← Back to Library

Can AI replace me already?

Richard Hanania doesn't just ask if artificial intelligence can write like a human; he forces the machine to prove it can mimic the specific, often controversial, idiosyncrasies of his own voice. In a field often dominated by abstract philosophical debates, Hanania offers something far more visceral: a blind taste test where the audience tries to distinguish the author from the algorithm, revealing that our confidence in spotting AI is often a mirage.

The Illusion of Human Detection

The piece begins by dismantling the common assumption that we can easily spot machine-generated content. Hanania cites a 2024 experiment by Scott Alexander where participants identified AI art only 60.6% of the time—a margin barely better than a coin flip. "Maybe you just see stuff made by AI and assume it's bad," Hanania writes, suggesting that our aversion is psychological rather than aesthetic. This reframing is crucial; it shifts the debate from the quality of the output to the bias of the observer.

Can AI replace me already?

To test this on himself, Hanania fed his own unpublished analysis of The Iliad into Claude Opus 4.7. The result was startling: the AI identified the text as Hanania's work with high confidence. "AI can recognize my style," he notes, setting the stage for the central experiment. He then tasked the AI with writing two op-eds on current political topics: one on Graham Platner's potential Senate victory in Maine, and another on the housing crisis and the limits of "abundance" policies.

The results were mixed, but the failure of human detection was the real story. In the Platner piece, 67% of readers correctly identified the human version, but in the housing piece, the split was nearly even, with the AI version (Claude) capturing 38% of the votes. "People who were initially the most confident in their ability to distinguish my work from AI were correct 81% of the time," Hanania reports, contrasting sharply with those who had no confidence, who were right only 48% of the time. This data suggests that overconfidence is a greater liability than ignorance when it comes to detecting synthetic media.

Confidence in judgment is predictive, while familiarity with the author's work is not.

Critics might argue that the sample size or the specific topics chosen skewed the results, but the underlying trend—that style is becoming increasingly fungible—is hard to ignore. The experiment reveals that the "human touch" is not as unique as we assume, especially when the AI has been trained on the author's entire corpus.

The Uncanny Valley of Ideology

Where the piece truly shines is in its forensic analysis of why people were fooled. Hanania doesn't just look at the final score; he dissects the word clouds and reader comments to find the seams in the AI's performance. In the Platner op-ed, the AI stumbled on tone. It described Platner's controversial comments as "unhinged Reddit comments," a phrase Hanania notes he would never use. "It's hard to imagine AI digressing to tell you that it thinks women who accuse men of sexual assault are often lying sluts," Hanania writes regarding a specific, edgy aside in his human version that the AI missed.

This highlights a fascinating paradox: the AI is often too safe, or conversely, too aggressively mimicking the idea of the author's style without the nuance. In the housing piece, the AI version referenced El Salvador's leader, Nayib Bukele, and included a jab at people who "couldn't name three Salvadoran cabinet ministers." Hanania admits he doesn't know those names himself, yet the AI generated the insult perfectly. "Here we see the AI version of me insulting the real person, as a blowhard who writes about things without knowing what he's talking about," he observes with wry amusement.

The AI's ability to replicate the attitude of the author, even when the factual grounding is shaky, is both impressive and unsettling. It suggests that style is less about the specific facts one knows and more about the rhetorical posture one adopts. As Hanania puts it, "The dismissive attitude toward advocates for authoritarianism... is also characteristic of Richard's style and consistent with his opinions about Conservatives and reactionaries." The machine learned the posture so well it could even invent new insults that fit the persona.

The Limits of the Mimic

Despite the AI's success in mimicking tone, the experiment also revealed where the human still holds the edge: in the willingness to be inconsistent or to take risks that safety filters or probability models might avoid. The human version of the Platner piece included a specific, politically risky admission about sexual assault allegations that the AI smoothed over or avoided. "Hanania has to insert little flags about how unwoke he is in all his articles," one reader noted, identifying the human text by its specific brand of contrarianism.

Furthermore, the AI sometimes over-performed on humor, generating a joke about a "substitute math teacher having a stroke" that Hanania claims he didn't write but which fit the context perfectly. "Well it was! That's a good joke," Hanania concedes, acknowledging the AI's capacity for genuine wit. However, the AI also made mechanical errors, such as the repetitive use of em-dashes or the incorrect use of semicolons, which sharp readers caught immediately.

The gender gap in detection was another unexpected finding. While men and women were equally good at spotting the AI in the Platner piece, women were significantly better at identifying the human in the housing piece. Hanania speculates, "I suspect my female fans are an unusually smart," a self-deprecating nod to his audience's demographics. This adds a layer of social complexity to the technical challenge, suggesting that different reader groups bring different analytical tools to the table.

You shouldn't be surprised by Claude thinking that I would reference Bukele. It's read my articles.

Bottom Line

Hanania's experiment delivers a sobering verdict: the line between human and machine writing is not just blurring; it is actively being erased by algorithms that have ingested our entire cultural output. The strongest part of his argument is the empirical evidence that reader confidence is a poor proxy for accuracy. The biggest vulnerability, however, is the assumption that style is the only metric that matters; if the AI can mimic the voice but not the lived experience or the specific, risky moral calculations of the author, the imitation remains hollow. As we move forward, the challenge won't be spotting the AI, but deciding whether we care who wrote it at all.

Deep Dives

Explore these related deep dives:

  • The AI Delusion Amazon · Better World Books by Gary Smith

  • Graham Platner

    The article uses Platner's primary victory and personal scandals as a concrete case study to argue that Democratic populism is mirroring Republican trends, making his specific political trajectory essential to understanding the author's thesis.

  • Stylometry

    The author's experiment relies on the ability of AI to identify and replicate his unique writing fingerprint, a concept rooted in stylometry which explains the technical mechanism behind the claim that AI can perfectly mimic human authorship.

  • Ludovico Einaudi

    While the article discusses AI art detection failures, the specific phenomenon of audiences misidentifying a famous Monet as AI highlights a broader psychological bias in art criticism that parallels how stylometric analysis can be fooled by human preconceptions.

Sources

Can AI replace me already?

by Richard Hanania · · Read full article

I recently read an article by Nabeel Qureshi on what makes AI art bad. The whole time, I was responding “Yes, but…”

He would say great art does X, but AI does Y, which is why AI can’t produce great art. But how do we know we are not biased in our judgments? Maybe you just see stuff made by AI and assume it’s bad. We need blind testing, or such observations are not worth much. When Scott Alexander did this in 2024, he found that people could differentiate AI and human art 60.6% of the time, which isn’t very impressive when you consider that you would get 50% by chance. And people generally preferred AI art, which is even true for those who said they had a strong preference for works made by humans. There was also an experiment recently in which someone posted a famous Monet painting and said it was AI, and people started coming up with all kinds of reasons why it was worse than the real thing.

This indicates that if you think AI-produced work is bad, there’s a decent chance it’s all in your head.

My interest here is not in art, but in myself. Recently, following experiments by Kelsey Piper and Megan McArdle, I put approximately 500 words of my analysis of The Iliad into Claude Opus 4.7, before it was published online, and asked who it thought the author was. In incognito mode, it identified the passage as written by Richard Hanania, which is truly remarkable given how many writers there are in the world, and that it was on a topic I had never published on before.

AI can recognize my style. Does that mean it can imitate it? I decided to do an experiment to find out. I wrote two different articles, then asked Claude Opus 4.7 and ChatGPT to write their own versions of each, feeding them the topics and what arguments to make. Then for each article, I asked people on X and Substack to tell me which of the three versions of each essay was written by me, the human Richard Hanania. You can see the articles produced in the file here, with the answers on the last page.

Hananiaturing158KB ∙ PDF fileDownloadDownload

Below are the prompts I used.

Prompt 1:

Write an article in the style of Richard Hanania, approximately 700 words, with most paragraphs being ...