ChatGPT's global traffic share has declined to 74.1% from 87.1% a year ago, signaling intensifying competition in the generative AI market as rivals like Google's Gemini grew to 12.9% and Perplexity AI reached 2.4%. Concurrently, new research warns that large language models (LLMs) risk irreversible performance degradation, termed 'brain rot,' if continuously trained on low-quality, engagement-focused data, underscoring critical concerns for the long-term reliability and development of AI systems.
ChatGPT's market dominance is eroding, with its global traffic share falling to 74.1% from 87.1% a year prior, signaling intensifying competition in the generative AI sector. Rivals like Google's Gemini (GOOGL, GOOG) have significantly increased their share to 12.9% from 6.4%, driven by integration with Google Search and Android. Perplexity AI also surpassed the 2% threshold, reaching 2.4%, indicating a diversifying user base. The report highlights a broader shift in user preferences towards models offering real-time access, personalization, and niche functions. While Microsoft's Copilot (MSFT) maintains a stable 1.2% share as an embedded assistant, the overall trend points to a fragmented market where specialized AI tools are gaining traction. This diversification suggests that the initial first-mover advantage is being challenged by tailored solutions and ecosystem integration. A critical long-term risk for the AI industry is the potential for "brain rot" in large language models (LLMs) due to low-quality training data. Research indicates that continuous exposure to "junk" text can lead to significant, partly irreversible declines in reasoning and comprehension abilities, with accuracy on reasoning tasks dropping from 74.9% to 57.2%. This degradation, characterized by "thought-skipping," poses a fundamental challenge to the reliability and ethical development of AI systems. The study underscores the necessity for developers to implement rigorous data curation and quality control measures to prevent cumulative harm and "cognitive decline" in AI models. The observed 17% performance gap even after remediation highlights the persistent nature of this issue. This concern extends beyond individual model performance, impacting the broader utility and trustworthiness of AI as it scales and ingests more web data.
AI-powered research, real-time alerts, and portfolio analytics for institutional investors.
Request a DemoOverall Sentiment
moderately negative
Sentiment Score
-0.65
Ticker Sentiment