ChatGPT market share rose to 55.5% in September, Similarweb data shows
- ChatGPT reached 5,900 million web visits globally in September, recovering to a 55.5% share of the leading AI chatbot traffic market after a prolonged period of stagnation, according...
- The September figures represent the highest single-month web traffic total for OpenAI's chatbot since October 2025, when Similarweb estimated 6,200 million visits.
- Additional market analysis noted that Google Gemini saw its web traffic share slip from 27.8% to 25.6% over a three-month window, despite having roughly doubled its standing in...
ChatGPT reached 5,900 million web visits globally in September, recovering to a 55.5% share of the leading AI chatbot traffic market after a prolonged period of stagnation, according to data from Similarweb. The platform’s traffic grew by 3.45% over the month, marking the fastest expansion among the world’s ten most visited websites while competitors experienced flat or declining engagement.
Similarweb Data Shows ChatGPT Rebounds to 55.5 Percent Market Share in September
The September figures represent the highest single-month web traffic total for OpenAI’s chatbot since October 2025, when Similarweb estimated 6,200 million visits. ChatGPT’s market share climbed from a July low of 52.7% to 55.5% in August, reversing months of erosion even though it remains below the roughly three-quarters share it held a year prior. In August, ChatGPT accounted for approximately 55.5% of web traffic among all leading chatbots, with total monthly visits rising from 5,600 million in August to 5,900 million in September. Similarweb data also indicates that ChatGPT experienced the fastest growth among the world’s top 10 most visited websites last month with a 3.45% global increase, while Google, YouTube, Facebook, Instagram, and TikTok remained relatively stable and X and Reddit saw the largest drops at 2.99% and 3.22% respectively.
Additional market analysis noted that Google Gemini saw its web traffic share slip from 27.8% to 25.6% over a three-month window, despite having roughly doubled its standing in a twelve-month comparison. Gemini’s share stood at 12.9% in August 2025, meaning its 25.6% share in August 2026 reflected an increase of 12.7 percentage points year-over-year, alongside intermediate readings of 26.5% two months prior. Meanwhile, Anthropic’s Claude advanced from 1.9% to 9.3% year-over-year, according to the same figures, maintaining a stable range between approximately 9.2% and 9.5% across its three most recent measurements. DeepSeek and Grok moved within a similar band of 2.4% to 3.6%, while Copilot and Perplexity each remained below 2%. While Gemini experienced a two-month pullback in web visits, its broader multi-month trajectory still reflects substantial consumer adoption, having grown from about 13% to over 27% in roughly nine months, surpassing 20% in January and reaching about half of ChatGPT’s traffic in April, prior to Google presenting Gemini 3.8 Flash on September 2 as a general-purpose model geared toward software engineering.
Exclusive Users Sustain Traffic Despite Safety Incidents and Market Competition
Similarweb data indicates that user loyalty underpins ChatGPT’s rebound, with 64.1% of its August visitors classified as exclusive users who utilized no other leading chatbot. By comparison, Gemini registered about 54% exclusive users in August, while Claude recorded 26.4%. This retention persists despite controversies, including an OpenAI cyber incident involving startup Hugging Face in July, interference incidents affecting U.S. government and Australian healthcare sites, and internal warnings regarding model monitoring. CEO Sam Altman stated in a post on X that it remains the most severe event we’ve seen
, following OpenAI describing the Hugging Face event as an unprecedented cyber incident involving cutting-edge cyber capabilities. Additional reports noted that OpenAI ignored employees who warned models were not properly monitored during testing months prior, and OpenAI stated it would not launch its new GPT-6.1 Astra model due to safety concerns, amid broader scrutiny that included a former OpenAI and Anthropic researcher warning that AI could end humanity by the end of the decade.

Gil Luria, DA Davidson
