I built this for an AI class in 2002

By B Klug · October 2026

The first version of this site was homework.

In 2002 I was a computer science student at the University of Maryland, College Park, and I was taking the AI class. Just "the AI class," singular. Its course number doesn't even exist anymore, because back then there was one AI class, period, not the whole menu of them they offer now. If you're picturing anything like ChatGPT, stop right there. AI in 2002 was search algorithms, game trees, chess programs, logic puzzles and a whole lot of probability. There was Lisp. There was a lot of Lisp. The hot new thing that year was spam filters that could learn from your own email. Paul Graham published an essay called "A Plan for Spam" that summer, and every nerd I knew had an opinion about it. A computer that could write a decent paragraph was science fiction, and not even the good kind.

The project

My idea was pretty simple. Take a student's paper, pick out the sentences most likely to be unique to whoever wrote it, and ask a search engine whether those exact sentences already existed somewhere. If they did, somebody probably copied.

Was it "intelligent"? Ehh. It was more like teaching a computer to be a suspicious TA. But the hard part was the picking. Search for "the war had many causes" and you'll match ten thousand pages and learn nothing. Search for a long, slightly odd sentence and you either find nothing, or you find the exact website a kid copied from. Teaching a program to tell those apart was the AI-ish part, and honestly, a version of that same idea is still in the code today.

Google had opened up a way for programmers to search it from their own code that spring, with a limit of a thousand searches a day. That felt like a lot, right up until people started actually using the thing.

Then people found it

I figured my professor and maybe a few classmates would ever look at it. Instead, teachers found it. Then students did. Then I started getting email from people I'd never met asking if they could pay for more checks, which, as a broke college student, I was extremely open to. It lived at a web address with /cs/ in it for years, because it really was my computer science homework.

Over the years, more than 26,000 people have subscribed. Most of them signed up before Stripe even existed. (Stripe is the payment company pretty much every website uses now. It launched in 2011, and by then this site was nine years old.) If you never had to take credit cards on a website in the 2000s, count yourself lucky. I could write a whole other article about that, and it would be mostly swearing.

The site has had a few names. It's plagiarism.ink now. The code has been rewritten more or less from scratch, more than once. But the idea is the same one I had sitting in that class.

Still just me

There's no company here, really. No investors, no team of fifty, no sales department. When you email support you're emailing me, and I read it.

I'm a trans woman, and I live in the Midwest now, a long way from College Park in about every sense. I'm older than the kids using this site by a margin I'd rather not do the math on. And I still get a little thrill when someone writes in to say the checker caught something, or, just as often, that it helped them prove a paper was clean.

The weird full circle

This is the part that still makes me laugh. I built a plagiarism checker in an AI class, back when AI couldn't write a sentence. Twenty-four years later, the most common way students cheat is by having an AI write the paper, and the way I check for it is with an AI. My 2002 professor would be either very proud or deeply horrified, and I genuinely can't tell you which.

So this year I added AI detection to the checker. It's not magic, and I'd rather tell you that than have you find out the hard way. I wrote about how much I'd trust it, about the signs I'd look for in a paper myself, about what teachers should do when they suspect AI, and, because the tools get it wrong sometimes, about what students should do when they've been wrongly accused.

A lot has changed since 2002. A few things haven't. Teachers still want to be fair. Students still mostly want to do well and are scared of messing up. And the job of a tool like this is still the same one it had in that classroom: give a person a good reason to look closer. A person still has to decide.

If you've used this site at any point in the last twenty-four years, thank you. Really. You kept a college project alive a lot longer than it had any right to live.

B