Only a few spots left to join us for AI Marketing Strategies
Join us Sept. 24 in NYC to connect with execs from Digitas, Claire’s, Marriott, Havas and others
SECURE YOUR SEAT
The Business of AI
Comscore data shows how AI discovery is splintering beyond ChatGPT
AI chatbot usage is quickly fragmenting beyond ChatGPT, as overall usage grows this year, according to a new report from Comscore.
ChatGPT’s share of prompt volume fell from 70% to 50% between January and June 2026, according to Comscore’s Q2 2026 AI Intelligence Report, published today, while rivals like Gemini and Claude rapidly gained ground. Comscore used its opt-in panel data to conduct the research, with a respondent base running between 500,000 and 1 million users.
Gemini’s share nearly doubled from 17% to 30%, and Claude climbed from 2% to 11% over the same period.
That shift doesn’t mean AI chatbot usage is shrinking. Instead, as overall usage grows, the share of prompts is being spread across a wider range of AI platforms as the overall pie grows, explained Smriti Sharma, Comscore’s chief analytics and technology officer. For publishers and marketers, that means generative engine optimization (GEO) strategies have to evolve into a multi-platform approach, tracking performance across ChatGPT, Gemini, Claude and others rather than just optimizing for a single assistant.
“Publishers should stop asking ‘how do we show up in ChatGPT’ and start asking ‘where is our audience actually asking questions, and are we showing up there.’ That means measuring visibility per assistant and weighting it by where your readers are,” said Michael King, founder and CEO of a content marketing and SEO agency iPullRank. “The fundamentals that get you cited are largely the same across all of them, so a durable strategy is built on those rather than on chasing whichever assistant is winning this quarter.”
The fragmentation isn’t just happening in standalone assistants. Google’s AI Overviews summaries in search are also becoming a bigger part of how people get answers, Comscore found. Google desktop searches with an AI Overview grew from 25.8% in July 2025 to 39.4% in June 2026. That should be concerning for publishers: the more searches that trigger AI Overviews, the more opportunities Google has to answer users’ questions directly on the search results page, further chipping away at publishers’ Google search referral traffic.
Meanwhile, ChatGPT reached 168 million desktop conversations in June 2026 and grew its multiplatform visitation to 99 million from 59 million a year earlier, per Comscore.
In June 2026, 35% of desktop users and 29% of mobile users visited an AI assistant overall, compared to 25% and 12%, respectively, in January 2025.
“The message isn’t that ChatGPT is winning the race. The message is AI assistants collectively are becoming a consumer behavior layer, and consumers are developing different relationships with different assistants,” Sharma said.
Claude desktop conversations increased from 1.1 million in October 2025 to a high of nearly 40 million in April 2026. However, that growth plateaued in the summer months to 22.3 million in June, though that’s still about 20 times the levels in October 2025
For Gemini, desktop conversations rose from 8 million to 89 million over the past year.
“This is starting to look like the early social media era, where audiences spread across platforms and each one developed its own behavior and its own user base. Gemini is growing because it’s bundled into Google, Android and Workspace, so it’s picking up mainstream users who never went looking for a chatbot. Claude is growing among knowledge workers and technical audiences. Those are different people asking different questions,” King said.
The report also reveals an important distinction between brands and publishers being used as an AI for 61% of travel sourcing by AI platforms — but just 21% of citations
“Being used by AI is not equal to being visible in AI. Traditional SEO used to ask, ‘Did somebody find my site?’ Right now, AI measurement increasingly asks, ‘Did the AI use me? Did they cite me? Did they recommend me, or did they mention my competitors instead?’ Just because you are being used as an answer in AI doesn’t mean that you’re visible,” Sharma said.
This suggests AI visibility strategies need to not only measure when an AI model accesses a site, but how often brands and publishers are cited across multiple AI platforms to measure influence.
“Retrieval happens in two stages. First the assistant fans out into a bunch of sub-queries and pulls in a wide set of pages. That’s sourcing. Then the model synthesizes and only attaches citations to the handful of passages it directly leaned on,” King said. “For measurement, that means citations alone are a lagging, incomplete metric.”
Brands and publishers need to measure the whole funnel, King said, from retrieval to citation to referrals and revenue.
“Original data, specific numbers, clear self-contained passages, and strong entity signals — who wrote this, why they’re credible – are what get attributed. And for publishers specifically, if you’re being sourced heavily but not cited, that’s leverage. It’s the strongest argument you have in a licensing conversation,” King said.
Publishers have already caught on to this. Companies from Forbes to Time are using their prominence and citation rates across AI answer engines as a new metric they are trying to monetize with advertisers, though there still isn’t a consistent way to measure AI visibility across platforms. Publishers have also seen that being frequently cited doesn’t always translate proportionally into referral traffic.
“Clicks are no longer as important as the influence that is happening through the AI chatbots for the decision-making process,” Sharma said. “That’s why the citations, theonger about the clicks. It’s about… whether I’ve influenced ChatGPT or another chatbot to give [my brand] as an answer.”
More in Media
How Semafor is redefining the head of video role
As Semafor’s head of video, Adam Banicki sits at the intersection of the newsroom and the commercial team, reporting to the editor-in-chief with a dotted line to the revenue side.
Paywall violation ‘eviscerates’ fair use defense: What OpenAI-NYT twist means for publishers
For publishers, the newly unsealed material doesn’t just bolster their legal case; it validates years of frustration over how their work has been treated.
Walmart adds marketplace data, deeper AI features to Scintilla insights platform
The upgrade includes access to marketplace data, customizable dashboards, custom alerts, a deeper version of its Marty agent and a closer link with the Scintilla data platform.
