started · updated
AI market share shifts as Gemini and Claude gain ground on ChatGPT
Recent data highlights shifting dynamics and reliability concerns within the artificial intelligence sector. According to a Comscore report, ChatGPT's dominance in AI prompt volume is eroding, falling from 70% to 50% in the first half of 2026. During the same period, Google's Gemini saw its share nearly double from 17% to 30%, while Claude's share grew from 2% to 11%.
Reliability remains a significant issue, particularly in specialized domains. A study by Saturn revealed that popular AI models provided incorrect answers to financial queries in an average of 57% of cases. While paid models generally outperformed free versions, errors included fabricated rules and ignored tax changes. For instance, Claude Haiku 4.5 provided incorrect information regarding UK pension rules.
Furthermore, research by Chu and Hou suggests that AI recommendation engines exhibit a brand notoriety bias. In cosmetic product testing, AI assistants tended to favor established brands over unknown ones even when product characteristics were identical. The study noted that a minor difference of 0.1 stars in customer ratings can be enough to shift these AI-driven recommendations.
Entities
Claims
What the coverage asserts, and how many sources carry each claim.
- [○ 1 SOURCE] ChatGPT's share of AI prompt volume fell from 70% to 50% during the first half of 2026. www.advanced-television.com
- [○ 1 SOURCE] Popular AI models provided incorrect answers to financial questions in an average of 57% of cases. haszon.hu
- [○ 1 SOURCE] Paid AI models demonstrated higher accuracy than their free counterparts. haszon.hu
- [○ 1 SOURCE] Claude's share of total AI prompt volume increased from 2% in January 2026 to 11% in June 2026. www.advanced-television.com
- [○ 1 SOURCE] Google's Gemini share of AI prompt volume grew from 17% to 30% between January and June 2026. www.advanced-television.com
- [○ 1 SOURCE] A rating difference of just 0.1 stars can significantly alter AI recommendation outcomes. dailydigital.fr
- [○ 1 SOURCE] A study by Chu and Hou found that AI assistants often favor established brands over lesser-known ones when characteristics are identical. dailydigital.fr
- [○ 1 SOURCE] Claude Opus 5 performed best in reasoning mode but still erred in 39% of its responses. haszon.hu