Skip to content

Compare AIs

Four tools, one set of moves

The overview that sits above the four deep-dive guides. Learn the language of one platform, then read across to find what the same thing is called in the others. Then see clearly which one to reach for and when.

Background image generated with Google Nano Banana Pro.

Start here

One line on each

Before the detail, the single sentence that captures what each platform is built around. The vocabulary differs, but understanding the core philosophy tells you why each one names things the way it does.

Part 1

The terminology translator

All four platforms have converged on the same core concepts. Each calls them something different. If you are fluent in one and lost in another, this is your Rosetta Stone. Read across each row to find the equivalent term. Where a mapping is loose rather than exact, there is a note.

Core, everyone hits these
The conceptChatGPTClaudeCopilotGemini
The AI brain
The model doing the thinking
GPT-5.5Claude Opus / Sonnet / HaikuGPT-5.5 or Claude OpusGemini 3.5 Flash / 3.1 Pro
Standing instructions
How you want it to always behave
Custom InstructionsCustom InstructionsCustom InstructionsInstructions for Gemini

NoteGemini's version lives inside Personal Intelligence and works on personal Google accounts with an AI plan only. It is not available on work or school Workspace accounts. For Workspace users, the equivalent is set inside the Gemini side panel in Google Docs.

Long-term memory
Remembers you between chats
MemoryMemoryMemoryPersonal Intelligence

NoteGemini splits this. Personal Intelligence pulls live context from your Gmail, Drive and Calendar, which is broader than the others' Memory. The closest pure-memory equivalent sits inside that feature.

Organised workspace
A home for one project's chats and files
ProjectProjectCopilot NotebookNotebookLM / Gem

NoteClosest match for project continuity is NotebookLM. But Gemini has no single perfect equivalent: a Gem is closer to a Custom GPT, while NotebookLM is grounded only on sources you add. See the agentic tier for custom assistants.

Custom AI assistant
A reusable AI built for one job
Custom GPTSkillCopilot AgentGem

NoteClose cousins, not identical. A Custom GPT and a Gem package instructions plus knowledge into a reusable persona. A Claude Skill is more procedure-shaped, a repeatable capability. A Copilot Agent can take autonomous actions across your org. Same family, different centre of gravity.

Live editable output
A document the AI builds with you
CanvasArtifactCopilot PagesCanvas
Deep research
Multi-source cited report
Deep ResearchResearchResearcher (agent)Deep Research
Live web access
Pulls current information
Web SearchWeb SearchWork IQ (web grounding)Google Search grounding
Voice conversation
Talk to it out loud
Voice ModeNot built-inCopilot VoiceGemini Live
Image generation
Make a picture from a prompt
Images 2.0Not built-inDesigner / CopilotNano Banana

NoteClaude does not generate images natively. It can build image-producing tools and design assets via Claude Design and Artifacts, but for direct text-to-image, the other three lead.

Advanced, agentic and power-user
The conceptChatGPTClaudeCopilotGemini
Step-by-step reasoning
Shows its working on hard problems
ThinkingExtended ThinkingGPT-5.5 ThinkingDeep Think
Connect to other apps
Link to external tools and data
Apps / ConnectorsMCP ConnectorsConnectors / GraphExtensions

NoteClaude's MCP (Model Context Protocol) has become the open industry standard. Several other platforms now support MCP connections too, so this term increasingly travels across all four.

Long-running autonomous work
Delegate a multi-step task that unfolds over time
Agent ModeCowork / Computer UseCopilot CoworkGemini Spark

NoteThis is the fastest-moving category in 2026. Gemini Spark is still rolling out (US beta, Workspace preview this summer). Copilot Cowork is Frontier-programme only. Maturity varies a lot, even though the concept is shared.

Build agents without code
Platform for making your own AI agents
GPT BuilderSkill / Agent SDKCopilot StudioWorkspace Studio
Scheduled tasks
Set work to run automatically later
Create TaskVia CoworkScheduled (in agents)Scheduled Actions
Coding agent
AI that writes and runs code
CodexClaude CodeGitHub CopilotAntigravity / Code Assist
Grounded knowledge base
AI answers only from sources you give it
Project filesProject knowledgeCopilot NotebookNotebookLM
Team / shared workspace
Colleagues work from the same context
Shared ProjectsTeam ProjectsShared Notebooks / PagesShared Gems / Drive
Meeting intelligence
Transcribe and summarise meetings
Record ModeNot built-inCopilot in TeamsGemini in Meet
Part 2

Capability snapshot

Not a scorecard with an overall winner, because there is no honest overall winner. Each platform leads somewhere. This shows relative strength by job, so you can match the tool to the task. More filled dots means stronger for that specific job, as of June 2026.

The jobChatGPTClaudeCopilotGemini
Deep reasoning and analysis
Hard, nuanced thinking
Leads
Leads
Long-form writing
Quality and voice control
Leads
Research with citations
Multi-source synthesis
Leads
Data and spreadsheets
Analysis in your files
Leads
Image and video
Visual content creation
Leads
Build your own AI
Custom assistants and tools
Leads
Agentic / autonomous work
Delegating multi-step tasks
Leads
Meetings
Transcribe and summarise
Leads
Works in your existing apps
Lives where you work
Leads
Leads
How to read this fairly

Dots reflect relative strength for a specific job, not overall quality. All four are excellent general assistants. The differences show up at the edges, and the edges are where the right choice gets made. Capability also moves monthly. Treat this as a June 2026 snapshot, not a permanent ranking. The strategic move is rarely pick one and abandon the rest. It is knowing which to reach for, for which job.

Part 3

Which should I actually use?

The honest answer usually starts with what you already pay for and where your data lives. Most people in a business already have one of these by default. Here is how a strategist would decide, by scenario.

Your organisation runs on Microsoft 365
Default to Copilot

You are paying for it inside your licence, and its Work IQ context advantage is real. It already knows your emails, files and meetings. Get fluent here first. Reach for ChatGPT or Claude when you need stronger reasoning or want the newest capabilities before Microsoft ships them.

Your team lives in Google Workspace
Default to Gemini

It is bundled into your Workspace plan and lives inside Gmail, Docs and Drive. Workspace Intelligence plus live Google Search is a real edge. Watch the agentic features arriving this summer. Pair with Claude when a task needs the deepest reasoning.

You want the best thinking and writing
Reach for Claude

For nuanced analysis, high-quality long-form writing, and building your own tools through Skills and MCP, Claude leads. It is also the strongest for serious agentic work via Cowork and Claude Code. The trade-off: no native image generation, no meeting transcription.

You want range and the newest features first
Reach for ChatGPT

The broadest toolbox and the largest third-party ecosystem. New capabilities tend to land here first, and it has a strong answer for nearly every job. A safe, capable default if you are not locked into Microsoft or Google. Excellent all-rounder.

You need image or video generation
Gemini or ChatGPT

Gemini's Nano Banana and Omni video lead in 2026, with ChatGPT's Images 2.0 a strong rival. Copilot handles basic visuals through Designer. Claude does not generate images directly, though it can build design assets and image-producing tools.

You need meeting intelligence
Copilot or Gemini

Only Copilot (in Teams) and Gemini (in Meet) transcribe and summarise meetings natively. If meetings are central to your workflow, your platform choice may simply follow your video tool: Teams points to Copilot, Google Meet points to Gemini.

The strategist's view

The biggest mistake is treating this as a single winner-take-all choice. The top 1% use the right tool for the job. Most professionals end up with a default driven by their employer's stack, plus one deliberately chosen second tool for the work their default does poorly. Master your default first. Add a second for its edge. Ignore the rest until a real job demands them. That is how you get leverage without drowning in tools.

Turn tool choice into team capability

Knowing the platforms is step one. Justin Kabbani trains businesses, leaders and teams to put them to work, so AI saves real hours and unlocks what your people are truly capable of.

Work with Justin