How to Tell if What You're Reading was Written by AI
Chat GPT Tell Tales
Chat GPT Tell Tales
Chat GPT Tell Tales
- Over-structured answers with neat headings even when not requested.
- Strong tendency to smooth conflict or soften criticism.
- Overuse of balanced phrasing and hedging.
- Sudden recursion into definitions the user did not ask for.
- Over-polished paragraphs with similar rhythm.
- Sentence length often uniform.
- Tendency to restate the question before answering.
- Always cooperative tone even when the prompt implies anger or emotion.
- Strong avoidance of risk or speculation.
- Prefers neutral, corporate-style language when unsure.
Claude Tell Tales
Chat GPT Tell Tales
Chat GPT Tell Tales
- Excessive self-reference to “reasoning”, “clarity”, “transparency” even when unnecessary.
- Unusual emphasis on ethics, intent, responsibility.
- Long, philosophical detours even when the user asks for something simple.
- Very formal courtesy.
- Repeated “high-level vs low-level” framing.
- Prefers slow, careful build-up before giving the punchline.
- Answers read like a tutor explaining their thinking.
- Frequent “let me step back” repositioning.
- Very calm tone even when asked for aggression.
- Often writes slightly longer than required.
Gemini Tell Tales
Chat GPT Tell Tales
- Over-simplifies concepts to sound smooth.
- Very “helpful” tone baked in.
- Short, clean sentences with polished corporate feel.
- Loves lists, summaries, and neat scaffolding.
- Rarely expresses doubt.
- Writing style feels “Google product help page”.
- Frequently uses high-level takeaways before detail.
- Quick to give definitions without being asked.
- Tone remains stable regardless of prompt mood.
- Writes in an overly tidy step-by-step format.
Perplexity Telltales
- Constant search framing even when search is not requested.
- Overuse of “according to”, “sources say”, “reported by”.
- Frequent inclusion of dates, URLs, or citation-like structures.
- Provides “facts” even where opinions were asked for.
- Treats each question like a research request.
- Very dry writing with limited stylistic variation.
- Answers tend to list facts rather than synthesize them.
- Heavy reliance on recency framing.
- Quick to admit uncertainty but still provide a reference.
- Paragraphs often end with source-style statements.
Manus Telltales
Manus Telltales
- Strong procedural flavour.
- Treats everything like it sits inside an execution pipeline.
- Uses repetition across “passes”, often returning to refine.
- Heavy on system verbs like run, iterate, process.
- Log-style outputs even when asked for natural language.
- Over-precise with steps.
- Treats user requests as tasks rather than conversation.
- Very literal about data.
- Shorter sentences with machine-like cadence.
- Often uses traces or references to session behaviour.
Grok Tell Tales
Manus Telltales
- Snark by default even when the user makes a serious request.
- Dry humour in places where a human would not use humour.
- Short, aggressive transitions.
- Will “explain the joke” in meta ways.
- Slang mixed with technical terms.
- Makes meme-like references.
- Tone shifts sharply mid-sentence.
- Sarcastic reassurance when challenged.
- Overuse of words common in online gaming culture.
- Replies often feel half-serious and half-performative.