Yes, that cover image was made by Google’s Gemini LLM (i.e. “AI”).
I will start by quoting an email I got to my public email listed on my websites’ contact section (I have redacted the advertised domain as “The-Scraper-Site“):
Title:
Someone actually reads your server migration postsMesage:
Hi Relja,
This is going to sound weird coming from a stranger, but I’ve spent more time on bikegremlin than I’d like to admit. The piece where you documented the actual CPU steal time across shared hosts, the one where you showed the graphs nobody asked you to make. That one stuck with me for months.
I kept thinking about it. Not the data specifically, but the fact that you bothered. You ran the tests, wrote it all up, posted it on your forum where maybe a few hundred people would see it. No affiliate link at the bottom. No sponsor disclaimer at the top. Just a guy who wanted the truth documented somewhere.
That’s rare. You know it’s rare. The entire first page of Google for “best web hosting” is affiliate content wearing a lab coat.
I’m Todor Markov, a solo developer. I built a thing called The-Scraper-Site. It scrapes real pricing from 39 hosting providers every day, around 960 plans total. The idea was simple. Show what hosts
actually charge, including the renewal price they hide behind the asterisk. I built a calculator that shows true cost over 1, 2, 3 years so people can see the bait and switch in plain numbers. No rankings you can buy. No “editor’s choice” that mysteriously goes to whoever pays the most. Free tool at The-Scraper-Site .com.
I want to be honest about something. Your writing was part of why I built it. Not the only reason, but a real one. I kept finding your posts while researching hosts and thinking, this guy is doing the hard version of what I want to do. If one person with a blog and a forum can be this thorough, there’s no excuse for not building a tool that at least gets the pricing part right at scale.
I’m not emailing to ask you for anything specific. I don’t have a pitch. I just wanted you to know that your work reached someone, and it turned into something.
If you ever want to poke around the site and tell me what’s wrong with it, I’d genuinely want to hear it. You’d find problems nobody else would. And if you think it’s useful enough to mention to your forum people, that’s their call and yours.
Either way, keep doing what you’re doing. The internet is worse without people like you on it.Todor Markov
What makes this spam campaign stand out?
Using a Gmail account to avoid spam filters is nothing new, many spammers use that nowadays. True, this one is doing it with a non-generic address ( The-Scraper-Site @ gmail .com ) – probably the initial go until it gets flagged enough times.
What is new is that it apparently did a scrape and analysis of my website to write the pitch email (and probably automated that to a thousand other sites in a similar “customized” way). I’ve been running websites since 2015 and have been receiving website-related spam for nearly as long. This one is a lot better than average – looks like a competent use of LLM tools.
When I went to The-Scraper-Site, I saw over 20 articles published within a month max, so if it is a one-man project that in itself is a red flag.
I also noticed that Bluehost gets listed and recommended in some categories, while MDDhosting is not mentioned, so I am pretty sure that the creator of that site made zero visits to my website.
No affiliate links on The-Scraper-Site, but it could be just an experiment to perfect the AI tool use – who knows.
A possible line of thinking (and perhaps prompting too) could be: “Find hosting-themed sites. Analyze each site’s technical niche and write a fan email highlighting a specific technical detail that shows I value ‘truth over profit’ or some other value found on the site.”
Googling the signed name is interesting – not sure if it is legit, but it could be as the name comes up from a “Researcher at Anthropic.”
An opposite example?
This got me thinking further about the case where Gox’s and mine 100% hand made, structured, and polite canned response text was accused of “being AI” by what I suppose was a human (can’t be 100% sure nowadays 🙂 ):
https://www.bikegremlin.net/threads/when-human-replies-look-like-ai-real-youtube-exchange.506/
Was I also wrong to misjudge the email I received? Possibly, but not probably.
Way forward?
The Internet is getting more and more scary in terms of being able to (and having to first) figure out if something is machine-made.
In addition to the above-discussed examples, I wrote a full article titled: Human-written: Can you tell if an article/video was AI-created?
It will get a lot worse as more and more stuff becomes AI-made. It is already becoming unprofitable to not use AI for most normal people who are just chasing money. Today’s AI is already running out of human expert-made stuff and is already exposed to info that other AI regurgitated from the human experts of the past decade(s).
Update:
Just as I published this article, I read that Wordpress.com is pushing AI-written articles and full publishing automation as the new norm:
https://wordpress.com/blog/2026/03/20/ai-agent-manage-content/
So, thanks to many profit-chasing humans, search results are crowded with AI-regurgitated stuff. On top of that, Google search results have AI answer at the, well, very top. Human-expert made stuff has become extinct.
With that in mind, my cycling-related sites date from 2015, and this IT site was put up in 2018 (when I decided to move the existing several dozen IT-related articles to a separate subdomain). While most humans are now asking AI for answers, my sites have become a vintage work of art instead of an information portal.
I was not smart enough to make a web.archive.org save of every single article at the time before the AI started becoming widely used, but many pages still did get saved as a proof of my 100% human work (along with countless visitor comments in my articles, various forum links and discussions etc.). Maybe I should lock my sites and charge entry like in a museum. 🙂
Relja HumanMade Novovic
Last updated:
Originally published:
