Claude Artifacts vs. Projects: Which AI Sidebar Feature to Learn First

"What's the difference between Claude's Artifacts and Projects?"

Short answer: one hands you a result, the other stockpiles material. Artifacts opens whatever Claude makes for you in a dedicated window, separate from the chat. Projects lets you upload material and instructions into one workspace, and every conversation started there automatically references them. Artifacts is the side that hands things out; Projects is the side that stockpiles them. If you picked a tool, uploaded a file, and even swapped models in the previous track, today is about the sidebar items you've been glancing past this whole time. If you came here just to learn the difference between these two features, this lesson alone will answer that. All three companies reshuffle their sidebars every few months. ChatGPT's Canvas, for instance, was dropped from GPT-5.5 this past May. So instead of memorizing labels, read for what each spot is actually standing in for.

Artifacts / Projects

Artifacts opens whatever Claude makes for you in a dedicated window, separate from the chat. Projects lets you upload material and instructions into one workspace, and every conversation started there references them automatically. Artifacts hands out a result; Projects stockpiles the material behind it.

If All You Use Is Chat, You're Using Half the Tool

Whether it's claude.ai, chatgpt.com, or gemini.google.com, the left side of the screen is lined with menu items. Most people never touch any of them, they click New Chat, ask a question, and call it a day. And the rest of those items don't explain themselves. Their names alone don't tell you what they do.

Have you ever clicked one of these once and closed it right back? There's a good reason for that. These screens are usually blank the first time you open them. Claude's help center states plainly that artifacts created during a conversation don't automatically get added to the sidebar list; you have to click Publish first. Projects work the same way: the list stays empty until you actually create one. These aren't screens that explain what they do, they're screens that collect what you've already made, so a first-time visitor gets no explanation at all.

That's why this lesson draws a map before you go digging through screens. You don't need to memorize the items by name. Group them by what they actually do, and there are only four buckets.

One Map for All the Features: Grab, Organize, Connect, Delegate

Grab, Organize, Connect, Delegate. This order isn't arbitrary, it's the order in which what you need to prepare and the weight of the judgment call both increase. Let's walk through why each one sits where it does.

Step 1, Grab: Artifacts and Canvas

The first step is Grab because it asks almost nothing of you upfront. Setting up a project means deciding what to upload before you start. Connecting a connector means reading through a permissions screen. Artifacts skips all of that. The only thing worth knowing ahead of time is a single switch in settings. Claude's help center states plainly that Claude stops supporting artifacts if Code execution and file creation is turned off under Settings > Capabilities. That switch defaults to on, so you usually don't need to touch it, but check there first if artifacts refuse to open. Claude's help center also explains that once your content runs past about fifteen lines and would make sense pulled out of the conversation, Claude opens it as an artifact on its own. Keep chatting the way you normally would, and at some point a window just appears on the right. There's nothing to operate here, only something to notice. That's why this step gets the very next lesson.

Step 2, Organize: Projects and Gems

Organize comes second because there's almost nothing new to learn conceptually. You've already done two things in the previous track: uploaded a file into a chat window, and written your own context into custom instructions. A project just makes both of those permanent in one room. Instead of uploading the same file every time, you upload it once to the project. Instead of re-explaining your situation every time, you write it once into the project's instructions. Every conversation you start inside that room references both automatically. The screen is new, but the concept is just a recombination of things you already know. The one thing that puts this a step above Artifacts is that you now have to decide for yourself what belongs in the room. This step gets covered in Lesson 3.

Step 3, Connect: Connectors

The third step is where you need a genuinely new skill for the first time: reading a permissions screen. Through projects, you're only ever uploading copies of files you picked yourself, so a mistake stays contained inside that room. Connectors are different. You're issuing AI a badge that lets it reach into the original files on your Google Drive or Gmail. You need to be able to read exactly what you're authorizing, and know how to revoke it later. What goes up isn't the difficulty of clicking buttons, it's the difficulty of judgment. That's why this step belongs to the next track, not this one.

Step 4, Delegate: Code and Cowork

The last step changes what you're actually reviewing. In the first three steps, you give the instruction and you check the result, every time. Delegation means handing over a goal and letting the process run on its own. Claude's help center describes Cowork as analyzing your request, building a plan, breaking the work into smaller pieces, and running several of them at once when needed. What you review widens from a single reply to an entire workflow. Starting here before you have a feel for what's safe to hand off leaves you with no basis for trusting the outcome. That's why it waits for the CLI track and the advanced material after it.

Some of you will want to skip ahead. My advice is still to start with Grab, for one reason. The first two steps show you the result immediately, and there's nothing to undo if something goes wrong. From Connectors onward, you're granting access to your own accounts, so you need to know how to revoke that access too. It's safer to let how much you trust a tool grow at the same pace as how much responsibility you hand it.

Sidebar Comparison Across the Three Companies (as of August 2026)

These tools do the same job under a different name at every company. That's exactly where search results start to diverge, so let's line them up in a table first.

Only one cell in that table needs unpacking. Gemini splits the Organize slot into two. Gems is your own personalized version of Gemini, built to hold instructions and knowledge files. Notebooks is a space that remembers your sources, your requests, and an ongoing conversation together. Google's help center even titles its Notebooks article "Organize your projects with notebooks in Gemini Apps." Neither one requires a paid plan to create. Notebooks first opened to Google AI plan subscribers on the web in April 2026, and Google's current help center listing no longer mentions a plan requirement at all. Where Gems only requires you to be 13 or older, Notebooks carries three conditions: you must be 18 or older, use a personal Google account, and have Gemini Notebook connected to the Gemini app. That connection happens automatically in most regions, but users in the European Economic Area have to set it up manually, and the rollout is still gradual enough that some accounts don't see it yet.

Three things worth knowing before you read the table too literally.

First, a different name doesn't mean a different feature. Artifacts and Canvas occupy the same slot. Learn Artifacts on Claude, then move to Gemini, and the only thing left to learn is how to open it.

Second, not everything in the table lives in the sidebar. Gemini's Canvas isn't a sidebar item at all, you switch it on from the menu that opens when you click "add files" below the input box. ChatGPT's writing blocks aren't a sidebar item either, they appear directly inside a reply. That's why people report searching the sidebar and coming up empty. Read the left column of the table as a grouping by function, not by screen location.

Third, this table will go stale. ChatGPT's Canvas is proof. OpenAI's release notes state that as of May 28, 2026, GPT-5.5 Instant and GPT-5.5 Thinking no longer offer Canvas, with writing and coding support moved into writing blocks and code blocks inside the reply itself. Paid users can still use Canvas, but only temporarily, until the legacy models it depended on are retired. When autocomplete suggests "chatgpt canvas gone," that's not a rumor. It actually happened.

What's the Difference Between Claude Code and Cowork?

One item still sits near the bottom of the sidebar: Code. Cowork doesn't get its own sidebar slot at all, you pick it right next to Chat in the message box. Some reassurance first: you don't need to open either of these today. All you need right now is to know where they are and what they do.

Code links to claude.ai/code. Its full name is Claude Code on the web. You pick a code repository you've pushed to GitHub, describe what you want done, and Claude pulls it down and works on it inside an isolated remote environment that Anthropic manages. It runs without anyone watching, and hands back changes for you to review when it's done. According to the official docs, this is a research preview for Pro, Max, and Team plans and qualifying Enterprise users, and it isn't open to the free plan. If you don't have a repository to point it at, clicking through gets you nowhere past the first screen.

Cowork is for handing off work that isn't code. Claude's help center describes it as Claude carrying out complex, multi-step tasks on your behalf: breaking down the request, building a plan, splitting the work into pieces, and running several of them in parallel when it makes sense. Think file cleanup, pulling research together, or building out a spreadsheet with formulas or a slide deck. It's paid-plan only. On the web that means Pro, Max, or Team, and Enterprise only if an admin has turned it on. Chat and Cowork share the same screen, so you start a Cowork session by selecting it from the message box.

The difference comes down to what you're handing off. Code takes a code repository, Cowork takes files and documents. Both belong to Step 4, Delegate, and both sit outside this track. Code gets covered in the CLI track, and Cowork comes later still. What matters right now is that a locked sidebar item doesn't mean you're missing out on something. Its turn just hasn't come yet.

What You'll Build in This Track: A Monthly Client Content Report

A map alone doesn't leave you with anything to show for it. This track works toward one concrete deliverable: a content report you send a client every month. You don't need to be a consultant for this to apply. Any document you rebuild in the same format every period fits here, whether that's a weekly team report or a monthly sales summary for your store. The structure is the same either way.

• Lesson 2: Grab the report skeleton as an artifact. Instead of disappearing into a chat scroll, you end up with one editable document. (Step 1, Grab)
• Lesson 3: Move that skeleton into a project. Upload your client's material and your writing rules to the room once, and next month you start with a single line: "Draft this month's report." (Step 2, Organize)
• Lesson 4: A checklist to confirm what you've built so far and flag the common mistakes.
• Next track: Eliminate re-uploading material every month altogether. Connect the source directly, and you stop carrying copies around. (Step 3, Connect)

I found this list embarrassingly late. I'd used ChatGPT for a long time before switching to Claude, and even after switching I didn't know Projects existed. ChatGPT had GPTs for the stuff I did over and over, so I assumed Claude had no equivalent and just used plain chat.

Here's what I did instead: I dug through my old conversations. Whenever I needed to do something similar to a past task, a document or a piece of text, I'd find that old thread and keep working inside it. Scrolling back through months-old conversations became part of my routine. Opening a new chat meant explaining my situation from scratch, and I wanted to skip that hassle, but looking back, I was doing by hand exactly what Projects was built to do.

The item had been sitting in the sidebar the whole time. I just never clicked it.