Microsoft has spent the last two years embedding Copilot into everything: Windows 11, Edge, Office 365, Bing, the taskbar, the Start menu. There is a dedicated Copilot key on new laptops. By pure distribution, Copilot should be the most-used AI assistant on the planet.
It is not. And the pattern should look very familiar to anyone who watched Internet Explorer slowly lose the browser wars despite shipping on every PC sold.
The Distribution Trap
Microsoft knows how to get software onto devices. That has never been the problem. Internet Explorer peaked at over 95% market share in the early 2000s because it came pre-installed on Windows. Bing processes billions of searches because it is the default in Edge. Neither dominance translated into loyalty. The moment users had a choice, they chose something else.
Copilot is following the same trajectory. Despite being one tap away on every new Windows machine, ChatGPT's mobile and desktop apps keep pulling ahead in daily active users. Claude is growing fast among professionals who need longer, more nuanced work sessions. Google's Gemini is integrated into Android and Search. When people actively choose an AI assistant rather than accepting the default, Copilot rarely wins.
Where Copilot Actually Falls Short
The core issue is not visibility. It is quality and identity. Copilot tries to be everything at once: a chatbot, a search engine, an image generator, a coding assistant, an Office automation layer. Each of those surfaces uses different underlying models with different capabilities, and the experience feels stitched together rather than coherent.
Ask Copilot a question in Edge and you get a Bing-flavored answer with citations. Ask it inside Word and you get a document-editing assistant. Ask it in Windows and you get a general chatbot. The branding is the same, but the product is not. Users who try Copilot in one context and find it mediocre have no reason to believe it will be better in another.
Compare that to ChatGPT, which is one product with one interface that does the same thing everywhere. Or Claude, which has a clear identity around careful, detailed reasoning. Copilot's identity is "the AI that Microsoft put in front of you."
The IE Playbook Does Not Work for AI
Bundling worked for decades because most users never changed their defaults. But AI assistants are different from browsers or search engines in one critical way: the quality gap is obvious within seconds. You paste the same prompt into ChatGPT and Copilot and you can immediately see which response is better. There is no learning curve hiding the difference.
Microsoft has the models (they fund OpenAI, after all), the distribution, and the enterprise relationships. What they lack is a product that makes people feel like Copilot is the best tool for any specific job. Until that changes, Copilot will keep playing the role Microsoft knows best: the thing that came with your computer that you replace as soon as you know better.