April 6, 2026
Spot the bot? Cue the plot!
Ask HN: How do systems (or people) detect when a text is written by an LLM
Humans say 'I can tell' vs 'No one can' as AI‑text witch hunt heats up
TLDR: HN asked how to spot AI-written text and got a split: some say it can’t be done reliably, others claim they can hear the “bot tone.” A touted detector, Pangram, enters the chat, while users worry false positives are turning good writing into a witch hunt—and changing how people write.
Hacker News threw a hot potato: “Can you detect when text is written by AI?” The thread instantly split into two camps. One side, led by Someone1234, insists detection is basically impossible and says false accusations are turning good writing into a witch hunt. Their advice? Sprinkle a few typos so you don’t look “too good,” and watch out for “Signs of AI writing” lists people now use as checklists.
The other side fired back that seasoned readers can smell “bot tone” a mile away. alex43578 says native speakers can spot “unpolished AI copy-paste,” citing telltale quirks: emoji bullet lists (lol), fake citations, and that meme-y rhythm: “It’s not X. It’s Y.” PufPufPuf calls that a default AI pattern; prmph adds that AI loves “verbose lists with fluff” and a hyper-consistent, punchy vibe. Then the style war hits: alex43578 mocks dramatic colon titles—“The Strait of Hormuz: Chokepoint or Opportunity?”—and Filligree claps back, “I’ve used titles like that for thirty years,” prompting a “show receipts” challenge. Drama achieved.
Meanwhile, tool fans drop receipts: moonu links to Pangram, a detector claiming low false alarms with a paper. dipb says machines do it by pattern training. But skeptics warn detectors can misfire, and humans are already writing worse to dodge them. The vibe? Equal parts sleuthing, skepticism, and meme-police energy—welcome to the “spot the bot” Olympics.
Key Points
- •An Ask HN post asks how to detect when text is written by a large language model (LLM).
- •The question covers both human and computer-based detection approaches.
- •The author specifically asks whether APIs exist for automated detection of AI-generated text.
- •The post provides no proposed solutions, serving as an open inquiry to the community.
- •The topic focuses on distinguishing LLM-generated content from human-written text.