ChatGPT daily visits fall 22%, Gemini holds steady: report

OpenAI’s ChatGPT saw a 22% drop in average daily visits over the past six weeks, falling from about 203 million to 158 million, according to data from web analytics firm Similarweb.

During the same period, Google’s Gemini held steady with around 55–60 million daily visits, narrowing the usage gap between the two major AI platforms.

Gemini’s users also spent more time per visit and viewed more pages, with an average session lasting 7 minutes 20 seconds and 4.3 pages visited, compared to ChatGPT’s 6 minutes 32 seconds and 3.8 pages.

Analysts cite ChatGPT’s decline to both seasonal holiday slowdowns and increased competition from Gemini, which rolled out a major model upgrade in November 2025 and new features in mid-December.

According to Similarweb, other AI models also saw notable activity in December, with Perplexity.ai recording 154.9 million visits, Grok reaching 247 million, and DeepSeek at 282 million.

🔗 Source: The Economic Times

🧠 Food for thought

Implications, context, and why it matters.

Web traffic alone misses the full picture of ChatGPT’s user engagement

Similarweb logged a 22% drop in ChatGPT desktop and mobile web visits. That metric excludes native app usage and API consumption.
To tell if this is a real drop or a channel shift, we need app downloads and monthly active users for November. We also need the same for December 2025, which Similarweb’s web visits miss.
Gemini from Google logs 7 minutes 20 seconds per session, while ChatGPT averages 6 minutes 32 seconds. Longer sessions can signal deeper task completion, not necessarily superior product–market fit (how well a product meets user needs).
In December, Similarweb counted 282 million visits for DeepSeek and 247 million for Grok. Perplexity.ai reached 154.9 million, which complicates cross-platform comparisons.

Multi-model strategies create opportunities in AI infrastructure and governance

With usage split across many chatbots, enterprises need multi LLM gateways. This middleware routes requests to different models based on cost, latency (response speed) and quality needs.
Vendors with routing and observability tools (systems that monitor and log performance and reliability), such as unified dashboards across providers, help teams manage AI spend as competition grows.
Investors can find upside in the infrastructure layer. Vendor-agnostic platforms (works across providers) that manage selection, fallback strategies (automatic switching when a model fails or degrades), and cost allocation will gain value.
Enterprise AI operations teams should benchmark specific use cases across platforms now. Pricing plus capability gaps between models create arbitrage opportunities (taking advantage of price and performance differences). These gaps may narrow as the market matures.