We know that ChatGPT remains the benchmark in the mainstream AI space, but I am afraid its dominance is becoming less striking, thanks above all to the growing prominence of Gemini, Google’s AI. That is, at least, what the Comscore AI Intelligence Report confirms, based on data from the second quarter of the year.
As you will see, the conclusions are based on data from the U.S. market, but they can still help us get a sense of what will (probably) happen here in a few months.
After reviewing data from the U.S. market, Comscore found that its share of prompt volume fell from 70% in January to 50% in June of this year: in other words, half of the prompts used across the main general-purpose AI platforms now take place in ChatGPT.
Which is still a lot, of course 😉
During that period, Gemini rose from 17% to 30%, and Claude from 5% to 11%. Much farther behind is Microsoft Copilot, steady at 5%. But be careful: we are not talking here about the number of users or conversations, but rather the number of prompts, an area where usage varies considerably across platforms. For example, the average conversation in Claude generates about 9.1 prompts, compared with 6.2 in Gemini and just 5.4 in ChatGPT.
In any case, losing prompt share does not necessarily mean losing users. In fact, Comscore estimates that ChatGPT’s mobile visitors in the United States grew from 25 million to 47 million between June 2025 and June 2026. Gemini, Copilot, and Claude also gained audience: Gemini added about five million mobile visitors in one year, rising from 14 million to 19 million. Copilot increased from 9 million to 11 million, and Claude tripled its mobile audience from 1 million to 3 million.
The share of desktop searches showing an AI-powered answer reached 39.4% on Google in June (+14 points over the past twelve months) and 17.3% on Bing (+5.5 points over the past twelve months). This is a logical conclusion, given the increasing weight that Google and Microsoft have gradually given to automated responses, such as those offered through AI Overviews.
Of course, not all media outlets play the same role in these answers. Combining data from both platforms, we find major differences by vertical:
The study looks at lodging-related queries to highlight an important difference: being listed among ChatGPT’s sources is not the same as being cited in the answer text itself. For example, Tripadvisor appeared among the sources in 61% of the responses analyzed, but it was shown as a cited link in only 21%. In the case of Hotels.com, the percentages were 50% and 18%, respectively. Expedia drops from 46% to 16%, while Booking.com falls from 33% to 9%.
Among those who bought home and decor items on the eCommerce sites analyzed, 35% had visited an AI assistant in the previous 30 days. The report also identifies a link with conversations specifically about that category: 14% of buyers had a conversation with an AI about furniture and decor during the same session in which they completed their purchase.
This habit appears with varying intensity depending on the category. Looking only at conversations related to what the user ultimately purchased, the share of users who had that type of conversation during the same shopping session was 12% in books, music, and video, 10% in office supplies, and 4% in toys and hobbies.
Image: ChatGPT
Your email address will not be published. Required fields are marked *
Δ