The news: OpenAI’s dominance of AI web traffic is weakening as Google’s Gemini gains speed.

• ChatGPT’s global AI web traffic share dipped to 64.6% in mid-January, its first time dropping below 65%, per Similarweb.
• Gemini’s traffic share has grown 315% over the past year to nearly one-quarter (22%) of the market.

Meanwhile, xAI’s Grok surpassed DeepSeek for the first time, marking a rankings shake-up for second-tier AI platforms. Use of Anthropic’s Claude remained stable, while Perplexity and Microsoft Copilot continued to lose share.

Zooming out: These platforms are competing for the attention of 133 million genAI users in the US alone in 2026, per our forecast—that’s about 39% of the US population and growing. Considering genAI’s growing role in the discovery and shopping journey, the AI service that pulls in the most users may be the most attractive to advertisers.

As AI experimentation grows, users are discovering which platforms best serve specific needs as they establish go-to generative engines.

• Gemini, ChatGPT, and Perplexity could be best suited for general information queries, though the latter isn’t as focused on complex reasoning.
• Claude is a strong coding tool and is especially popular among workplace users—80% of Anthropic’s business comes from enterprises, CEO Dario Amodei said at the World Economic Forum.
• Grok could be gaining traction for casual asset generation and increasing business adoption, though its emerging use for creating explicit deepfakes—which regulators are investigating—could dissuade new adopters.

The fight for traffic: As use cases diverge, these companies are also tackling user acquisition and retention in different ways.

• OpenAI is experimenting with a broad range of subscription tiers and soon, direct in-chat advertising.
• xAI emphasizes free access and premium feature bundling to boost Grok adoption.
• Gemini replaced Google Assistant on Android and will soon power Apple’s Siri, effectively doubling its reach among US voice assistant users.
• Microsoft relies on enterprise distribution by putting Copilot into existing software rather than competing on standalone consumer usage.

Implications for marketers: Difficulties measuring AI traffic consistently will arise as platforms fragment and attribution standards evolve. This could mean brands need to develop platform-specific content strategies rather than assuming a standardized generative engine optimization (GEO) strategy will work across all platforms.

Experimentation is key: A/B test a variety of content—such as contextualized web content and search-style answers like FAQ pages—to determine what surfaces best on each AI platform.