Slop-Scanning: The New Cognitive Tax

How the prevalence of AI-generated content induces mental fatigue

I am feeling uncharacteristically salty, so buckle up.

I was watching a presentation recently and it was clear that the slidedeck simply took the speakers speech and dumped it into Claude.

Within half a second of the first slide, my brain did this:

The reason I knew why is because Claude has very particular fonts, font sizes, spacing, colors/hex codes, and other formatting giveaways. Claude also likes to do these mini headings in all caps and summary statements at the bottom.

Maybe you have had a similar experience reading something with a ton of em dashes or “it’s not X, but it’s Y” sentence configurations and it messes with your head. I know I am weird and probably an outlier here, and that’s okay, but I can’t be the only one annoyed by this cognitive tax.

The whole thing is frustrating enough that I feel the need to coin a new technical term to efficient describe this new cognitive tax and so I am officially coining the term “slop-scanning.”

Slop-scanning is defined as the ongoing cognitive tax of having to perpetually ask yourself, “is this AI-generated?” The only time you can turn it off is when you have the luxury of reading or interacting with pre-2023 content (three cheers for old books!) and you don’t have to be triggered every time you see an em dash, over-hyphenation (see what I did there), or sentence constructions with three parallel clauses.

Maybe you are saying, “Mike, get over it, who cares if someone used Claude for their slides?

You have a point and on the one hand, I don’t particularly care if the speaker’s slides are made in Claude. On the other hand, as someone who is drowning in AI-generated content, it feels special when you get to interact with something that is entirely analog.

I wouldn’t have minded this particular instance but Claude seemed to have added additional content that wasn’t in the speech. It was small and subtle, mainly in the tiny headings but it also added summary statements that weren’t in the speech.

If we go further down the rabbit hole there are much more problematic cases. I am seeing all sorts of article pitches, email body content, summaries, reports, and other content that are being passed off as if it came from the authors mind… when it clearly didn’t. In my view, we are either running afoul or just in straight violation of the 9th commandment here. There is also a sense where this kind of communication can be insulting — do you think I am too ignorant to notice this AI-content being passed off as yours? Am I not worthy of the time it takes to hear your actual thoughts and wisdom from all of your knowledge and experiences?

I really don’t know how to turn off the slop-scanning. I am not sure I should turn it off even if I could. The prevalence of people taking cognitive shortcuts without attribution is so ubiquitous that prudence demands I keep a permanent and RAM-intensive browser tab in my mind.

I have some hope that we can establish transparency norms around usage and attribution but in the meantime we will all be paying the slop-scanning cognitive tax.

Next week I will coin another technical term, “aignosis” to pinpoint another frustrating new phenomenon in the era of ubiquitous AI-generated content. Make sure you subscribe if you haven’t already.

Hopefully, this is obvious and goes without saying, I never use AI to write any of these posts.


Here are a few developments in AI worth noting from this week (and speaking of old books):
AI Companies Are Buying Tons of Old Books Because They’re Free of AI Slop (404 Media)

Citadel Buys Situational Awareness’s Stock Portfolio After Big Losses in AI (WSJ)

AI and the Image of God: What Does it Mean to be Human? [Silicon Spiritualities - Ep. 3 with Stephen Driscoll]