AI Discovery Is Splintering Past ChatGPT. Here’s How to Adapt
ChatGPT's share of AI prompts fell from 70% to 50% in 2026 as Gemini, Claude and Google AI Overviews rose. Here's how to track real visibility.
If your AI visibility strategy still begins and ends with ChatGPT, the ground has already shifted. Comscore’s Q2 2026 AI Intelligence Report, released today, shows generative AI discovery is fragmenting quickly across assistants and search surfaces.
The numbers behind the splintering
ChatGPT’s share of prompt volume dropped from 70% in January to 50% in June 2026, while Gemini nearly doubled from 17% to 30% and Claude climbed from 2% to 11%. Overall usage is growing, but attention is spreading across more platforms.
Google’s AI Overviews are also expanding their reach. Desktop searches triggering an AI Overview rose from 25.8% in July 2025 to 39.4% in June 2026. That trend matters because every AI-generated answer on a search results page is another chance for Google to answer directly, not send a click.
ChatGPT still reached 168 million desktop conversations in June, and multiplatform visitation grew to 99 million from 59 million a year earlier. But by June 2026, 35% of desktop users and 29% of mobile users had visited an AI assistant, up from 25% and 12% in January 2025.
Sourced is not the same as cited
The report highlights a crucial distinction for brand managers and publishers: being pulled into an AI model’s retrieval pool is not the same as being visible in the final answer. Tripadvisor, for example, accounts for 61% of travel sourcing by AI platforms but only 21% of citations.
Comscore’s Smriti Sharma frames the measurement shift this way: “Did the AI use me? Did they cite me? Did they recommend me, or did they mention my competitors instead?” That question changes the objective from ranking to influence.
What this means for your AI visibility
Generative engine optimisation can no longer be a single-platform game. Marketers need to diagnose where their audience actually asks questions, then track retrieval, citation and referral across ChatGPT, Gemini, Claude and AI Overviews. Citations alone are a lagging metric.
Use this funnel as a rough scorecard:
- Retrieval: Is your page pulled into the assistant’s sub-queries as a source?
- Citation: Does the final answer attribute a claim or recommendation to your brand?
- Influence: Does your brand appear when the assistant answers a category question, even without a click?
- Revenue signal: Are those citations creating licensing leverage, referrals or measurable demand?
Michael King, founder and CEO of iPullRank, points to the same fundamentals that get any brand attributed: original data, specific numbers, clear self-contained passages and strong entity signals showing who created the content and why they are credible. King adds that publishers who are sourced heavily but cited rarely actually hold leverage: “It’s the strongest argument you have in a licensing conversation.”
Consumers are developing different relationships with different assistants. Gemini is growing because it is bundled into Google, Android and Workspace. Claude is gaining among knowledge workers and technical audiences. That is not one audience asking one type of question. Plan for a portfolio of AI touchpoints, not a single assistant.
Source: Digiday
