Creator Profile Avatar for Suganthan Mohanadasan
Suganthan Mohanadasan
Creator Profile Avatar for Suganthan Mohanadasan

Suganthan Mohanadasan

18.2K total followers
Profile generated by Google

Latest posts

ChatGPT now tells you its shortlist before it searches. There's a new Activity panel in the interface. Ask it "best web hosting" and the reasoning step shows up like this. "My likely shortlist is Kinsta, Rocket, Cloudways, SiteGround, and Hostinger." Then it goes and searches. I tried again with "best web hosting companies". That one didn't search at all. It named 12 hosting brands from memory and called that a shortlist itself. So the search isn't how it finds the candidates. It's how it checks the ones it already had. It doesn't happen on every query. Which means the real work is getting into that sentence. Is your brand written about, reviewed and compared enough for that association to exist in the training data? Adding this to fanoutfox.com so you can read it on any answer. Coming soon.

X (Twitter)
X (Twitter)
3 days

The July Common Crawl has 2.14 billion pages in it. 3 of them are from bbc.com, and all 3 are copies of robots.txt. The crawler asked for permission and got told no. So one of the biggest news sites on earth is missing from AI training data. The BBC did that on purpose. About half a million other sites also block CCBot in robots.txt, and nobody knows how many of them did it on purpose. Cloudflare has blocked AI crawlers by default on every new domain since July 2025. Its managed robots.txt is on 3.8 million sites that never wrote one. Ad plugins add blocklists. A lot of sites are blocked and nobody ever decided. Most LLM training data starts as a Common Crawl snapshot. Mozilla checked the LLMs released between 2019 and 2023 and 64% trained on Common Crawl data. GPT-3 was about 60% filtered Common Crawl. FineWeb added six 2025 crawls last summer, so this is still how models get built. In June, Common Crawl published a guide called The AI Visibility Audit. 18 pages on how to check your

X (Twitter)
X (Twitter)
1 wk

FanoutFox is a free generative engine optimization (GEO) tool and ChatGPT rank tracker. It reads your own ChatGPT session and shows the fan-out queries, the sources it fetched, who got cited, who got mentioned, and who beat you on each claim. Free, local, one click. For the last 2 teardowns on my blog I worked out how ChatGPT picks its sources by reading its raw network traffic by hand. It works, and it takes the better part of an afternoon per answer. So I built the whole recipe into a Chrome extension. It’s called FanoutFox, it’s free, and it’s live on the Chrome Web Store. (Please visit the website for updated information on this) What is FanoutFox? A free Chrome extension that reads your own ChatGPT session and shows you how the answer was built. The searches it actually ran, the pages it fetched, who it cited, who it only mentioned, and who beat you on each claim. Depending on the job you need doing, it’s a generative engine optimization (GEO) tool, a ChatGPT SEO tool, or an AI

X (Twitter)
X (Twitter)
2 wks