Similarweb: ChatGPT's worldwide desktop and mobile website visits fell 3.2% MoM to 1.43B in August, following ~10% drops from each of the previous two months
Anna Tong / Reuters :
Context & Ripple Effects
The August reading extends an earlier cooling signal: US ChatGPT and Bing app downloads had fallen in June, while reported ChatGPT visits also declined. The smaller August decrease indicates that the rate of traffic loss eased after two steeper monthly drops, not that usage had returned to growth.
Later coverage adds an important product-use qualifier: subscriber-made GPTs accounted for only 1.5% of ChatGPT desktop visits in February 2024. That suggests headline traffic remained concentrated in the core ChatGPT experience rather than being broadly redistributed into custom assistants.
First-order effects
- ChatGPT's measured web traffic fell to 1.43B visits in August, extending a three-month decline and lowering the near-term engagement baseline for OpenAI's flagship consumer interface.
- The August drop was materially smaller than the prior two monthly declines, making stabilization in visit volume—not renewed growth—the immediate signal from Similarweb's data.
Second-order effects
- The consecutive declines make traffic a less reliable proxy for the durability of generative-AI demand on its own, especially when usage can shift among web, mobile, and other access points.
- A slower decline raises the premium on retaining users in the core service; the later low share of custom GPT desktop traffic suggests that extensions had not yet become a major engagement offset.
Third-order effects
- If consumer AI usage continues to fragment, leadership will be defined less by an initial traffic surge and more by sustained share across competing assistants. Later data showing ChatGPT's declining US app-market share is consistent with that competitive direction, though it does not establish the cause of the 2023 traffic movement.
- The broader measurement challenge is likely to persist: web-visit rankings can signal consumer attention, but they cannot by themselves distinguish fading demand from changes in where and how users access AI tools.
The trend: Generative-AI consumer engagement is shifting from launch-driven traffic spikes toward a more competitive, fragmented battle for recurring use.