Recently we shared concepts with a client. A particular individual was hell-bent on dismissing ideas, claiming they could spot AI work. We provided iterations without any AI assistance. Dismissed as AI. We revised again, writing everything from scratch. Dismissed as AI.
So I played a trick.
I called them on video and suggested we brainstorm together. I looked like I was writing notes as we talked, nodding along, jotting things down. Then I read back what I had “just written” in real time.
It was the AI output from the beginning. The same work they had rejected twice.
“That’s what I’m talking about.” Applause. Sheer conviction.
Nothing about the output had changed. The words were identical. What changed was the story about where it came from. The moment they believed a human was laboring in front of them, the work transformed from worthless to valuable.
This got me thinking.
Welcome to AI-Gregation. A new form of segregation.
When Speed Becomes Evidence of Fraud
Psychologists have long studied what’s called the effort heuristic. In 2004, Kruger, Wirtz, Van Boven, and Altermatt ran experiments where participants evaluated poems, paintings, and medieval armor. Half the participants were told a piece took a long time to create. Half were told it took very little time. Same object. Different story about the labor involved.
When people believed more effort went into something, they rated it higher on quality, value, and liking. The effect was especially strong when the quality of the object was ambiguous or hard to evaluate. If you can’t tell whether something is good, you look for proxies. Time and effort become stand-ins for quality.
Some restaurants figured this out long ago. The meal is ready, but they hold it back a few extra minutes. If your entree arrives too quickly, you wonder whether it was microwaved. A short wait signals that someone in the kitchen is actually cooking. The food hasn’t changed. Your perception of it has.
AI has inverted this heuristic at scale.
Now, the assumption is that any polished output required zero effort. The mere possibility that AI was involved triggers an automatic deduction. Before the reviewer reads a single sentence with the intent to understand it, they’ve already concluded that minimal labor went in. And if minimal labor went in, the work must be worth very little.
Same mental shortcut, different starting point. We used to see polished work and assume someone labored over it. Now we see polished work and assume a machine produced it. The judgment follows the assumption.
Guilty Until Proven Human
The problem isn’t that people are becoming more critical. Critical evaluation would require actually engaging with the work. Reading it. Testing it. Checking whether the logic holds, whether the data supports the claims, whether the strategy makes sense.
What’s happening is different. The evaluation never occurs.
A 2024 study from researchers examining AI-generated content found that when content was labeled as AI-created, people judged it as less trustworthy even when the content itself was evaluated as accurate and fair. The label triggered the response. The quality was irrelevant.
MIT Sloan researchers found that when people didn’t know how content was created, they actually preferred AI-generated marketing copy. But the moment they learned AI was involved, their assessment of human-created work went up while AI work stayed flat. They called this “human favoritism” rather than “AI aversion.” The source mattered more than the substance.
Laziness that sounds like taste.
The reviewer doesn’t have to do the hard work of evaluating quality. They just have to identify the source. Or rather, they have to guess at the source and then treat that guess as a verdict.
Everyone Is Using AI (Except Me)
There’s a concept in psychology called affective realism. Your brain doesn’t passively receive reality and then react to it. Instead, your brain makes predictions about what it expects to encounter, and those predictions shape what you actually perceive. Your prior beliefs become your experience. If you expect to see something, you’re far more likely to see it, whether it’s there or not.
The person who walks into a room expecting AI-generated work will find it everywhere. Their brain is already primed. The prediction precedes the perception. They don’t evaluate and then conclude. They conclude and then look for confirmation.
The same person who assumes everyone else is using AI rarely includes themselves in “everyone.”
They use AI. Of course they do. Everyone does. But when they use it, they’re adding judgment. They’re synthesizing. They’re bringing expertise to bear. Somehow, through this alchemy, what comes out the other end is their original work. The AI was just a tool. A sounding board. A starting point. The output belongs to them.
When you use it? That’s just AI.
The fundamental attribution error wearing new clothes. When I use AI, I attribute the quality of my output to my skill, my judgment, my discernment. When you use AI, I attribute your output to the machine. My context matters. Yours doesn’t.
The reviewer who dismisses your strategy as “obviously AI” probably asked ChatGPT for help structuring their feedback. But that was just brainstorming. That was just getting started. That was legitimate.
You see the asymmetry?
“Everyone is using AI” is a belief held about everyone except the person holding the belief. A universal suspicion with a sample size of one exception: me.
Critics Who Can’t Drive
Think about the person who drives an automatic transmission, then mocks another driver for also using automatic. “You’re not really driving. Real drivers use manual.”
Meanwhile, they’ve never touched a stick shift in their life. They’re mediocre at operating their own automatic. They couldn’t parallel park without a backup camera. But they’ve developed an expertise in identifying other automatic drivers and dismissing them.
They can’t drive well. But they can judge.
This is the new skill. Not competence in the work itself, but competence in identifying how others did their work. The ability to detect shortcuts has replaced the ability to evaluate outcomes. The person who can’t assess whether a strategy is sound can absolutely tell you it “feels like AI.” The person who couldn’t write a compelling paragraph themselves knows exactly what AI-generated prose looks like.
Criticism without true skill.
Prejudice Without a Protected Class
Prejudgment in its most literal form. Pre-judging. Reaching a verdict before examining the evidence. Deciding what something is based on where you assume it came from rather than what it actually contains.
The word prejudice usually refers to judgments about people based on group membership. Race. Gender. Religion. Characteristics that tell you nothing about an individual’s actual qualities but trigger assumptions anyway.
AI-Gregation operates the same way.
The work is no longer evaluated as work. It’s evaluated as a member of a category: suspected AI output. And membership in that category, once assigned, determines the verdict. The content is irrelevant. The logic is irrelevant. The evidence is irrelevant. The category does all the work.
What makes this particularly absurd is that the categorizer is almost certainly a user of the same tools they’re using to condemn others. They’ve just constructed a mental partition. Their use is legitimate. Everyone else’s is suspect.
The same tribal psychology that produces every other form of in-group favoritism, just pointed at a new target.
I am thoughtful. You are lazy. My tools are instruments. Your tools are crutches.
When Excellence Looks Suspicious
Because everyone assumes everyone else is using AI, and because most people lack the ability to actually evaluate output quality, we’ve created a universal suspicion machine.
The excellent work gets dismissed because it seems too polished.
The mediocre work gets dismissed because it seems too formulaic.
The genuinely AI-assisted work gets dismissed because it seems too... whatever the reviewer decides AI seems like.
And the genuinely human work, crafted over hours by someone who actually knows what they’re doing? Also dismissed. Because it pattern-matches to what AI could theoretically produce.
Regression toward mutual dismissal. Everyone becomes an output skeptic. Everyone assumes the worst about everyone else’s labor. The baseline expectation is that nothing required real effort, that nothing reflects real expertise, that nothing deserves genuine engagement.
The better your output looks, the more likely it is to be dismissed without engagement. A bizarre incentive structure. It punishes competence.
The Shortcut That Replaced Evaluation
What’s missing from all of this is what was always missing: the ability to evaluate work on its merits.
The marketing director who dismissed that strategy document couldn’t have explained what made it good or bad. He couldn’t have identified whether the positioning was defensible, whether the competitive analysis was accurate, whether the recommendations followed logically from the insights. He lacked the evaluative framework to assess the substance.
But he had a new shortcut. “It’s AI.”
Two years ago, he would have nodded along to mediocre work delivered with confidence. He couldn’t tell the difference between a solid strategy and a recycled framework with new labels.
Nothing about his evaluative capacity has changed. He still can’t judge quality. He just has a new way to sound like he can.
The Evaluators Still Exist
If you produce excellent work, expect it to be dismissed by people who can’t evaluate it.
This was always true. The difference now is that they have a respectable-sounding reason. “It’s AI” carries more social weight than “I don’t understand this well enough to judge it.” One makes you sound discerning. The other makes you sound limited.
The strategic response isn’t to prove your work is human-made. That’s a losing game. You can’t prove a negative, and the suspicion will persist regardless.
The response is to recognize who you’re actually trying to reach.
The people who can evaluate your work will evaluate it. They’ll engage with the logic, test the claims, notice the nuances that reveal depth. These people exist. They’re just not the majority.
The people who can’t evaluate your work will find reasons to dismiss it. AI gave them a new reason. Before AI, they had other reasons. The reasons don’t matter. What matters is that they were never going to engage with the substance anyway.
Your audience isn’t everyone. It never was.
Discernment Theater
We’ve democratized the appearance of expertise. Anyone can now produce polished output that sounds knowledgeable, regardless of whether they understand the domain.
But we’ve also democratized the appearance of discernment.
Anyone can sound like they have high standards by dismissing work as AI-generated. It requires no knowledge, no skill, no engagement with the actual content. You just have to be suspicious, and suspicion is free.
The more you dismiss, the more selective you appear. The more selective you appear, the higher your perceived standards. Expertise theater performed through rejection.
But dismissing everything is not discernment. It’s the opposite. Discernment requires the ability to distinguish between things. It requires engaging with the work long enough to notice what makes one piece different from another.
The person who rejects every wine at the tasting isn’t a connoisseur. They’re just unpleasant.
None of this is a warrant for AI slop. Lazy prompting produces lazy output. The person who copies and pastes without thought, without editing, without judgment deserves the criticism.
But there are two kinds of slop now. Production slop and criticism slop. One is the careless generation of content. The other is the careless dismissal of it. Neither requires skill. Neither requires engagement. Neither moves anything forward.
A perpetual machine where low-effort output meets low-effort judgment, and both sides feel righteous about it.
Welcome to the era of AI-Gregation.