Generative AI in Transition: Platform Competition, Adoption Patterns and Emerging Governance Challenges
Executive Summary
Generative artificial intelligence (GenAI) has evolved rapidly from a novel technological curiosity into a mainstream digital infrastructure shaping how individuals and organisations search, create, communicate and work. While ChatGPT remains the dominant platform globally, the GenAI ecosystem has become increasingly diverse, with Microsoft Copilot, Google Gemini, Claude, Perplexity and DeepSeek each occupying distinct positions within the market. Survey and web-traffic data indicate that ChatGPT leads both consumer and enterprise adoption, although platform choice increasingly reflects ecosystem integration, governance requirements and task-specific capabilities rather than model quality alone.
The evidence also highlights important differences between individual and organisational use. Consumers primarily employ AI for information seeking, workplace assistance and creative tasks, whereas organisations focus on productivity gains, return on investment and integration with existing digital infrastructures. At the same time, rising adoption is accompanied by significant concerns regarding privacy, security, reliability, regulation and societal impact.
This blog reviews current evidence on GenAI platform adoption, compares leading AI tools, examines the factors driving ChatGPT's continued dominance, and explores the emerging governance challenges that will shape the next phase of AI diffusion across society and organisations.
1. Introduction
Since the launch of ChatGPT in late 2022, generative artificial intelligence has moved from technological novelty to an everyday tool used by millions of people worldwide. Yet the AI landscape is no longer defined by a single platform. Alongside ChatGPT, tools such as Microsoft Copilot, Google Gemini, Claude, Perplexity and DeepSeek have created an increasingly competitive market in which users choose platforms based on productivity needs, organisational ecosystems, specialised capabilities and trust considerations.
Understanding who uses these platforms, for what purposes, and why some succeed more than others is becoming increasingly important for educators, businesses, policymakers and researchers. Drawing on recent survey evidence and platform usage data, this blog examines the emerging GenAI ecosystem, explores the drivers of platform adoption, and discusses the opportunities and risks associated with the growing integration of AI into everyday life and organisational practice.
2. From Novelty to Infrastructure: The Rise of Generative AI
The popular history of GenAI begins with ChatGPT's release in late 2022, when a single product seemed poised to dominate AI chat as Google dominates search (Kaufman, 2025). Adoption was not linear. Global web traffic to ChatGPT plateaued at roughly two billion visits per month in summer 2024, before GPT-4o, voice interaction and a polished mobile application renewed growth; by June 2025 visits exceeded five billion, up 160 per cent year on year. The sector has also proved volatile: DeepSeek reached ten million users in twenty days in February 2025, twice as fast as ChatGPT, yet had lost about four-fifths of that traffic by June (Kaufman, 2025).
Within the United States, survey evidence shows steadier diffusion. Reported ChatGPT use rose from 18 per cent of adults in 2023 to 23, 34 and 44 per cent in 2024, 2025 and 2026, while 49 per cent now use any AI chatbot, against 33 per cent in 2024 (Figure 1) (Pew Research Center, 2026). Because Pew widened its 2026 question to all respondents, whereas earlier waves asked only those aware of the tool, the trend is indicative rather than strictly like-for-like. Daily use, at 24 per cent, suggests growing habituation. Figure 1 measures national adoption; the global figures below measure web activity.
3. The Platform Landscape: Who Uses What, and for What Purpose
Global web-traffic data for July 2025 indicate a concentrated market. ChatGPT recorded an estimated 416 million monthly unique visitors, against 122 million for Gemini, roughly 50 million for DeepSeek, 19 million for Perplexity and 16 million for Claude (Figure 2) (Kaufman, 2025). These Similarweb estimates cover desktop and mobile web only, so they understate activity on app-centred platforms.
Audiences also differ in character. ChatGPT users averaged 7.1 minutes and 4.3 pages per visit, against 6.3 minutes for Gemini, 5.4 for Perplexity and 6.6 for Claude (Figure 3). Claude is favoured by developers, with strong overlap with GitHub, Stack Overflow and Figma; Perplexity skews towards business and research; Gemini benefits from Google and device-default distribution; and ChatGPT users resemble the general internet public. Choice is not exclusive, with 40–60 per cent user overlap between ChatGPT and other platforms (Kaufman, 2025).
The American ranking is broadly consistent but measures something different. In February 2026, 44 per cent of adults reported using ChatGPT, 24 per cent Gemini, 17 per cent Copilot, 14 per cent Meta AI, 8 per cent Grok, 6 per cent Claude and 3 per cent Character.ai (Figure 4) (Pew Research Center, 2026). Copilot, absent from the global traffic table, registers 17 per cent, plausibly reflecting its embedding in familiar software (Section 4).
Searching for information (42 per cent) and work tasks (38 per cent of employed adults) are the leading purposes, followed by entertainment (25 per cent), image and video creation (24 per cent), medical advice and diet and fitness information (20 per cent each), news (13 per cent), emotional support (10 per cent) and companionship (4 per cent) (Figure 5).
4. Explaining ChatGPT's Ascendancy
Four explanations recur across the sources. The first is scale and first-mover advantage. ChatGPT attracts roughly three times the unique visitors of Gemini and about eight times its visits (5.4 billion against 650 million; author's calculation from Kaufman, 2025). Trackers cited by WindowsForum (2025) place its share of chatbot referral traffic at 70–80 per cent, although scope and method vary.
The second is product breadth: multimodal input, voice interaction and a polished mobile application opened new use cases, while workflows such as code review, spreadsheet formulas and study aids appear to have fostered daily habits (Kaufman, 2025). The third is ecosystem effects: developer tools, integrations and tiers from free to enterprise support adoption at individual and institutional levels (WindowsForum, 2025). The fourth is habit and switching costs, since organisations build prompts, templates and verification routines around a tool (WindowsForum, 2025).
The American survey confirms the demand side: ChatGPT's reach has more than doubled since 2023 and exceeds every other named chatbot by at least twenty percentage points (Figures 1 and 4). Public dominance does not, however, imply superiority everywhere; distribution and integration can outweigh raw model quality (WindowsForum, 2025).
5. Individual Experimentation and Institutional Adoption
Personal and organisational use differ in motivation, measurement and risk. Personal use in the United States centres on information seeking, entertainment and creative tasks (Figure 5), and adults are more likely to say chatbots help rather than hurt their productivity (30 against 5 per cent) and how informed they are (28 against 5 per cent); effects on happiness and relationships are mostly perceived as neutral (Pew Research Center, 2026).
Organisational use is framed in terms of measurable return. A Wharton Human-AI Research survey of business leaders, summarised by WindowsForum (2025), found that 46 per cent use generative AI daily and that 72–75 per cent of organisations report positive return on investment (Figure 6). Comparison with the American public's 24 per cent daily chatbot use is suggestive but not like-for-like: the populations differ, as do the objects measured (generative AI versus chatbots), and the Wharton sample's geographic scope is not stated. The return figures also reflect perception rather than audited outcomes.
Tool selection differs accordingly. Procurement turns on integration, governance and contractual guarantees such as non-training clauses. Copilot benefits from embedding in Word, Excel, Outlook and Teams; Gemini's Workspace integration has converted less readily into habit; and Claude, though well regarded, lags in public telemetry that may understate private deployments (WindowsForum, 2025). Claude's 6 per cent American usage (Figure 4) and 16 million global web visitors (Figure 2) both suggest a niche, but they belong to different contexts and cannot be combined.
6. Public Scepticism and Ethical Considerations
Rising adoption has not produced enthusiasm. Forty per cent of American adults expect AI's impact on society over the next twenty years to be negative, against 16 per cent positive; for personal impact the figures are 31 and 23 per cent, with many anticipating mixed effects or unsure (Figure 7). Scepticism is not confined to non-users: adults aged 18–29, who stand out for their use of AI, are the most likely to expect negative societal consequences (48 per cent, against 37 per cent of those aged 50 and over) (Pew Research Center, 2026).
Concern is broader still (Figure 8). Seventy-one per cent expect AI to make personal information less secure, and 63 per cent consider it is advancing too quickly, against 2 per cent too slowly. Sixty-seven per cent have little or no confidence in government regulation, up from 62 per cent in 2024, and 59 per cent lack confidence in companies to act responsibly. Republicans' lack of confidence in regulation fell from 70 to 61 per cent, while Democrats' rose from 54 to 74 per cent (Pew Research Center, 2026).
Several ethical issues follow. Privacy and surveillance are salient: 35 per cent of adults own a smart speaker and 18 per cent an AI-enabled doorbell, and Pew notes associated surveillance and child-safety concerns. Reliability matters, since one in five adults seek medical advice from chatbots while organisations stress human verification of hallucinations (WindowsForum, 2025). Ten per cent use chatbots for emotional support and 4 per cent for companionship, practices whose consequences remain contested. Organisations additionally face data-leakage and provider-concentration risks (WindowsForum, 2025). The sources identify these concerns without resolving them.
7. Prospects: Consolidation, Specialisation and Governance
American adoption has risen in each wave, suggesting further diffusion, although unease about speed and security may temper trust. AI is also becoming ambient: 60 per cent of adults read AI-generated search summaries (Pew Research Center, 2026).
Competition appears to be shifting from model quality towards distribution, through productivity suites and device ecosystems, while specialisation preserves niches: research tools, long-context offerings and privately hosted models where data sovereignty matters (WindowsForum, 2025). The market is diversifying beneath ChatGPT's dominance, and DeepSeek shows that novelty can fade quickly (Kaufman, 2025). Governance is likely to become decisive: with two-thirds of Americans doubting regulatory capacity and organisations demanding audit reports and contractual guarantees, trust may function as a competitive attribute alongside capability.
Evidence gaps remain: the sources lack comparable global survey data, web estimates exclude app use, and the material dates from mid-2025 to early 2026. Future research should pair population surveys with audited organisational data.
Dr Bidit L. Dey
Professor of Marketing
Northumbria University, UK
LinkedIn: Bidit Dey | LinkedIn
References
Kaufman, D. (2025, 8 July). Mid-2025 scorecard: Who’s actually using the top AI chatbots? Siteline. https://siteline.ai/blog/mid-2025-ai-chatbot-scorecard/
Pew Research Center (Gottfried, J., et al.) (2026, 17 June). Americans and AI 2026: Chatbots, smart devices and views on impact. https://www.pewresearch.org/internet/2026/06/17/americans-and-ai-2026-chatbots-smart-devices-and-views-on-impact/
WindowsForum (2025, 7 November). Enterprise AI 2025: ChatGPT dominates, Copilot gains ground. https://windowsforum.com/news/enterprise-ai-2025-chatgpt-dominates-copilot-gains-ground.388410/
