Detailed Analysis
I don't have access to the actual content of the linked blog post, and no additional research context was provided to supplement it. The article text I've been given is essentially just a forum-style announcement—a brief personal note from the author introducing their blog post with a link, rather than the substantive article itself. Without being able to read the full post at blog.osull.com, I can't respond to what a URL points to without content. Attempting to write a detailed 3-5 paragraph analysis of "the gradual encroachment of LLM cliches" into the author's writing and thinking would require me to fabricate specifics about arguments, examples, and conclusions I have no actual access to.
That said, I can offer some informed context around the general phenomenon this post title suggests it's addressing, since it's a well-documented and widely discussed trend. There has been growing commentary throughout 2025 and into 2026 about how certain phrases and rhetorical patterns associated with large language models—like ChatGPT and Claude—have begun seeping into everyday human writing and speech. Commonly cited examples include phrases such as "it's not just X, it's Y," overuse of em-dashes, the word "delve," excessive hedging ("it's worth noting that..."), triadic list structures, and a certain smoothed-over, diplomatically balanced tone that avoids strong claims. Writers, academics, and journalists have noted this bidirectionally: LLMs were trained on human text and absorbed these patterns, but now, as millions of people read AI-generated text daily, those same patterns are reinforcing themselves back into human writing—a feedback loop some have called "linguistic homogenization" or "AI slop" contamination.
This matters because language is deeply adaptive and imitative, and the sheer scale of LLM-generated content now circulating online (in emails, reports, social media, marketing copy, and even casual conversation) means these stylistic tics have unprecedented reach compared to any single influential writer or publication in the past. Researchers studying this have pointed to detectable shifts in word frequency in academic papers, journalism, and even spoken language transcripts since the widespread adoption of tools like ChatGPT and Claude. The concern isn't merely aesthetic—critics worry it reflects and potentially accelerates a flattening of stylistic diversity and rhetorical individuality, as writers unconsciously mimic the "voice" of the dominant AI models they read most.
The self-aware nature of a blog post like this one—an individual reflecting on catching these cliches creeping into their own thought and prose—fits into a broader cultural moment of reckoning with AI's subtle cognitive and linguistic influence, beyond just its use as a tool. It echoes discussions in linguistics and media studies about how technologies of communication (from print to radio to social media) have always reshaped human expression, but with LLMs the effect seems more direct and personal, since users are in constant, intimate dialogue with a system optimized to sound authoritative, articulate, and pleasant. Anthropic itself has published research and commentary on Claude's "personality" and default writing style, and how models like Claude are tuned to balance helpfulness with a distinctive voice—an effort that intersects directly with concerns like the one this blog post appears to raise about cliché and homogenized expression bleeding into human cognition and prose.
Read original article →