Summary

The article discusses the inefficacy of current AI-detection tools in reliably identifying AI-generated text, despite the ongoing arms race between AI content creation and detection.

Abstract

The article titled "AI-Detection Tools: Can We Uncover Who’s Grifting?" delves into the current landscape of AI-generated content and the tools designed to detect it. The author expresses frustration over the undisclosed use of AI in writing and the subsequent race to develop effective detection tools. After testing a dozen AI-detection tools, the author concludes that none provide consistent and reliable results, with the same text often yielding contradictory outcomes from different tools. The article also touches on the potential ulterior motives of some tool providers, suggesting that ad revenue and data collection could be primary goals over accuracy. Furthermore, it highlights the existence of tools designed to "humanize" AI content to evade detection, warning users about the risks of trusting such services.

Opinions

The author believes that the current state of AI-detection tools is insufficient and unreliable for accurately identifying AI-generated content.
There is skepticism about the intentions of some AI-detection tool providers, with suspicions that they may be more interested in ad revenue and data harvesting than providing a genuine service.
The author suggests that AI tools themselves might be necessary to detect AI-crafted content but indicates that the current offerings do not meet the mark.
A warning is issued to those using "AI humanizer" tools, as these services could potentially record and misuse the content provided by users.
The author implies a sense of resignation, advising readers to be cautious and not to rely heavily on the existing AI-detection tools.

The AI vs. HI Arms Race

AI-Detection Tools: Can We Uncover Who’s Grifting?

What you need to know about AI-detection tools — they’re not what you may be thinking

**Detect and humanize, huh? The war is on** | Screenshot from undetectable.ai

We human authors are angry about the undisclosed use of AI in stories here.

It’s clear to me we’re in a raging arms race with ever more sophisticated AI-generated text versus AI-detection tools.

Is there no one to save the day?

It seems Medium can’t help us. Factions are forming on both sides of whether we can win this fight with just our brainpower. See, for example, Seda Anbarcı’s story on Sentence Fragments that Scream AI “Landscape, tapestry, intricate dance of, captivating narratives….”

A few weeks ago, I suggested that it’s down to the tools. That is, it will take AI tools to help us detect AI-crafted content. I’ve spent the intervening time testing my hypothesis.

The conclusion, upfront

Having worked now with a dozen AI-detection tools, my conclusion is this: no one has a reliable tool to detect AI-written work. At least, none that I have been able to find.

CarolF, you were right and I was wrong!

I originally planned to do a review of the leading tools and tell you the pros and cons of each. I’ll list them at the end for the curious, but save yourself the time. In my experience, they simply don’t work reliably.

How do I know? Simple:

I ran samples of known original content and known AI-generated content through each of the detectors.
Every time, I got answers ranging from guaranteed human to guaranteed AI (and everything in between) on the exact same text.

So what’s the game?

A career spent practicing law taught me to be skeptical about people’s motives. Watch what they do, not what they say.

Quite a few of the AI-detection tools have home pages festooned with ads. Your time spent there is delivering ad revenue to them.

That’s not the worst that I suspect some companies are doing with the text we’re inputting into their tools.

The race never ends (AI Humanizers)

Search for “AI detection tools” and you’ll also be presented with a bevy of tools to help you “humanize” your AI content and avoid detection.

Here are a few prominently featured companies offering help in disguising your AI-generated content to fool the AI detection tools.

Undetectable.ai

**I wish their stories were undetectable** | Screenshot from contentatscale.ai/ai-content-detector

Stealthwriter.ai

Here’s a bit of advice for the unwary credulous. These companies have a record of everyone who puts content in to be “humanized” and all the output they generate. You’re giving them a pirate’s map of your misdeeds.

Don’t walk the plank in your rush to find treasure.

Be well.

Subscribe to read my stories, and you’ll gain a dedicated ally in the war of humans against machines.

List of AI Detection Tools

Scribbr, Copyleaks, GPTZero, Writer.com, Copyleaks, ZeroGPT, Content at Scale, Wordtune.com, Contentdetector.ai, Sapling.ai, Duplichecker.com, and Originality.ai