A Store-Level Breakdown of the Generative AI ‘Three Kingdoms’ Era Shifted in Six Months [Announced September 22, 2026, Comscore Survey]｜shishimaru（獅子丸）

To summarize: 1) ChatGPT’s market share dropped from 70% in January 2026 to 50% in June; 2) Gemini (17% to 30%) and Claude (2% to 11%) have gained share accordingly; and 3) the assumption that ‘AI just means ChatGPT’ is crumbling within six months, making the idea of using different AIs for different situations a practical reality.

Many store operators likely think, ‘I’ve only ever used ChatGPT,’ or ‘I hear about Gemini and Claude, but which one should I actually use?’ Honestly, it is a hassle to rethink and choose an AI every time you write a daily review response or create a social media post, and it is natural to think, ‘It’s safer to just use the most mainstream one.’ However, an overseas research firm has announced that there has been a clear change in generative AI usage over the past six months.

What happened?

According to the ‘Q2 2026 AI Intelligence Report’ released by the American research firm Comscore on September 22, 2026, there was a significant shift in desktop usage share for major generative AI chatbots between January and June 2026. While ChatGPT dropped 20 points from 70% to 50%, Google’s Gemini grew significantly from 17% to 30%, and Anthropic’s Claude grew from 2% to 11%. Although ChatGPT remains the leader, this report suggests that the market has shifted from a ‘one-horse race’ to a ‘state where the three major companies compete.’

The power map in numbers

The specific numbers are as follows (desktop usage share from January to June 2026):
1) ChatGPT: 70% → 50% (approx. 168 million desktop conversations in June)
2) Gemini: 17% → 30% (conversations grew from approx. 8 million to approx. 89 million, about 11 times in one year)
3) Claude: 2% → 11% (conversations grew from approx. 1.1 million in October 2025 to approx. 22.3 million in June 2026)
Regarding the background of this change, Comscore commented, ‘With intensified competition and the advancement of AI usage differentiation, it is no longer the case that a single AI experience defines the market.’

What changes for store operators?

More important than the individual numbers is the fact that the assumption that ‘using only ChatGPT is enough’ is collapsing. In previous Shishimaru AI utilization notes, I have introduced AIs that are suited for specific situations, such as Gemini for drafting review responses (due to its integration with GBP) and Claude for editing photos and documents (due to its integration with Adobe). The shift in market share can be seen as proof that this ‘usage-based differentiation’ is spreading beyond just a few experts to all users. You don’t need to force yourself to master everything, but if there is a task where the AI you always use isn’t giving great results, it is more worth trying a different AI than before.

Three points to keep in mind

1) These figures are data from January to June 2026 and may change further in the future.
2) Because numbers vary depending on the research firm and calculation method (desktop only vs. including mobile, etc.), simple conclusions about superiority cannot be made.
3) ‘Growing market share’ does not necessarily mean ‘it fits your store.’ It is important to choose based on whether it actually fits your work after trying it out.

Three things you can do starting today

1) Don’t limit yourself to one AI; try others using their free plans.
2) Keep a note of which AI is ‘compatible’ with each task, such as review responses, social media posts, or image editing.
3) When trying a new AI, maintain your usual precautions, such as not entering personal information or customer data.

Summary

I have organized the Comscore survey results, which show that ChatGPT’s market share dropped by 20 points in six months while Gemini and Claude grew significantly, from a store-level perspective. Please note that these figures are as of June 2026 and may change depending on future circumstances. Next time, I plan to delve into specific usage scenarios that are useful when actually differentiating between them.

Source