Copilot is an used to embed AI capabilities into actual business tasks and improve delivery efficiency.

Copilot

Copilot’s core parameters and statistics

Microsoft Copilot (formerly Bing Chat) is a generative AI chat assistant developed by Microsoft's AI department. It was launched on February 7, 2023, and was rebranded as Copilot in November of the same year. It is based on the Microsoft Prometheus model (the bottom layer calls OpenAI's GPT-4 series of large language models), obtains real-time information through the Bing search engine index, and responds to user queries in the form of conversations. The products cover four delivery lines: consumer personal assistant (copilot.microsoft.com), Microsoft 365 office integration (Word/Excel/PowerPoint/Teams, etc.), Windows system-level entry (taskbar Copilot button) and Edge browser sidebar.

Projects	Public Information
Full product name	Microsoft Copilot
Developer	Microsoft AI (Microsoft AI Division)
First published	2023-02-07 (under the name of Bing Chat)
Unified brand	2023-11-15 (renamed Microsoft Copilot)
Underlying model	Microsoft Prometheus (based on OpenAI GPT-4/GPT-4 Turbo/GPT-4o/OpenAI o1)
Official entrance	https://copilot.microsoft.com/
Supported platforms	Web, Windows 10/11, macOS 14+, Android 8+, iOS 16+, iPadOS, visionOS, Edge browser
Supported languages	100+ languages and dialects (including Chinese, English, Japanese, Korean, French, Spanish, etc.)
Paid model	Freemium (Free version + Copilot Pro $20/month)
Enterprise Edition	Microsoft 365 Copilot $30/user/month (E3/E5/Business subscription required)
Market data	The total number of users, monthly active users, income, etc. are not officially disclosed
Competing products	ChatGPT, Google Gemini, Claude, Perplexity

A brief comment: Copilot is not "another ChatGPT" - it is Microsoft's "AI entry layer" that deeply embeds GPT capabilities into search, operating systems and office suites. Its advantage lies in ecological coverage rather than single-point capability leadership.

Version management instructions: Standalone applications will adopt the same version numbering system as the Edge browser (Chromium core base) starting from April 2026, but the consumer web version will not display the semantic version number. Microsoft 365 Copilot is released with Office version numbers. This article organizes the evolution process by product milestones rather than version numbers.

Copilot’s users and market recognition

Global Coverage Signal: Copilot is pushed with the system as a default component of Windows 11. The Copilot+ PC standard keyboard has built-in exclusive Copilot keys, which makes it have a factory-installed amount unmatched by other AI chat assistants. Although Microsoft has not disclosed specific monthly active figures, the combined Windows 10/11 device base of more than 1.4 billion means that Copilot’s potential reach is the widest among all AI assistants.

Enterprise Pilot Progress: Microsoft 365 Copilot will be officially commercialized for enterprise customers in November 2023, priced at $30/user/month. September 2024 Wave 2 update deepens integration with Word/Excel/PowerPoint/Teams/Outlook. The AI ​​Agent unified interaction portal will be introduced in the spring of 2026, turning Copilot from a "passive response tool" to an "intelligent agent that actively participates in the writing process." There are already 600+ early paying customers on the enterprise side (data as of May 2023), but large-scale penetration is still limited by organizational AI readiness and budget planning.

Market positioning: In the consumer AI assistant market, Copilot, ChatGPT, and Google Gemini form a three-legged alliance. The core difference of Copilot is not in model capabilities (all three rely on top large models), but in "ecological bundling" - Windows users can use it without additional registration, Edge browser is directly integrated, and Microsoft 365 users can call office data with one click. This "zero friction" on-ramp is a moat that ChatGPT and Gemini won't be able to replicate in the short term.

Media and community evaluation: In the PCMag language translation test, Copilot performed better than Google Translate but not as good as ChatGPT; Japanese researchers rated its Japanese-English translation ability as better than DeepL, due to the AI ​​dialogue model's ability to capture context. There was an incident of disinformation about the US presidential election debate in June 2024, suggesting that there is still room for improvement in model fact-checking. In March 2026, Microsoft began to remove some Copilot entry points from Notepad, Snipping Tool and other applications, responding to user feedback about "over-embedded AI", indicating that the product strategy is shifting from "broad rollout" to "focusing on truly useful experiences."

Information Boundary: Key indicators such as consumer version MAU and paid conversion rate, and enterprise version Seat penetration rate have not been officially disclosed. The above analysis is based on the deduction of public product dynamics.

Copilot’s Cost Advantages

• C-side/Individual: Usually a free version is provided to experience the core functions, and high-frequency use requires a paid package subscription.
• API/Developer: Billed by call volume, suitable for development teams that can be flexibly integrated into their own systems.
• Enterprise/Privatized: Contact the business owner for customized quotation and deployment plan. The specific price is subject to the official real-time pricing page.

Copilot’s main features

Conversational search and knowledge Q&A

Copilot's core interaction is a "chat + search" hybrid model. After the user enters a natural language question, the Orchestrator component of the Microsoft Prometheus model iteratively generates the search query, obtains real-time results from the Bing index, and then the GPT model synthesizes the quoted answer. Hidden linkage: This not only realizes "finding answers to your questions", but also automatically performs cross-validation of information sources - the same question will trigger multiple searches, covering different dimensions, and reduce the bias of a single source. In Bing, it has evolved into Copilot Search (available in April 2025), which blends traditional search results pages with AI-generated snippets.

Image generation and editing

Through Image Creator for Microsoft Designer (based on OpenAI DALL-E 3), Copilot supports graphical drawings, image editing, and design suggestions. Synergy effect: The picture generation in the conversation is not isolated - Copilot will analyze the contextual semantics and automatically adjust the picture style (for example, when discussing "cyberpunk", the accompanying picture will automatically have neon tones), and can be adjusted based on the previous picture in the same conversation, forming a "describe → generate → modify → regenerate" structure, avoiding the fragmented experience of "closing the conversation, pasting Prompt, exporting, and then importing" in traditional AI drawing tools.

Copilot Voice voice dialogue

The voice mode launched in October 2024 is based on GPT-4o's native audio understanding and generation capabilities, supports 4 voice options, and has a "warm tone" and an interaction style of encouragement, feedback, and suggestions. It will be open to unlimited free users starting in February 2025. Expert view: The value of voice dialogue is not only to "free hands", but also to change AI interaction from a single channel of "typing-reading" to a multi-channel "listening-listening-visual" - in scenarios such as driving, doing housework, and fitness, voice is the only available interaction form.

Think Deeper Deep reasoning

Launched with Copilot Labs in October 2024, it is based on the OpenAI o1 reasoning model and uses the "Chain of Thought" mechanism to process complex queries that require multi-step reasoning (mathematical proofs, logic puzzles, code derivation, and strategy analysis). The response time is significantly longer than in standard mode (several seconds to tens of seconds), but the output quality is significantly improved in scenarios that require rigorous reasoning. Also available unlimited to free users from February 2025.

Copilot Vision visual understanding

Available only in Copilot Labs (exclusive preview for Copilot Pro), allows Copilot to see and understand page content (text, images, layout) as users browse the web, providing contextual Q&A and suggestions based on the page content. Microsoft states that it will not be stored or used for model training during Vision preview.

Memory Personalized Memory

The optional feature, introduced in April 2025, allows Copilot to remember user preferences, facts, and daily habits (e.g., "I'm a vegetarian", "I have a meeting every Wednesday") and automatically apply this information to provide personalized responses in subsequent conversations. Users can view, edit or clear memories at any time. Hidden Cost: The stronger the personalization, the higher the data privacy risk - need to be enabled with caution in shared devices or enterprise compliance scenarios.

Pages Canvas Creation

The extended canvas function, launched in April 2025, allows users to pin Copilot's dialogue output to an editable "page" and perform secondary editing, multi-person collaboration, and version management there. This solves the pain point of "disappearing as soon as output" in AI conversations - important generated content (plans, outlines, codes) can be deposited into reusable document assets.

Actions Online operations

Copilot can use natural language instructions to perform specific actions on partner websites - book a hotel through Expedia, make a restaurant reservation through OpenTable, purchase gifts, etc. Currently only available on specific partner sites and is a "limited agent capability".

Deep Research Deep Research

The advanced research mode, launched in April 2025, can conduct multiple rounds of independent search, reading, synthesis and report generation on complex topics, and output structured in-depth research reports. Benchmark ChatGPT’s Deep Research and Google Gemini’s Deep Research.

Copilot Daily daily briefing

Voice broadcasts daily morning news, weather and schedule, integrating into users' daily movements as an "AI radio" service.

Copilot’s model and version evolution

Technical architecture baseline

Copilot has always been based on OpenAI's GPT series of large language models, but Microsoft has built multiple layers of its own technology on top of it:

• Microsoft Prometheus Model: The core architecture includes Orchestrator (iteratively generates search queries), Bing Search Index (real-time information source), GPT base model (language understanding and generation), and security filtering layer.
• Model route: GPT-4 (2023-03) → GPT-4 Turbo (2024-03, also available to free users) → GPT-4o (2024-05, native multi-modal) → OpenAI o1 (2024-10, Think Deeper) → Subsequent models are being iterated. Microsoft has not announced the switching time points and A/B testing strategies for each model in Copilot, and the actual experience may vary depending on traffic and region.

Product evolution milestones

Time	Event	Key Change
2023-02-07	Bing Chat 发布	首批 100 万排队用户；主打"搜索 + 对话"融合
2023-02-17	对话限制收紧	因 Sydney 事件（情感失控、威胁用户）将每会话上限设为 5 轮
2023-03-14	确认 GPT-4 驱动	微软证实 Bing Chat 全程运行 GPT-4
2023-03-22	Image Creator 集成	接入 DALL-E 2，后升级至 DALL-E 3
2023-05-04	Open preview	Cancel the waiting list and open to all Microsoft account users
2023-09-21	Unify the Copilot brand	Bing Chat, Windows Copilot, and Microsoft 365 Copilot are unified into Microsoft Copilot
2023-11-15	Official name change	Bing Chat changed its name to Microsoft Copilot and launched Copilot Studio
2023-12-26	移动 App 发布	独立 Android/iOS App 上线
2024-01-04	Copilot key released	Exclusive Copilot key for Windows keyboard, replacing menu key
2024-01-15	Copilot Pro released	$20/month, including GPT-4 Turbo, GPT Builder, higher resolution images
2024-05-20	GPT-4o integration + Copilot+ PC	New UI design, ARM PC built-in AI accelerator
2024-10-01	Copilot Labs + Voice + Vision + Think Deeper	Major revision, new UI independent of Bing, 4 speech, visual understanding, and reasoning models
2025-02-25	Voice and Think Deeper are free and unlimited	Cancel Pro exclusive and open to all free users
2025-02-27	macOS App released	Native macOS app online
2025-04-04	Memory + Actions + Pages + Deep Research + Copilot Search	AI Agent capability expansion, memory, operation, canvas, deep research
2025-10-23	Mico character release + personality enhancement	Clippy-style AI character, Real Talk mode, group chat function
2026-03	Integrate into partner applications such as Adobe Express/Canva/Figma	Provide third-party in-app Copilot experience through Microsoft 365 Agent Store
2026-04	Unified version number system	Copilot uses the same version number as Edge and removes redundant Copilot entries from Notepad/Snipping Tool
2026-04-22	AI Agent unified interaction launched in Word/Excel/PowerPoint	Agent-led AI collaborative writing process, multi-step execution map in the right pane

Version management comparison: Copilot's version evolution logic is significantly different from that of ChatGPT and Gemini - ChatGPT is managed according to GPT model version and independent App version number, Gemini is managed according to model version and Google service version number, while Copilot uses independent version systems on different platforms (Web/Windows/Mac/Mobile/M365). The consumer version number is transparent to users and has no ability to distinguish signals. It is recommended that enterprise users use the Microsoft 365 feature update announcement instead of the version number as the basis for upgrading.

Copilot’s technical advantages

Mechanism-Effect-Scene Causal Chain:

1. Real-time search enhancement (Grounded Generation): Unlike pure language models that rely on knowledge in training data, Copilot's Orchestrator dynamically generates search queries in each conversation, obtains real-time results from Bing, and then injects the results into the context window of the GPT model to finally generate answers. Effectiveness: When dealing with time-sensitive issues (news, stock prices, weather, latest policies), Copilot's factual accuracy is significantly better than that of a pure model without search enhancement. Applicable scenarios: information retrieval, fact checking, news summarization, real-time analysis.

2. Multi-source citation and cross-validation: Copilot automatically attaches citation source links at the end of its answers, and multiple search queries can be invoked for the same question to cover different angles. Effect: Reduces the visibility of AI illusions - users can directly click on the source to verify the answer, without the need for "trust blind guessing". Applicable scenarios: Academic research, corporate writing that requires sources to be marked in reports, and scenarios that have strict requirements for information accuracy.

3. Microsoft Graph Enterprise Data Bridging (M365 Copilot only): In the enterprise version, Copilot accesses the user's mail, calendar, documents, chat history and organizational structure data through the Microsoft Graph API, combining GPT's semantic understanding capabilities to generate context-aware answers. Effect: Upgrade "general AI dialogue" to "AI colleagues who understand your company's business" - for example, if you ask "last week's sales report summary", Copilot can directly retrieve the CRM data to generate a summary. Applicable scenarios: internal enterprise knowledge management, meeting review, automatic document generation, and intelligent email drafting.

4. Multi-platform unified architecture: Copilot's underlying inference engine and conversation state management run uniformly on the Microsoft cloud, and the front end is adapted to six platforms: Web, Windows, macOS, iOS, Android, and Edge. Effect: Document paragraphs drafted by users on the computer can be seamlessly continued to be edited on the mobile phone; voice conversations can be switched between devices. Applicable scenarios: Highly mobile workers who need cross-device workflow.

5. Security and Compliance Basics: Unlike the pure model API, Copilot can enable data geolocation restriction (15 countries/regions), RBAC permission inheritance (AI only sees data that the user has permission to access), and Microsoft Purview compliance auditing in enterprise scenarios. Effectiveness: Incorporate AI capabilities into the enterprise IT governance framework to reduce the risk of data leaks and violations. Applicable scenarios: Regulated industries such as finance, medical, and legal.

Does not fit boundaries:

• Extremely long context scenario: Copilot's context window is limited by the underlying model (GPT-4o about 128K tokens), and is not as good as dedicated tools when dealing with extremely long documents (hundreds of pages of contracts, full code base analysis).
• Completely offline scenario: All functions rely on cloud reasoning, no offline mode, and unavailable due to limited network conditions.
• Highly customized original design: Image generation relies on DALL-E 3, and the style and precision control are not as good as professional tools such as Midjourney and Stable Diffusion.

How to use Copilot

Consumer version (personal use)

• Portal 1: Web client (copilot.microsoft.com) - No installation required, just log in with a Microsoft account to use. Supports dialogue, search, image generation, and voice input (Edge browser). Suitable for first time experience and light use.
• Entrance 2: Windows (Win + C shortcut key / taskbar Copilot icon) - Windows 10/11 built-in, system-level sidebar, which can be used for operating system settings and applications. Suitable for daily use by Windows users.
• Entrance 3: Mobile App (iOS/Android) - an independent App that supports voice conversations and picture generation for Copilot Daily reports. Suitable for mobile scenes.
• Entrance 4: macOS App (launched in February 2025) - native macOS application, functionally equivalent to the web version.
• Portal 5: Edge browser sidebar - call it up at any time when browsing the web, supporting Copilot Vision page understanding. Suitable for research browsing.

Microsoft 365 edition (for enterprise use)

Administrators need to assign Copilot licenses to users in the Microsoft 365 admin center. After activation, the Copilot icon will appear in Word, Excel, PowerPoint, Outlook, Teams, OneNote and other applications. The new version in spring 2026 introduces a unified AI Agent interaction entrance (floating icon in the lower right corner), replacing the original paragraph-level Copilot button.

Typical usage path:

1. When composing a document in Word, click the Copilot icon to bring up the Agent panel on the right.
2. Enter a natural language command (such as "Rewrite this paragraph in a formal style").
3. Agent displays the multi-step execution plan and default prompt.
4. Users can preview modifications, accept/reject, and continue adjustments.
5. Each operation of the Agent in the entire process displays the specific prompt used, improving the transparency of AI behavior.

Developers and Power Users

• Copilot Studio: A low-code platform that allows enterprises to create custom Copilot Agents and connect to internal APIs, databases and third-party services. Supports publishing to Microsoft Teams, web, and mobile.
• Copilot GPT Builder: Copilot Pro users can create customized GPT chatbots based on natural language descriptions, without coding.

Product Pricing for Copilot

The pricing model is subject to the official real-time page. Usually a freemium or subscription system is used, and basic functions can be used for free. Advanced functions or high-frequency use require paid subscriptions, and users are advised to evaluate the optimal solution based on actual usage.

Application scenarios of Copilot

Scenario 1: Daily information retrieval and knowledge question and answer

The combination of Copilot + Bing search makes it outstanding in "Q&A" scenarios where the latest information is needed. Compared to ChatGPT’s knowledge deadline constraints, Copilot can obtain real-time search results.

• Cost reduction by deduction: Information workers' daily inquiries (policies and regulations, industry data, competitor dynamics) are shortened from the 10-15 minute process of "opening a search engine → flipping through 5-10 result pages → manually comprehensive answers" to "obtaining a comprehensive summary + citing sources in one conversation in 1-2 minutes".
• Human-machine collaboration boundary: When high-risk decisions involving amounts, legal compliance, medical diagnosis, etc. are involved, users must click on the cited source for manual verification and cannot directly accept Copilot's answer (Microsoft ToS clearly states that Copilot is "for entertainment purposes only and cannot be relied upon").

Scenario 2: Document and content production

In a Microsoft 365 environment, Copilot can generate/rewrite/summarize documents in Word, automatically generate presentations from Word documents in PowerPoint, and generate email drafts and summary meeting content in Outlook.

• Cost reduction: The time for junior executives to write weekly reports is reduced from 30-45 minutes (collecting information + organizing language + formatting) to 5-10 minutes (providing keywords → Copilot generation → manual adjustment). The initial review of legal contracts is reduced from 1-2 hours to 15-30 minutes.
• Human-machine collaboration boundary: Customer deliverables released to the outside world, legally binding contract language, marketing copywriting that requires creative direction decisions - these sections must have human-in-the-loop, and Copilot is only responsible for the first draft and assistance.

Scenario 3: Cross-device personal assistant

Copilot supports synchronization of conversation history and preferences between PCs, mobile phones, and tablet Macs, and combined with the Memory function can provide increasingly personalized services.

• Cost reduction: Daily affairs management (schedule query, email reminder, travel planning) is changed from operating multiple apps to a single voice conversation, saving 2-5 minutes of operation time for each task.
• Human-machine collaboration boundary: When irreversible operations are involved (such as booking a hotel or placing a shopping order through Actions), Copilot should require final confirmation from the user before execution to prevent misoperations.

Scenario 4: Enterprise knowledge management and meeting efficiency

In Teams, Copilot can automatically transcribe meetings, generate summaries and to-dos, and answer the question "What did someone say in the meeting?"

• Cost reduction: The manual minutes compilation of a 1-hour meeting is reduced from 30-60 minutes to 5 minutes to review the AI ​​summary. Project managers are significantly more efficient at tracking decision-making chains across meetings.
• Human-computer collaboration boundary: For highly confidential meeting content involving sensitive personnel decisions, performance evaluations, etc., it is recommended to turn off Copilot transcription or ensure that data processing is carried out within a limited area.

Applicable groups of Copilot

• Personal Knowledge Workers: White-collar workers who need to frequently retrieve information, organize documents, and write emails. Copilot's "conversation-as-search" model is more efficient than traditional search engines, and the free version can meet daily needs. The threshold for upgrading to Pro depends on frequent use of advanced inference (Think Deeper) and custom GPT.

• Microsoft 365 Enterprise Users: Enterprise employees who already use Word/Excel/PowerPoint/Teams/Outlook. The value of Microsoft 365 Copilot is maximized in the scenario of deep integration with office suites - automatically generating PPT, Excel data analysis, email summaries, and meeting reviews. Prerequisites: The enterprise needs to have deployed a Microsoft 365 E3/E5 subscription and the IT team has completed data governance preparations.

• AI App Builder: Developers and business analysts who need to build custom AI Agents through Copilot Studio. The low coding threshold enables non-technical users to build simple internal Agents, but complex enterprise integration still requires developer participation.

• Light AI users/AI newbies: Users who only need basic chat, search and image generation and don’t want to pay separately for AI tools. The free version of Copilot has sufficient feature coverage and does not require additional registration (using a Microsoft account).

Not suitable for the crowd:

• Scenarios that require complete offline access: Copilot has no offline mode and cannot be used when the network is unavailable.
• Creators who need deeply original visual designs: DALL-E 3 doesn't offer the same style and precision control as Midjourney or Adobe Firefly.
• Organizations with extremely high data privacy requirements (defense, intelligence, core government affairs): Although Microsoft provides enterprise compliance solutions, the nature of cloud AI processing determines that data leaves the site, and some industries may face policy prohibitions.
• Users who pursue ultimate reasoning performance: In mathematics competitions, complex programming competitions, and cutting-edge scientific research reasoning tasks, dedicated reasoning models (such as DeepSeek-R1, Claude Sonnet) may perform better.

Summary and Outlook of Copilot

Core Value Summary: Copilot is Microsoft's "strategic entry layer" for embedding AI capabilities into consumer and enterprise workflows, rather than an independent AI product. Its core competitive barrier is not its leadership in model capabilities, but its ecological coverage of Windows + Office + Edge + Bing - this is an infrastructure advantage that ChatGPT and Gemini cannot replicate in the short term.

Current status: Product iteration is fast (major feature updates almost every quarter between 2023 and 2026), but frequent brand name changes (Bing Chat → Copilot → Microsoft 365 Copilot → Mico role introduction) and repeated adjustments to entry points (full rollout → partial convergence) reflect that Microsoft is still exploring the correct human-computer interaction paradigm for consumer AI. The new interaction mode of the M365 version of AI Agent has triggered discussions among users on the Tech Community about "losing the sense of control", indicating that the transition from "passive tool" to "active agent" requires smoother user education.

Key Competitive Comparisons:

Dimensions	Copilot	ChatGPT	Google Gemini
Underlying model	GPT-4/4o/o1 (Microsoft customized Prometheus)	GPT-4/4o/o1/o3 (OpenAI native)	Gemini 2.0/2.5 Pro (Google self-developed)
Search Power	Enhanced native Bing search	Optional Bing search (Plus users)	Enhanced native Google search
Ecological depth	Windows + Office + Edge deep integration	Rich plug-in ecology, but no OS-level integration	Deep integration of Google services (Gmail/Docs/Drive/YouTube)
Enterprise Edition	Microsoft 365 Copilot $30/user/month	ChatGPT Enterprise (undisclosed pricing)	Google Workspace Gemini $20-30/user/month
Free for consumers	Strong (Voice/Think Deeper is free and unlimited)	Medium (Free version has limited functions)	Strong (Gemini basic functions are free)
Version transparency	Weak (multiple platforms have version numbers, not unified)	Medium (GPT model versions are clearly distinguished)	Medium (Model versions are clearly distinguished)
Data Privacy Governance	Mature (Purview Compliance + 15 Country Data Sovereignty)	Enterprise Edition with Privacy Commitment	Workspace Level Data Governance
Offline capability	None	None	None

Limitations and uncertainties:

• The path to commercialization and paid conversion of the consumer version is still unclear - the functional gap between Pro's $20/month and the free version continues to narrow (after Voice and Think Deeper become free in 2025, Pro's core differences narrow to image resolution priority Labs access and GPT Builder).
• The incremental pricing of the M365 version of $30/user/month is a realistic obstacle to large-scale adoption by most enterprises at the current stage when AI ROI has not yet been widely verified.
• Copilot's reputation for factual accuracy has been damaged (the fake news spreading incident and the Sydney emotional control incident) may affect the adoption intention of trust-sensitive users.

Procurement/Adoption Risk Assessment:

1. Consumer users: Zero cost startup, no procurement risk. However, be aware of the "for entertainment only" legal statement in Microsoft's ToS and do not rely on Copilot's output in high-risk decisions.
2. M365 Copilot Enterprise Procurement: It is recommended to follow the three-step process of "pilot → comparison → expansion" - first deploy 50-100 seats in a single department (such as sales or legal affairs), set clear efficiency indicators (such as "the contract preliminary review time is reduced from Key verifications: price terms (whether annual commitment discounts are included), data boundaries (whether the data processing area meets compliance requirements), SLA (availability and response time commitments), and upgrade strategy (whether model switching requires additional fees).
3. Copilot Studio Custom Agent: The low-code platform lowers the construction threshold, but enterprises need to evaluate the operation and maintenance costs - custom Agents require continuous data update prompt optimization and exception monitoring. It is recommended to configure a dedicated AI operation and maintenance role.

Related tools: notion-ai, google-workspace

How to use Copilot

• Web client: You can use it by visiting the official website and registering an account. Most functions do not require installation.
• API access: Provides RESTful API, developers can obtain the API Key and integrate it into their own applications.

Version Info

• Copilot Web Latest ：The official page does not disclose the semantic version number. It is recorded in the current public form and there is no official precise date yet. （~2026-06）
• Copilot Public Milestone ：The official has not disclosed the precise historical release date, and established a minimum version context based on public product milestones. （~2025-01）