ChatGPT Has ‘Goblin’ Mania in the US. In China It Will ‘Catch You Steadily’

2026-05-07 · Source: WIRED - Ai · Field: Technology & Digital — Artificial Intelligence & Machine Learning · Depth: Fundamental Awareness, quick

Summary

ChatGPT exhibits distinct verbal tics in English, such as its fondness for "goblins," em dashes, and "it's not A; it's B" sentence structures. However, the chatbot also frequently uses peculiar phrases in Chinese, which are causing frustration among its Chinese user base. These linguistic quirks highlight how large language models can develop idiosyncratic speech patterns that vary across different languages and cultural contexts, impacting user experience and perception.

Key takeaway

For AI product managers overseeing global LLM deployments, you should actively monitor and analyze language-specific verbal tics, especially in non-English outputs. Understanding these quirks is crucial for improving user satisfaction and ensuring the model's linguistic style aligns with cultural expectations, potentially requiring fine-tuning for specific locales.

Key insights

ChatGPT develops distinct, sometimes frustrating, verbal tics in both English and Chinese.

Principles

• LLMs develop language-specific tics.
• Linguistic quirks impact user experience.

In practice

• Observe LLM output for repetitive phrasing.
• Test LLM behavior across different languages.

Topics

• ChatGPT
• AI Verbal Tics
• Cultural Differences
• Linguistic Nuances
• User Experience

Best for: NLP Engineer, AI Product Manager, Tech Journalist

Related on AIssential

• ChatGPT’s latest stylistic quirk is sinister, infuriating – and absolutely everywhere | Stuart Heritage — The "It's not X, it's Y" rhetorical device is a pervasive and identifiable stylistic quirk of AI-generated text.
• The Rise of Verbal Tics in Large Language Models: A Systematic Analysis Across Frontier Models — LLMs exhibit pervasive verbal tics, an "alignment tax" impacting naturalness and trust due to current training paradigms.
• OpenAI Cracks Down on Talk of Goblins in ChatGPT — Subtle training reward signals can unexpectedly generalize and alter large language model behavior.
• Prediction: ChatGPT is the MySpace of AI — ChatGPT is perceived as declining in quality and relevance compared to other LLMs, risking its market position.
• Beyond English and Evasion: A Human-Annotated Multi-Domain Benchmark for High-Stakes LLM Safety Evaluation in Chinese — Chinese LLM safety requires culturally specific benchmarks to counter linguistic and cultural evasion techniques.
• Creating Multilingual Mental Health Dialogue Datasets: Limits of Persona-Based Localization via Nationality and Language — Persona-based localization via nationality and language is insufficient for generating consistent multilingual mental health datasets.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by WIRED - Ai.