on the open web. Within weeks Google indexed the pages, but two algorithm updates erased its visibility, leaving the site invisible to traditional search.
Despite Google’s purge, the site has become a hotbed for LLM retrieval. Cloudflare recorded over 600,000 bot requests in the last month, 72,000 of them from the ChatGPT‑User agent that fetches pages in real time. The site has attracted more than 10,000 AI sessions, 330+ external links—including Shopify, DHL and Bluehost—and citations in an academic paper and a bank’s investment research.
The experiment reveals a stark asymmetry: search engines can de‑index a site, but LLMs can keep it alive in their answer pipelines for months, reshaping how misinformation spreads in AI‑driven ecosystems.
How the fake‑stat site was built
In March 2026 Yesilyurt bought the domain stateglobe.com and used GPT‑4.1 to generate 50,000 statistics pages in a single day. The pages claimed to cover SEO, content marketing, social media, e‑commerce and web technology across 200 countries, but every number was invented. He added 15,000 listicles later, spending roughly $30 for the site and $20 for the listicles—about $50 total. The site launched with a basic AI web design template and was indexed by Google within a week, showing healthy early traffic.
Why Google dropped the site but ChatGPT kept it
Two Google updates hit within days of launch, collapsing the site’s visibility. Index impressions fell to five‑to‑twenty per page and the site has been invisible on Google for many months. By contrast, ChatGPT’s retrieval agent continued to fetch pages. Cloudflare logs show over 600,000 bot requests in the last 30 days, with 72,000 coming from the ChatGPT‑User agent alone. These are real‑time fetches that power answers, not training crawls. The divergence highlights how LLMs rely on their own indexes independent of traditional search rankings.
Why are fake stats still surfacing in ChatGPT answers?
ChatGPT searches for seemingly reliable data when answering time‑sensitive queries. Statistics fill knowledge gaps, and the model treats any indexed page as a source. Because stateglobe.com remains in the ChatGPT/OpenAI index, the model pulls its fabricated numbers whenever a prompt matches the site’s topics. The experiment recorded more than 10,000 AI sessions directed to the site, and GA4 shows 10,000+ sessions from ChatGPT alone. The model’s reliance on indexed content, not on verification, lets false data persist in user‑facing answers.
What the fallout means for AI‑driven research
The site now boasts 330+ referring domains, including Shopify, DHL and Bluehost, and its numbers appear in an academic paper and a bank’s investment research. On October 7, stateglobe.com ranked second for visibility on specific prompts, behind only Statista. The experiment proves that once LLMs ingest a source, removal from Google does not erase its influence. Brands must monitor LLM citation pathways, not just search engine rankings, to protect data integrity in AI‑augmented workflows.
Frequently asked questions
How many fake statistics were created on stateglobe.com?
The site was seeded with 50,000 fabricated statistics generated in a single day.
How many domains link to the fake‑stat site?
According to Ahrefs and Semrush, more than 330 referring domains, including Shopify, DHL and Bluehost, link to the site.
Did Google continue to index the site after launch?
Google indexed the site initially, but two updates shortly after launch collapsed its visibility, leaving it invisible for many months.
Yesilyurt’s stunt forces a reckoning for anyone relying on LLMs for factual output. The persistence of fabricated stats in ChatGPT answers shows that AI models will continue to surface data from any indexed source, regardless of its credibility. As more brands discover that their misinformation can survive beyond search engine control, the pressure will mount on LLM providers to implement stronger source verification. The next wave of AI governance will likely focus on cleaning the model’s index, not just polishing SERP rankings.
Sources: Hacker News article "Publish fake stats, get cited in LLMs" by Metehan Yesilyurt, Oct 7 2026