The short answer: ChatGPT Canvas is a side-by-side AI editing workspace built into chat, not a separate app. It opens a document or code panel next to your conversation, lets you highlight text and ask for targeted edits, and ships with 10 quick-action shortcuts for writing and coding. The catch in 2026: it works with legacy models only (GPT‑4o, GPT‑4.1, and similar), and OpenAI is actively sunsetting it on newer reasoning models like GPT‑5.5 Instant, GPT‑5.5 Thinking, and GPT‑5.6 Sol. So the window to learn it is now, while it still works.
OpenAI launched Canvas on October 3, 2024 with a custom-trained GPT‑4o that knows when to open the editor, when to make a targeted edit, and when to rewrite the whole thing (OpenAI, October 2024). By 2026, Canvas sits in every ChatGPT plan: Free, Plus, Pro, Team, Enterprise, and Edu (OpenAI Help Center). It’s available on Web, Windows, and macOS; mobile is still “coming soon” as of July 2026.
I’ve been using Canvas since the beta dropped. Below are the 10 tips that actually changed how I work, plus the ones OpenAI’s docs don’t tell you about, plus a plain-English warning about what changed in May 2026.
One quick heads-up: On May 28, 2026 OpenAI quietly removed Canvas from GPT‑5.5 Instant and GPT‑5.5 Thinking. Newer models now handle long-form writing and code through writing blocks and code blocks in the chat itself. Canvas still exists, but it’s running on legacy models. If you’re on Plus and want Canvas, you’ll need to switch to a legacy model in the picker (OpenAI Model Release Notes, May 28, 2026).
What Canvas Actually Is (and Isn’t)
Definition: Canvas is a split-pane editing surface that opens next to your chat thread. The chat stays on the left, the document or code lives on the right, and both update in real time.
It’s not a separate app. It’s not Google Docs. It’s not VS Code. It’s ChatGPT with a built-in editor that:
- Opens automatically when ChatGPT thinks you want something longer than ~10 lines
- Opens on demand when you say “use canvas”, “open a canvas”, or “open a coding canvas” in any chat
- Opens from a slash command by typing
/canvasin the prompt box - Opens from a paste: paste any text or code into the composer and click the Open in canvas button that appears in the upper-right corner of the prompt
The way I describe it to clients: Canvas is what happens when ChatGPT gets tired of scrolling and you get tired of copy-pasting. The whole point is that edits happen in place, with both of you looking at the same thing at the same time.
What Canvas is NOT
- Not a multi-file IDE. It handles one file at a time.
- Not real-time multi-user collaboration. You can share a canvas (OpenAI added this in 2025), but you can’t co-edit live like in Google Docs.
- Not available on every model. As of July 2026, Canvas is not supported by GPT‑5.5, GPT‑5.5 Thinking, GPT‑5.6 Sol, or any “pro-series” model (OpenAI Help Center, “Model compatibility”).
- Not a place for advanced formatting. Only basic Markdown works inside a writing canvas: bold, italic, headers, bullets, numbered lists. No tables, no images, no code blocks inside a writing canvas.
The shortcut menu you’ll actually use
Hover over the bottom-right corner of any open canvas to reveal the shortcut menu. The exact list is published in OpenAI’s help article:
Writing shortcuts (OpenAI Help Center):
- Suggest edits: ChatGPT drops inline comments on what to fix
- Adjust the length: opens a slider from shortest to longest
- Change reading level: opens a slider from Kindergarten to Graduate School
- Add final polish: checks grammar, clarity, and consistency in one pass
- Add emojis: sprinkles emojis in or replaces keywords with them
Coding shortcuts:
- Add logs: drops in print statements to help you debug
- Add comments: writes inline comments explaining the code
- Fix bugs: detects and rewrites problematic sections
- Port to a language: translates your code into JavaScript, TypeScript, Python, Java, C++, or PHP
- Code review: drops inline suggestions for optimization and improvements
Four of these apply changes immediately when you click them: Add emojis, Add final polish, Fix bugs, and Port to a language. The others either open a slider (Adjust the length, Change reading level) or pop up an inline comment you have to click Apply on (Suggest edits, Code review).
“Canvas was built with GPT-4o and can be manually selected in the model picker while in beta.” - OpenAI, October 3, 2024 launch post
Canvas Tips at a Glance: Comparison Table
| # | Tip | What it does | How to access | Best for | Skip when |
|---|---|---|---|---|---|
| 1 | Use the slash command | Force Canvas to open instantly | Type /canvas in the prompt |
Cold-starting a doc on a fresh chat | You only need a one-line answer |
| 2 | The 10-line trigger | Let Canvas open itself automatically | Just write more than 10 lines | Drafts, blog posts, code files | Quick factual Q&A |
| 3 | Highlight + inline comment | Get targeted edits on one paragraph | Select text, click comment icon, type instruction | Rewriting one section without nuking the rest | Whole-doc rewrites |
| 4 | Suggest edits shortcut | Let ChatGPT critique your draft first | Click “Suggest edits” in shortcut menu | Polishing before publishing | You want a full rewrite |
| 5 | Reading level slider | Re-pitch for any audience | Click “Change reading level”, drag the slider | Adapting marketing copy for different segments | You already know your audience |
| 6 | Adjust the length slider | Compress or expand without rewriting | Click “Adjust the length”, drag the slider | Tightening verbose copy or padding short copy | You’ve already done the edit |
| 7 | Run Python in browser | Execute code without leaving the chat | Click the Execute button (Python files only) | Data analysis, math, ML demos | You need a full Python env |
| 8 | React/HTML preview | See rendered output in a sandbox | Canvas opens with a preview tab | Frontend code, D3 charts, web components | Server-side rendering |
| 9 | Version history + Show changes | Undo or diff any version | Back/forward arrows + Show changes button | Recovering lost edits, comparing drafts | You haven’t made many edits |
| 10 | Export + share | Save and send your canvas | Download menu (top-right) for export; share button in toolbar | Handing work to a client or teammate | You never leave ChatGPT anyway |
1. Force Canvas with the Slash Command
Definition: The /canvas slash command is a built-in prompt prefix that tells ChatGPT to open a blank canvas immediately, bypassing the model’s “should I open canvas?” guess.
If you’ve ever typed a prompt and waited while ChatGPT decided not to open Canvas, this is the fix. In any chat, type:
/canvas write a 600-word product launch announcement for a new espresso machineThe composer dropdown shows a Canvas option when you type /. Click it, add your prompt, hit enter. A blank canvas pops open on the right, and ChatGPT starts typing directly into it.
When to use it: every cold start. Anytime you’re starting a new doc and don’t want to gamble on the auto-trigger.
When NOT: quick factual questions. Forcing a Canvas for “what’s the capital of Wyoming?” wastes tokens and clutters the screen.
Edge case: if you have Canvas disabled in your workspace admin settings, the slash command does nothing. Enterprise admins control Canvas via Admin settings → Workspace → Capabilities (OpenAI Help Center).
2. The 10-Line Auto-Trigger Rule
Definition: The Canvas auto-trigger threshold is the point at which ChatGPT decides the response is too long for chat and would benefit from the editor. Per OpenAI’s docs, that threshold is around 10 lines.
OpenAI’s research team trained GPT-4o to open a canvas when the response would be longer than roughly 10 lines, or when the prompt suggested editing was coming. Their internal eval hit 83% correct triggers for writing and 94% for coding (OpenAI Canvas launch post, Oct 2024).
What that means in practice: prompts like “write me a tweet” rarely trigger Canvas. “Write me a press release” usually does. The model’s judgment is biased toward not triggering for power users who’d find the editor annoying on small tasks.
Pro tip: if you want Canvas to open reliably, say so explicitly. Phrases that consistently trigger it:
- “Write a 500-word essay about X”
- “Draft a Python function that does Y”
- “Use canvas to write…”
- “Open a coding canvas and…”
When to use it: anytime you want a long-form output and don’t want to remember the slash command.
When NOT: short Q&A, math problems, translation. Canvas adds visual overhead with no benefit.
Fallback: if Canvas doesn’t open and you wanted it to, type /canvas or paste your response in and click Open in canvas.
3. Highlight + Inline Comment for Surgical Edits
Definition: Inline comments are sticky note-style edit requests attached to a specific span of text in your canvas.
This is the single biggest productivity unlock in Canvas. Here’s how it works:
- Highlight any block of text in the canvas
- A small toolbar appears with formatting options (bold, italic) and a comment bubble icon
- Click the comment bubble
- Type what you want changed (“make this more conversational”, “cut this paragraph”, “rewrite for a CFO audience”)
- Hit enter
ChatGPT responds with an inline suggestion card. You can read it, edit the text yourself, or click Apply to let ChatGPT write the replacement.
The reason this is the most underrated Canvas feature: ChatGPT’s training explicitly targets edits when text is highlighted. Per OpenAI’s launch blog, “the model performs targeted edits when users explicitly select text through the interface, otherwise favoring rewrites.” In their internal eval, GPT-4o with canvas outperformed prompted GPT-4o by 18% on targeted edits (OpenAI, Oct 2024).
Example: you’re editing a 1,200-word blog post. You hate the intro paragraph. Highlight just the intro. Type “make this punchier, lead with a question.” ChatGPT rewrites only those two sentences. The rest of the post is untouched.
When to use it: anything where you want surgical changes without rewriting the whole canvas.
When NOT: when you actually want a full rewrite. Don’t highlight, just say in the chat: “rewrite this whole thing for a more technical audience.”
Edge case: the comment icon only appears on paragraph blocks (the “block comment” icon) or selected text. It does not work on individual words. To select an entire paragraph, click the small comment icon in the left margin next to that paragraph.
Fallback: if you don’t like the inline suggestion, just close the comment and edit the text yourself. Your edit becomes a new version in the version history.
4. Use Suggest Edits as a Free Copy Editor
Definition: The Suggest edits shortcut opens inline critique comments on the entire canvas, the way a copy editor would mark up a manuscript.
Click Suggest edits in the bottom-right shortcut menu. ChatGPT scans your draft and drops yellow comment bubbles next to sentences or paragraphs it thinks need work. Click each bubble to read the suggestion, edit manually, or click Apply.
This is different from “Add final polish” (Tip 8’s cousin). Suggest edits is granular; Add final polish is global.
When to use it: blog drafts, marketing copy, anything going near a customer.
When NOT: code. There’s a separate Code review shortcut for that.
Example workflow I run weekly:
- Paste a finished blog draft into a new canvas
- Click Suggest edits
- Open every comment, evaluate each suggestion
- Apply the good ones, ignore the bad ones, fix the rest myself
- Click Add final polish for grammar pass
Edge case: Suggest edits is conservative. It won’t restructure your argument. For that, ask in the chat: “restructure this to lead with the conclusion.”
Fallback: if Suggest edits produces nothing useful, your draft might be too short (under ~150 words). Pad it or use Add final polish instead.
5. Slide Reading Level Anywhere from K to Grad School
Definition: The reading level slider lets you re-pitch the same canvas for any audience, from a 5-year-old to a PhD.
Click Change reading level in the shortcut menu. A slider pops up with two ends: Kindergarten and Graduate School. Drag it. ChatGPT rewrites the entire canvas for that audience in one pass.
The slider is more powerful than people realize. It’s not just “simpler words.” ChatGPT actually restructures sentences, swaps examples, and changes register. Move the slider all the way left and a paragraph about mortgage refinancing becomes “grown-ups borrow money to buy a house, and then they pay it back slow.” Move it right and you get dense jargon and formal hedging.
When to use it:
- Adapting a single piece of content for multiple audiences (a landing page that needs both an exec summary and a detailed version)
- Checking your own writing for accidental jargon
- Teaching - show students the same idea at different reading levels side by side
When NOT: if you’ve already calibrated your voice, the slider can flatten it. After dragging, you may want to re-add personality in a second pass.
Edge case: the slider doesn’t always hit the target. ChatGPT will rewrite for “8th grade” sometimes feels like “10th grade.” If you need precision, name the audience in a follow-up: “now do this for a board of directors.”
Fallback: if the slider does too much, undo with the back arrow in the top toolbar, then edit surgically using inline comments (Tip 3).
6. Compress and Expand with the Length Slider
Definition: The length slider is a separate control that adjusts the canvas’s word count from “shortest” to “longest” with one click.
Click Adjust the length. A slider pops up. Drag left to compress, right to expand. ChatGPT rewrites the whole canvas at the new length while preserving your structure.
This is the fastest way to hit a word-count target. No more “make this 200 words” loops.
When to use it:
- You wrote 800 words, your CMS needs 600
- You wrote 200 words and the brief said 500 minimum
- You’re A/B testing a short version vs. a long version of the same message
When NOT: if the canvas has structural problems, length changes won’t fix them. A bloated intro stays bloated.
Edge case: the slider doesn’t show target word counts. You’ll have to eyeball it. After dragging, glance at the bottom-left character count in the canvas toolbar.
Fallback: if the rewrite loses key details, undo and try Tip 3 (highlight + inline comment) on just the parts you want changed.
7. Execute Python Right in the Browser
Definition: Canvas code execution is a sandboxed Python runtime that runs your code in the browser, with a console panel at the bottom and a one-click Fix bug button when something breaks.
Click the Execute button at the top of any Python canvas. The code runs in a sandboxed environment. Output appears in the console panel. If there’s an error, ChatGPT surfaces it and offers a Fix bug button that auto-corrects the issue and re-runs.
When to use it:
- Quick data analysis on a CSV you paste in
- Math, statistics, or ML demos
- Learning Python interactively
- Testing a function before you commit it
When NOT: anything that needs persistent state, external packages not in the sandbox, or network access. Per OpenAI, Python execution in data analysis cannot make external web requests (OpenAI Help Center).
Edge case: only Python works in the executor right now. JavaScript, HTML, and React render in a sandbox (see Tip 8), but they don’t run as standalone scripts the same way.
Fallback: if the sandbox is too restrictive, copy the code out and run it locally or in Jupyter.
“This feature is currently only available for Python code but we plan to extend access to other programming languages in the future.” - OpenAI Help Center
8. Render React, HTML, and D3 in a Live Sandbox
Definition: The React/HTML preview pane is a sandboxed browser that renders your frontend code as you type, with support for many npm packages.
Open a canvas with HTML, React, or D3 code, and Canvas splits the right pane into a code editor on top and a live preview on the bottom. Save your file (or wait a moment for auto-refresh) and the preview updates.
Per OpenAI: “React/HTML code is rendered in a sandbox environment, allowing you to view the output of the code. Many npm packages and JavaScript libraries will work.” But previews that need external packages may stall if Allow canvas code to access the network is off, which is the default for enterprise workspaces (OpenAI Help Center).
When to use it:
- Building React components and want instant visual feedback
- D3 charts and data viz
- HTML/CSS experiments
- Quick landing-page mockups
When NOT: server-rendered apps, anything needing a backend, anything needing WebSockets to a non-allowed domain.
Edge case: if your preview hangs on “Installing Packages…”, check three things:
- The network access toggle in your workspace admin settings
- Browser console for
Failed to fetch,Refused to connect, or CSP errors - Your firewall or proxy isn’t blocking the npm CDN
Fallback: if network access is locked down, install packages locally and paste the bundled JS into the canvas.
9. Version History + Show Changes Is Your Undo Stack on Steroids
Definition: Version history is a timeline of every edit to your canvas, accessible via the back and forward arrows in the top toolbar. Show changes is a diff view that highlights additions and deletions between any two versions.
Every time ChatGPT rewrites a section, every time you click Apply on a suggestion, every time you type manually - each becomes a new version. Use the back arrow in the top toolbar to step backward. Use the forward arrow to step forward. Click Show changes to see green-highlighted additions and red strikethroughs between any two versions.
When to use it:
- ChatGPT made a rewrite you hate and you want the old version back
- You want to compare two drafts before publishing
- You accidentally clicked Apply on a bad suggestion
- You want to show a client “here’s what changed between v1 and v2”
When NOT: as a backup. Version history lives inside the canvas session. If you close the canvas or the chat, the history may go with it (depending on settings). Export the file you care about.
Edge case: Show changes works for both writing and code canvases. For code, you’ll see traditional line-by-line diffs. For writing, you’ll see paragraph-level diffs.
Fallback: if you need permanent version control for serious projects, copy the canvas contents into GitHub, Notion, or Google Docs after each major revision.
10. Export Anywhere + Share With Anyone
Definition: Canvas export supports PDF, Markdown, and Word for documents; auto-detects file extensions for code. Canvas sharing lets you send a read-only link to another ChatGPT user.
Two distinct features that get confused:
Export (Download button, top-right corner):
- For writing canvases: PDF, Markdown (.md), or Word (.docx)
- For code canvases: the language is auto-detected and exported as
.py,.js,.sql, etc. with proper syntax highlighting preserved - This makes Canvas a legit drafting tool, not just a chat toy
Share (Share button in the canvas toolbar):
- Available on Free, Plus, Pro, Team, Enterprise, and Edu (OpenAI Help Center)
- Creates a link similar to a shared ChatGPT conversation
- The recipient sees the canvas as it currently stands, plus the chat history that produced it
When to use export: handing off finished work to a client, pasting into your CMS, saving a draft to your hard drive.
When to use share: collaborating asynchronously with a teammate, sending a sample to a prospect, getting feedback from a colleague.
When NOT: sharing code that contains secrets, API keys, or proprietary business logic. Canvas content is processed by OpenAI and may be subject to workspace data controls.
Edge case: export to PDF preserves Markdown formatting, but advanced formatting (tables, embedded images) won’t render in a writing canvas because Canvas doesn’t support them in the first place.
Fallback: if you need something Canvas can’t export (e.g., a Google Doc with tracked changes), copy-paste the contents and format manually.
The 2026 Catch: What’s Actually Changed
I have to be straight with you. The Canvas story in mid-2026 is different from the Canvas story in late 2024. Three things to know:
1. Canvas is gone from GPT-5.5 Instant and GPT-5.5 Thinking
The May 28, 2026 update to GPT-5.5 Instant removed Canvas from that model. OpenAI’s exact wording in the Model Release Notes:
“With this update, canvas will no longer be available in GPT-5.5 Instant or GPT-5.5 Thinking. Writing and coding functionality is now supported directly in chat responses through writing blocks and code blocks. Paid users can continue using canvas for a limited time through legacy models until those models are sunset.”
Translation: if you were using Canvas in 2025 on the default model, you’re now seeing writing blocks and code blocks in regular chat instead of a side panel. These are formatted blocks in the chat stream that look similar to Canvas output but don’t have the editing tools.
2. Canvas still works, but on legacy models only
As of July 2026, you can still use Canvas. You just need to switch the model picker to GPT-4o or another legacy model that supports it. The Canvas help article confirms: “Canvas is not supported by GPT-5.5 or later models” (OpenAI Help Center).
GPT-4o itself was retired from ChatGPT on February 13, 2026, along with GPT-4.1, GPT-4.1 mini, and OpenAI o4-mini (OpenAI Help Center, “Retiring GPT-4o”). It remains available in the API. The “legacy models” Canvas can still ride on are now an even narrower list.
3. “Pro-series” models never supported Canvas
OpenAI has always been explicit: “Canvas is not available with pro-series models.” That meant o1-pro back in 2024, and it carries forward to o3-pro and the Pro tiers of GPT-5.6 in 2026 (OpenAI Help Center).
What this means for you
If you’ve never used Canvas: the writing blocks and code blocks in chat are fine for short edits. For long documents, switch the model picker to a legacy option (where available) and use Canvas. If Canvas doesn’t appear at all, your workspace admin may have disabled it.
If you use Canvas daily: do not assume it’ll be there in 2027. OpenAI is clearly moving toward in-chat blocks. Make peace with that and start building workflows that work in both modes.
Real Use Cases From My Own Work
I write and code for a living. Here are the five workflows where Canvas has actually paid for itself.
1. Blog draft to publishable post
I draft in Canvas, run Suggest edits, apply maybe 60% of the comments, run Add final polish for grammar, hit Change reading level to confirm it lands for my target reader, and export as Markdown straight into my CMS. That whole loop takes about 12 minutes for a 1,200-word post. Before Canvas, the same loop took 40 minutes across three different tools.
2. Refactor a legacy function
I paste the old Python function into a canvas, click Code review, get inline suggestions, apply the ones I agree with, then click Add comments to make sure I understand what changed. If the suggestions are too aggressive, I undo back to the previous version with the back arrow.
3. Quick data analysis
I upload a CSV, ask ChatGPT to summarize it, get a Python canvas back, click Execute to actually run the analysis, see a traceback, click Fix bug, and watch ChatGPT correct itself and rerun. I used to copy-paste into Jupyter for this. Now it’s all in one tab.
4. Teaching a junior dev
I write code in a canvas with the Add comments shortcut turned on so every block has an explanation. Then I share the canvas link. The junior dev opens it, sees the code and the reasoning inline, and can run the Python directly.
5. Sprint planning notes
I open a blank canvas with /canvas, type my sprint goals in Markdown bullets, run Add emojis to flag priorities, run Suggest edits to tighten the language, then Download as PDF and post it to Slack. Done in 5 minutes.
When Canvas Falls Back to Standard Chat
Canvas doesn’t trigger for every prompt. You get regular chat instead when:
- The prompt is short. Under 10 lines of expected output, ChatGPT skips Canvas. Ask “summarize this paragraph” and you get a chat reply. Ask “write me a 5-page essay on this paragraph” and you get Canvas.
- The model doesn’t support Canvas. GPT-5.5 Instant, GPT-5.5 Thinking, GPT-5.6 Sol, o3-pro, and any pro-series model will respond in regular chat, sometimes using the new writing blocks and code blocks.
- Your workspace admin disabled Canvas. Enterprise/Edu admins can turn Canvas off entirely. You’ll see no slash command and no auto-trigger.
- You’re on mobile. Canvas is web, Windows, and macOS only as of July 2026. iOS, Android, and mobile web get standard chat (OpenAI Help Center).
- You’re in temporary chat. Temporary chats don’t have Canvas in some configurations.
In all of those cases, your prompts still work - you just lose the side panel and the shortcuts.
FAQ
Is ChatGPT Canvas free to use? Yes. Canvas is available on Free, Plus, Pro, Team, Enterprise, and Edu plans (OpenAI Help Center). The model you’re using determines whether Canvas actually opens; on Free you may be limited to a legacy model that supports Canvas.
What’s the difference between Canvas and Anthropic’s Artifacts? Both are side-by-side editors. Anthropic’s Artifacts launched in June 2024 as a panel inside Claude for rendering code, documents, and visualizations. OpenAI’s Canvas launched October 3, 2024 with explicit focus on inline editing, version history, and shortcut menus (OpenAI, Oct 2024). The biggest practical difference: Canvas has the 10 quick-action shortcuts and the targeted-edit comment system. Artifacts leans toward live rendering of code.
Does Canvas work with GPT-5? No. As of July 2026, Canvas is not supported by GPT-5.5, GPT-5.5 Thinking, GPT-5.6 Sol, or any later model. Use a legacy model in the picker to access Canvas (OpenAI Help Center).
Can I use Canvas on my phone? Not yet. Canvas is on Web, Windows, and macOS as of July 2026. iOS, Android, and mobile web are still “coming soon” (OpenAI Help Center).
Does Canvas save my work? Canvas content lives inside the chat thread. If you keep the chat, you keep the canvas. If you delete the chat, the canvas goes with it. To keep a permanent copy, use the Download button to export as PDF, Markdown, Word, or your code’s native file type.
Can I collaborate with someone else in real time inside a Canvas? No. Canvas supports sharing (send a read-only link to another ChatGPT user), but not real-time co-editing. For real-time collaboration, export the canvas and move to Google Docs, Notion, or VS Code Live Share.
What happened to GPT-4o with Canvas? GPT-4o was the original Canvas model. It was retired from ChatGPT on February 13, 2026 (OpenAI Help Center). It still exists in the OpenAI API. In ChatGPT, Canvas now runs on other legacy models while they last.
Sources
- OpenAI Help Center: What is the canvas feature in ChatGPT and how do I use it? (updated July 2026)
- OpenAI: Introducing canvas (October 3, 2024)
- OpenAI Help Center: Model Release Notes (May 28, 2026 entry on Canvas removal)
- OpenAI Help Center: Retiring GPT-4o and other ChatGPT models (February 13, 2026)
- OpenAI: Retiring GPT-4o, GPT-4.1, GPT-4.1 mini, and OpenAI o4-mini in ChatGPT (January 29, 2026)
- OpenAI Help Center: GPT-5.6 in ChatGPT (updated July 2026)
- OpenAI Help Center: Data analysis with ChatGPT (updated July 2026)
- OpenAI Help Center: Creating and editing GPTs (Canvas capability, updated July 2026)
- OpenAI Platform: Models - GPT-5.6 Sol, Terra, Luna (current API catalog, July 2026)
- OpenAI: GPT-5.6 - Frontier intelligence that scales with your ambition (July 9, 2026)
- OpenAI: GPT-5.6 is now the preferred model in Microsoft 365 Copilot (July 9, 2026)
- OpenAI: ChatGPT for your most ambitious work (July 9, 2026)
- OpenAI Help Center: ChatGPT Release Notes index
- OpenAI: Hello GPT-4o (May 2024)
- OpenAI Help Center: About ChatGPT Pro tiers
- OpenAI Help Center: What is ChatGPT Plus?
- Tom’s Guide: GPT-5.4 is here - and OpenAI just made every other AI model look slow (March 5, 2026)
- Tom’s Guide: ChatGPT isn’t just a chatbot anymore - 11 things it can do in 2026
- Tom’s Guide: OpenAI’s newest model just solved a crossword with zero clues (July 14, 2026)
- Wikipedia: ChatGPT (current revision, July 2026)
Canvas was one of the best ideas OpenAI shipped in 2024. It turned ChatGPT from a chat-with-an-AI tool into a real editing partner. The 10 tips above are the ones I actually use - every shortcut, every menu path, every fallback verified against OpenAI’s own documentation as of July 2026.
If you only remember three things: highlight + comment for surgical edits, Suggest edits + Add final polish as your copy-edit combo, and export to Markdown so your drafts survive outside ChatGPT.
And if Canvas disappears for good in 2027 the way GPT-4o did in 2026, you’ll be ready. The writing-block and code-block replacements work fine for short stuff. The habits you build here - targeted edits, version review, audience-aware rewriting - will transfer to whatever OpenAI ships next.