Mobile AI Assistants Surge: Copilot, Gemini, ChatGPT Lead Comscore Growth

Mobile usage of AI assistants has taken a measurable lead over desktop in recent months, with Comscore reporting mobile reach for AI tools rising to 73.4 million users (a 5.3% increase) while PC usage fell roughly 11.1%, and the largest mobile growth rates concentrated in Microsoft Copilot, Google Gemini and — to a lesser degree — OpenAI’s ChatGPT.

Background / Overview​

Comscore introduced an AI usage tracker earlier this year that tracks visits across 117 AI tools in nine categories, measuring deduplicated audience reach on mobile and desktop. That dataset — sampled from a digital panel and aggregated across web and native apps — is the basis for the March–June window that produced the headlines about rapid mobile adoption and platform-specific growth rates.
Across the March–June snapshot Comscore reported:

• Total mobile reach (web + native apps): 73.4 million users (up 5.3%).
• Desktop/PC usage: a decline of approximately 11.1% in the same period.
• Mobile adoption of AI assistants from November 2024 to June 2025 grew 82% overall.
• Fastest mobile growth (March–June 2025): Microsoft Copilot +175%, Google Gemini +68%, OpenAI ChatGPT +17.9%.

Comscore’s cross-visitation and deduplication analysis also shows that more than 85% of heavy AI assistant users primarily use a single platform, and that OpenAI’s mobile user base appears comparatively more loyal than Google’s or Microsoft’s, whose users explore across platforms more often. Smriti Sharma of Comscore summarized the broader behavioral shift as users treating mobile AI assistants as personal, always-available companions and gravitating to tools that feel native to phones.

---

What the numbers actually measure (methodology and caveats)​

Panel-based visitation vs other telemetry​

Comscore’s numbers are drawn from a panel-based measurement that captures visits and in‑app usage on a representative device panel. This method emphasizes deduplicated reach — unique users across devices — which answers a different question than referral or session-share trackers that count outbound web traffic. Independent tracking services that measure web referral and session share (for example, StatCounter) have historically shown ChatGPT commanding a dominant share of chatbot-driven web referrals, which explains why different trackers can tell complementary but distinct stories about market leadership.

Why percentage growth can mislead​

A headline like “Copilot up 175%” is dramatic but needs absolute context. High percentage growth from a smaller base can still be well below the raw audience of market incumbents. Comscore’s growth rates were first amplified through media partners that translated percentages into estimated absolute user counts (e.g., Copilot’s mobile user base reported in some writeups at roughly 8.8 million and ChatGPT at ~25.4 million for the same window), but those absolute figures come from secondary reporting and should be treated as estimates unless Comscore or the vendors publish explicit counts. Cross-checks with other telemetry vendors are essential to avoid misreading competitive positions.

What Comscore’s tracker covers — and what it doesn’t​

• Includes native apps and mobile web visits for a curated set of 117 AI tools across consumer and productivity categories.
• Produces deduplicated audience reach by platform and cross-visitation patterns.
• Does not directly measure API query volume, cloud inference counts, or paid-subscription revenue; those require vendor disclosures or cloud-provider financials for confirmation.
  Because of differences in definitions, readers should treat Comscore’s reach as a robust indicator of user behavior on devices rather than a full accounting of backend usage or revenue.

---

The mobile pivot: Why phones are winning casual AI interactions​

Short sessions, multimodal inputs, and device-native UX​

Mobile sessions are characteristically shorter and more context-specific. Users increasingly employ assistants for quick productivity tasks — drafting emails, summarizing a meeting note, snapping a photo and asking for analysis — and those flows favor assistants that are optimized for voice, camera/image inputs, and tight OS integration. Comscore’s analysts attribute the mobile tilt to these behavioral patterns: convenience plus multimodal inputs make phones the natural home for many assistant interactions.

Performance and latency considerations​

Mobile-first AI experiences place a premium on latency, efficient on-device routing, and streamlined UI. Assistants that can give succinct results with minimal friction win quick-use scenarios. That means vendors will prioritize smaller, optimized models (or hybrid on-device/inference routing) and UX polish over raw model size in mobile product roadmaps.

---

Platform dynamics: Copilot, Gemini, ChatGPT — growth drivers and strategic differences​

Microsoft Copilot — enterprise wedge, rapid mobile lift​

Microsoft’s Copilot shows the sharpest percentage gains on mobile in Comscore’s window. The explanation is largely structural: Copilot is not only a consumer app but also deeply embedded in Microsoft 365, Windows, Edge and enterprise workflows, which creates multiple low-friction distribution paths. Admin controls, single sign‑on, and enterprise licensing mean IT organizations can enable Copilot at scale across employee devices — including mobile — producing fast adoption in environments where the assistant is provisioned rather than discovered. That enterprise-first distribution is a core reason behind Copilot’s dramatic percentage growth.Key implications:

• Copilot benefits from bundling with productivity apps and OS-level integration.
• Enterprise enablement can convert into quick mobile adoption when organizations allow or push Copilot to employee devices.
• Rapid growth can continue as Microsoft ties Copilot into licensing and device management channels.

Google Gemini — device distribution and Android positioning​

Google’s Gemini growth is linked to device preloads and ecosystem placement (Pixel devices, Android integration, and Google Workspace touchpoints). Preinstallation and default presence on devices are powerful user-acquisition levers that can quickly translate into millions of mobile users without the same enterprise rollout mechanics Microsoft uses. Gemini’s strengths also include multimodal reasoning and long-context features that align well with creative and search-centric mobile use cases.

OpenAI ChatGPT — scale incumbent, high retention​

ChatGPT remains the scale leader in many independent referral and session-share trackers, even when Comscore shows faster percentage growth for competitors. Where Copilot and Gemini are growing quickly from smaller bases, ChatGPT’s absolute audience remains very large and sticky. Comscore’s cross-visitation data also indicates OpenAI mobile users show higher platform loyalty, meaning they are less likely to hop among assistants compared with some competitors’ users. That retention is a competitive moat: scale begets ecosystem effects (developer integrations, plugins, and content partnerships) that keep core audiences engaged.

---

Distribution, bundling and the new user-acquisition economics​

Large vendors have weaponized distribution:

• Microsoft uses enterprise licensing and embedded productivity integrations to push Copilot.
• Google leverages device preloads and Android defaults to expand Gemini’s reach.
• OpenAI retains consumer mindshare via direct-to-consumer product experiences and a broad developer ecosystem.

These channel plays create varied acquisition economics. Device preloads and enterprise enablement can produce fast user counts and percentage growth, but they also change the nature of consumer consent and product discovery. The practical effect: an assistant can become the default experience inside a company or on a phone, creating switching costs that extend beyond simple product preference.

---

User behavior: loyalty, cross-visitation, and habit formation​

Comscore’s deduplication shows heavy users frequently stick to one assistant: more than 85% of top users primarily use a single platform. That suggests AI assistants are evolving into habitual workflow components rather than novelty experiments. Habit formation favors:

• Assistants that are defaulted by the OS or IT.
• Assistants with deep integrations into frequently used apps.
• Experiences that feel personal and quickly solve common tasks.

The loyalty finding should be read with nuance: while top users tend to be platform-loyal, casual users still explore multiple tools, particularly when trying new multimodal features (image input, live camera analysis, or voice interfaces) that may be unique to specific assistants.

---

Commercial and infrastructure implications​

The commercialization of assistant usage translates directly into cloud demand and vendor economics. Independent reporting tied the usage surge to larger financial dynamics: Azure’s reported growth and OpenAI’s subscription/API revenue milestones have been cited as evidence that usage is converting into meaningful commercial outcomes. Those business realities create incentives for deeper integrations and for vendors to prioritize features that increase cloud consumption, subscriptions, or enterprise seat attachments. That feedback loop accelerates feature rollouts and ecosystem engineering, but also concentrates bargaining power among platform owners who control distribution channels.

---

Risks and governance concerns​

Measurement and reporting risks​

• Growth-rate headlines can obscure absolute scale; percentage figures require baseline context to be meaningful.
• Different measurement methods (panel reach vs session/referral tracking vs API/inference counts) can yield divergent narratives. Decision-makers should triangulate across trackers rather than rely on a single headline.

Privacy and data controls​

Mobile assistants often require camera, microphone, and location permissions to deliver rich multimodal experiences. Those permissions increase the attack surface for data misuse and raise complex compliance questions for regulated enterprises. Vendors and IT administrators must ensure:

• Clear consent flows for end users.
• Contractual clarity on data usage and retention.
• Configuration controls that prevent unintended data leakage from corporate apps to third-party services.

Platform lock-in and procurement risk​

The habit formation and distribution advantages enjoyed by big vendors can produce vendor lock-in, particularly for organizations that standardize on a single assistant across their productivity stack. Procurement teams must weigh short-term productivity gains against long-term strategic flexibility, focusing on data portability, auditability, and contract provisions that let them switch or dual‑footprint as requirements evolve.

Unverified and hyperbolic claims​

Some viral lists and rankings in the wider market have inflated absolute “user” numbers that are not verifiable with independent telemetry. These claims — often framed as cumulative or ambiguous “users” — should be flagged and treated cautiously until vendors or neutral samplers publish transparent methodologies. Where Comscore provides panel-derived reach, other independent trackers measure different slices (like web referrals), and some headline “billions of users” claims lack corroboration. Mark these as unverified in procurement or strategic analysis.

---

What this means for Windows users, IT admins and enterprises​

For Windows and Microsoft‑centric environments​

• Copilot’s mobile growth is a strategic opportunity if the organization already standardizes on Microsoft 365: admins can enable Copilot to accelerate productivity gains across devices, including mobile.
• Governance remains critical: deploy with policies that restrict sensitive data exposure and ensure audit trails for assistant queries that touch corporate documents or identity systems.

Practical actions for IT teams​

• Inventory AI touchpoints: identify where assistants are embedded across apps (Outlook, Teams, Word, mobile apps).
• Establish clear governance: set DLP rules, retention policies, and role-based access for assistant features.
• Pilot & measure: run controlled pilots to instrument latency, hallucination rates, and productivity impact before broad rollouts.
• Contractual protections: require transparency on data usage, provide for portability, and include SLA metrics for uptime and model behavior when relevant.
• Consider multi-assistant strategies: maintain the ability to route sensitive workflows to vetted, auditable models while letting less-sensitive workflows use consumer-grade assistants for convenience.

For everyday Windows users​

• Expect more AI features to appear in mobile and desktop apps; learn how permission dialogs affect privacy and be cautious about sharing sensitive screenshots, files, or account credentials in assistant prompts.
• Use paid tiers or enterprise configurations when possible if data residency and stronger SLAs are priorities.

---

How journalists, analysts and buyers should interpret the Comscore window​

• Treat the Comscore dataset as one high-quality panel-based lens on user behavior that is particularly valuable for mobile and app usage patterns.
• Complement Comscore with session/referral trackers (e.g., StatCounter-style telemetry) and vendor disclosures for a fuller picture: panel reach answers the “how many unique people used this on a device?” question while referral/session data answers “which assistant is sending traffic or producing outbound referrals?”
• Demand absolute figures where decisions depend on scale (for example: licensing costs, API capacity planning, or migration timelines), and insist that vendors provide transparent definitions for terms like “user”, “session”, and “visit.”

---

Looking ahead — product and market outlook​

Mobile-first assistant design will accelerate:

• Vendors will optimize latency and build hybrid local/cloud inference strategies to improve responsiveness on phones.
• Multimodal capabilities (camera + voice + text) will become baseline expectations for mobile assistants.
• Distribution deals (preloads, carrier bundles, enterprise enablement) will remain the fastest path to user growth, so procurement and antitrust scrutiny are likely to follow if default placements materially disadvantage competitors.

At the same time, measurement sophistication must improve. The industry needs:

• Standardized definitions for user metrics across panels, referral trackers, and API counts.
• Independent audits of vendor-reported usage where those numbers materially affect procurement or public policy.
• Clearer privacy-preserving telemetry frameworks so product teams can measure adoption without compromising user data protections.

---

Conclusion​

Comscore’s March–June snapshot provides a clear signal: mobile is now a dominant venue for many AI assistant interactions, and the ecosystem winners will be those that combine product excellence with distribution muscle and trustworthy governance. The headline growth figures — Copilot’s 175% surge, Gemini’s 68% increase, ChatGPT’s steadier 17.9% rise — are real and instructive, but they must be read alongside absolute scale, different telemetry methodologies, and the commercial incentives shaping where assistants live on devices and in enterprises. Decision-makers should use Comscore as a valuable behavioral lens, triangulate with other telemetry, and treat rapid mobile adoption as a call to harden privacy, governance, and procurement practices before default assistants become hard-to-replace defaults.

---

Source: Research Live Mobile AI tool usage increasing, says Comscore | News | Research live