Text-similarity researcher: LLM writing isn't plagiarism — humans have always been remixing memes

On September 23, Rex Douglass, a scholar who has done large-scale text similarity analysis, posted a series of tweets pushing back against the idea that writing with LLMs is plagiarism. His core argument: originality itself is an illusion, and LLMs merely do more transparently — and with a record — what humans have always done.

Confirmed

• Drawing on his practical experience with large-scale text similarity computation, he points out that culture itself is an evolutionary process in which memes and content are copied, rearranged, and remixed thousands of times with slight variations; the human brain likewise keeps recycling memes, just without leaving a trace, while LLMs leave a record.
• He argues that human text being "predictable" is exactly why LLMs work: people spend their lives slightly rewording others' sentences to avoid "plagiarism," and all of humanity doing this simultaneously produces massive amounts of n-grams, while genuinely new "ideas" are rare.
• He rejects the nostalgia-driven narrative of "authentic creation," saying those resisting AI content are defending a past that never existed — essentially protectionism against the next generation of competitors; the market doesn't care whether content is LLM-generated, just as people never cared that humans copy memes from each other.
• He criticizes academia for systematically overestimating the originality of research output: almost no one does metascience, so few realize that the vast majority of research is mediocre — derivative, copycat, vague, and rushed — which happens to make it a "very good baseline" for LLMs.

Why it matters

• This view shifts the AI plagiarism debate from "do machines steal" to the more fundamental question of "does human originality even exist," offering a counterintuitive, empirically grounded defense of LLM-generated content that also indirectly explains why LLMs perform well on academic and research writing.

2026-09-23 ~ 2026-09-23 · 5 related posts

• [source] Anyone who's done text similarity at scale never believed in originality — RexDouglass · 2026-09-23
• Text-similarity researcher: culture is memetic copying, so LLM 'plagiarism' ring hollow — RexDouglass · 2026-09-23
• Researcher argues LLM 'plagiarism' critique is ironic: all humans regurgitate memes constantly — RexDouglass · 2026-09-23
• Researcher: most research output is terrible — and that's exactly why LLMs work — RexDouglass · 2026-09-23
• [source] Market doesn't care that content is LLM-generated, argues Rex Douglass — RexDouglass · 2026-09-23