Microsoft Copilot’s mobile usage skyrocketed 175% between March and June 2025, making it the fastest-growing major AI chatbot, according to new Comscore data. Yet that triple-digit surge didn’t knock ChatGPT from its throne—the OpenAI chatbot still commands the largest user base and the highest loyalty among consumers. The contrasting trajectories reveal a market shaped by distribution muscle, platform lock-in, and a rapidly shifting balance between desktop and mobile AI use.
Comscore’s tracking of 117 AI tools across desktop and mobile during that three-month window painted a picture of explosive growth for Microsoft’s assistant, especially on phones. Copilot’s 175% gain left rivals in the dust: Google Gemini grew by 68%, and ChatGPT managed a relatively modest 17.9% increase. But percentages can mislead without absolute scale. Media partners who translated Comscore’s growth rates into raw user estimates pegged Copilot’s mobile audience at roughly 8.8 million, while ChatGPT’s mobile reach stood around 25.4 million. On desktop, ChatGPT still attracted 36.1 million users, compared with Copilot’s 23.1 million. The numbers underscore a velocity-versus-scale tension that defines today’s AI assistant wars.
Why the growth rates matter—and what they hide
Headline percentages like “175% growth” are catnip for tech readers, but they need careful unpacking. Smriti Sharma, SVP of Custom IQ at Comscore, told ZDNET that ChatGPT’s lower growth rate reflects its head start: “ChatGPT was the first mover, building a very large user base early on. That means its adoption curve is further along, and the growth now appears steadier. By contrast, Gemini and Copilot are still in a newer stage of adoption, so their growth rates are higher.”
A 175% increase from a base of 3.2 million monthly mobile users yields 8.8 million—significant, but still just a third of ChatGPT’s mobile audience. ChatGPT’s 17.9% growth on a base of tens of millions added far more absolute users. This distortion is a frequent trap when comparing AI chatbots: velocity can overshadow sheer magnitude. Moreover, Comscore’s panel methodology captures deduplicated audience reach across devices, not raw query volume. That differs sharply from referral-based trackers like StatCounter, which show ChatGPT holding a commanding 80–83% share of chatbot referral traffic. The two datasets answer different questions: who is using the tools, versus which tool is sending users back to the wider web.
Copilot’s ecosystem turbocharger
Copilot’s mobile surge didn’t happen by accident. It rides on Microsoft’s aggressive embedding of the assistant into Windows, Microsoft 365 apps, Edge, and enterprise management consoles. For IT departments, rolling out Copilot across an organization is often a matter of flipping a switch in admin portals—a frictionless path that consumer discovery can’t match. “Copilot’s mobile growth comes mainly from Microsoft’s ecosystem integration and productivity use cases,” Sharma explained. “The emphasis from Microsoft on its mobile-first approach fits well with lightweight productivity tasks.”
That framing is critical. Users aren’t necessarily seeking out a chatbot; they’re drafting emails in Outlook, summarizing Teams meetings, or editing documents on the go—tasks where Copilot appears as a natural extension. This bundling economics model creates growth that looks more like a feature uptake curve than a standalone product launch. It also explains why Copilot’s mobile gains were so concentrated: quick, work-related queries on a phone align perfectly with Microsoft’s productivity narrative.
ChatGPT’s unshakable lead—for now
Despite slower percentage growth, ChatGPT’s grip on the chatbot market remains formidable. It reached 36.1 million desktop users and 25.3 million mobile users in the March-to-June window, according to Comscore, dwarfing its rivals. StatCounter’s referral data reinforces this dominance: roughly four out of five chatbot-driven website visits still come from ChatGPT, signaling not just a large user base but the kind of habitual, intent-driven usage that brands crave.
Loyalty is another moat. Comscore found that more than 85% of mobile AI users stick with a single assistant, and ChatGPT users were the most loyal. That retention advantage creates a self-reinforcing cycle: more users beget more data, better models, and a richer plugin ecosystem. While Copilot and Gemini users hop between assistants more often, ChatGPT’s audience treats it as a destination, not a trial.
Mobile overtakes desktop: a tectonic shift
Overall AI usage on mobile grew by 5.3% in the period, reaching 73.4 million users, while desktop usage shrank 11% to 78.4 million. The gap is closing fast, and the implications are profound. Mobile sessions are shorter, more context-sensitive, and often tied to device sensors like microphones and location. Assistants optimized for snappy, lightweight interactions—like Copilot’s meeting summary generation or quick email replies—stand to gain the most as users shift away from deep-research desktop sessions.
For Microsoft, the mobile moment aligns with its “productivity anywhere” pitch. But Google Gemini, too, has a distribution card to play: pre-installed on Pixel devices and integrated into Android at the system level, Gemini’s 68% growth benefited from being in front of users at device setup. The 14.2 million mobile users it garnered show that when assistants come baked into hardware, adoption curves steepen quickly.
The rest of the field and the fragmentation ahead
Beyond the big three, Comscore’s tracker highlighted how specialized AI tools are carving out substantial audiences. Canva’s AI-powered design assistant pulled in 15.1 million desktop users, illustrating that for many, an “AI chatbot” isn’t a general conversationalist but a vertical tool embedded in a familiar workflow. Grammarly, Octane AI, and Voicemod also ranked high, each solving discrete problems. This fragmentation suggests the market for AI interactions won’t coalesce around a single winner but will splinter along use-case lines—a dynamic that favors platform owners like Microsoft and Google who can weave assistants into multiple surfaces.
Commercial incentives and the bundling endgame
The user growth isn’t just a vanity metric; it’s revenue. OpenAI reportedly hit $10 billion in annual recurring revenue by mid-2025 from subscriptions and API access, while Microsoft’s Azure revenue surged past $75 billion, with AI workloads flagged as a key driver. These numbers create a powerful incentive to push assistants deeper into enterprise and consumer products, turning every interaction into a cloud consumption event. Copilot’s growth, then, isn’t just about user numbers—it’s a strategic lever to boost Microsoft 365 stickiness and Azure compute demand.
That monetization imperative has a dark side. Vendors are increasingly tying assistant features to premium tiers and cloud contracts, risking lock-in. For enterprises, the line between a helpful productivity tool and a dependency that’s hard to unwind is becoming perilously thin.
Measurement minefields and the truth behind the claims
Any analysis of AI assistant growth must grapple with methodology. Comscore uses a consumer panel to estimate deduplicated reach across apps and sites; StatCounter counts referral traffic. Both are credible but incomparable. Copilot can show 175% growth on Comscore while barely moving the needle on StatCounter because its users often don’t click out to websites—they stay inside the Microsoft app ecosystem. That’s not a failure; it’s a design choice. But it means no single metric tells the whole story.
Then there are the wild vendor claims. Viral posts and press releases occasionally tout multi-billion-user numbers that evaporate under scrutiny, often conflating cumulative app installs with monthly active users or counting every API call as a unique “user.” Without standardized, transparent measurement, the AI landscape will remain a fertile ground for hype. The remedy: always ask for methodology, base sizes, and deduplication rules before drawing conclusions.
Privacy, governance, and concentration risks
As assistants creep into enterprise workflows and mobile devices, they bring thorny governance questions. Copilot’s integration with Office 365 means it can access sensitive emails, documents, and meeting transcripts. IT leaders must map data flows, classify information that shouldn’t touch public models, and demand contractual guarantees around data residency, retention, and deletion. The same goes for Gemini’s device-level permissions and ChatGPT’s data handling policies.
Market concentration is another concern. ChatGPT’s dominance in referral traffic suggests a single point of failure for businesses reliant on AI-driven customer interactions or content discovery. If that concentration persists, it could stifle innovation and reduce the diversity of models available in the mainstream. Policymakers and open-source advocates are already raising alarms about platform power and the need for interoperability standards.
What Windows users and IT leaders should do now
For everyday Windows users, Copilot’s deepening integration means its features will appear more frequently in taskbars, email clients, and document editors. Lean into it for lightweight tasks—drafting, summarizing, scheduling—but review app permissions and Microsoft account settings to control data sharing. If privacy is a top concern, choose the Copilot plan that aligns with your needs, and disable features that send data you’re uncomfortable with.
For IT and procurement teams, the message is caution. Pilot Copilot with a limited user group to measure real productivity gains, not just perception. Validate governance controls: audit logs, single sign-on, data retention policies, and the ability to restrict certain features. And don’t bet the farm on a single assistant. Keep testing alternatives so you’re not locked in if pricing or functionality changes.
Power users should adopt a multi-tool strategy. ChatGPT remains the heavyweight for deep research, creative writing, and plugin-driven tasks. Copilot shines when you’re inside Word, Excel, or Outlook and need context-aware help. Gemini excels at Android-centric queries and tight integration with Google’s services. Mix them based on the workflow, and always validate critical outputs—no assistant is infallible.
The road ahead
The Comscore snapshot captures a market in transition. Copilot’s 175% mobile sprint shows what happens when a platform giant flexes its distribution arteries. But ChatGPT’s scale and loyalty prove that being first and best still counts for a lot. Going forward, expect the battleground to shift further toward mobile, with assistants becoming ambient features of operating systems rather than stand-alone apps. Monetization will intensify, and with it, the blurring of lines between productivity tool and vendor lock-in.
The lesson for everyone watching this space is simple: look past the headline growth numbers. Demand transparency, diversify your AI toolkit, and adopt based on evidence, not enthusiasm. The bot wars are far from over, but the winners will be the ones who earn long-term trust, not just a spike in downloads.