ChatGPT, Claude, and Gemini do mostly the same things under different names. A Gem in Gemini is roughly a Custom GPT in ChatGPT. Claude’s Artifacts are ChatGPT’s Canvas. Once you learn the translations, you can move between the three platforms without relearning much of anything.
If your company pays for more than one of these tools (most do at this point), the naming differences cause real friction. Someone documents a workflow in ChatGPT terms, a colleague on the Claude plan can’t follow it, and you end up answering the same “what’s the Gemini version of this?” question every week. The table below is the answer to that question, in every direction. After the table, I’ve added notes on which equivalents are close enough to swap and which ones will trip you up.
The terminology translation table
Rows are grouped by what the feature does, because that’s the part that stays constant when vendors rename things.
| What it does | ChatGPT (OpenAI) | Claude (Anthropic) | Gemini (Google) |
|---|---|---|---|
| Persistent workspace: a folder of chats with shared files and instructions for one client, campaign, or initiative | Projects | Projects | Gems (closest thing, see note below) |
| Custom reusable assistant: a saved persona with its own instructions and knowledge files, like a “Brand Voice Editor” | Custom GPTs | Projects with custom instructions, plus Skills | Gems |
| Reusable instruction modules: packaged how-to instructions the AI applies to a recurring task | Skills / GPT instructions | Skills | Gem instructions |
| Side-by-side editing pane for drafting documents, code, or prototypes | Canvas | Artifacts | Canvas |
| Long-running research agent that produces a cited report | Deep Research | Research | Deep Research |
| Cross-chat memory (remembers facts about you between conversations) | Memory (saved memories) | Memory | Saved Info + Personal Context (past chats) |
| Standing preferences about tone, format, and context | Custom Instructions | Preferences + Styles | Saved Info |
| Scheduled automations (daily briefings, weekly reports) | Tasks | Scheduled Tasks (via Cowork) | Scheduled Actions |
| Third-party integrations with your CRM, drive, calendar, etc. | Connectors + Apps (and GPT Actions) | Connectors (built on MCP) | Apps / Workspace integration (Gmail, Docs, Drive) |
| Agentic browsing, where the AI operates a browser or desktop for you | Agent Mode (formerly Operator) | Claude for Chrome + Cowork | Agent Mode |
| Coding agent that works in your codebase or terminal | Codex | Claude Code | Gemini CLI + Jules |
| Marketplace / distributing custom assistants to other people | GPT Store | Org-level sharing on Team/Enterprise; no public store | Gem link sharing; no public store |
| Voice conversation | Voice Mode (Advanced Voice) | Voice mode (mobile) | Gemini Live |
| Underlying model family | GPT series (GPT-5.x) | Claude (Opus, Sonnet, Haiku tiers) | Gemini (Pro and Flash tiers) |
Names and availability change often and vary by plan tier. Last checked June 2026.
Where the names map cleanly
Several of these rows are near drop-in replacements. Canvas and Artifacts solve the same problem: instead of dumping a document into the chat stream, the AI opens a side pane you can edit and iterate on. A team that drafts landing pages in ChatGPT Canvas will adapt to Claude Artifacts in a few minutes.
Deep Research is the most standardized term of the bunch. ChatGPT and Gemini use it verbatim and Claude just calls it Research. All three take a brief, search the web in multiple passes, and come back with a cited report a few minutes later.
Scheduled automations also translate one-to-one. A prompt like “send me a competitor news digest every Monday at 8am” works the same way whether you set it up as a ChatGPT Task, a Gemini Scheduled Action, or a Claude Scheduled Task. The naming is annoyingly similar-but-different, which is exactly why teams mix them up, but the concept is identical.
Where the names will trip you up
Gems vs. Custom GPTs vs. Claude Projects
This is the row that causes the most confusion, because the three features overlap without lining up. A Custom GPT and a Gem are both persona-shaped: you configure an assistant once, with instructions and optional knowledge files, and reuse it whenever that role is needed. A Claude Project is a container for many chats that all share the same files and instructions, which fits “this client” or “this campaign” better than “this role.” ChatGPT complicates things by offering both: its Projects work like Claude’s, while its Custom GPTs work like Gems. Gemini has no real Projects equivalent. Gems are the closest thing it offers, and they’re built around personas rather than workspaces.
The translation rule I give teams: if you’re thinking “this role,” build a Custom GPT or a Gem. If you’re thinking “this engagement,” use a ChatGPT Project or a Claude Project.
Skills run deeper than the name suggests
Claude’s Skills are packaged instruction modules. Think multi-page playbooks with templates and decision rules that Claude applies automatically when a task matches. They hold far more detail than a Gem’s instruction field, which caps out at short behavioral guidance, and they follow an open standard that also works in several coding tools outside Claude. If your mental model is “a Skill is a saved prompt,” it’s worth updating: a Skill is closer to a standard operating procedure the AI loads on demand.
Memory works differently on each platform
All three platforms remember you, with different mechanics. ChatGPT’s Memory is largely automatic; the model decides what to save from your conversations. Claude’s approach is more explicit and document-driven, since Project files get read every time and you control exactly what persists. Gemini’s memory is really three things stacked together: Saved Info (facts you tell it), references to past chats, and context pulled from connected Google apps like Gmail and Drive. That last part is the real differentiator. The catch for teams: none of this memory transfers between vendors, so whichever platform accumulates your context becomes harder to leave.
Integrations reflect three different bets
ChatGPT went for breadth, with a large directory of apps, connectors, and GPT Actions for calling external APIs. Anthropic built Claude’s integrations on MCP, an open protocol for connecting the assistant to outside tools, and added agentic surfaces like Claude Code and Cowork on top. Google went deep on its own ecosystem instead: native Gmail, Docs, Sheets, and Drive integration that the other two can’t match. If your company runs on Google Workspace, that alone may settle the question for some workflows.
What to do with this if you run multiple platforms
Standardize your internal docs on function names instead of vendor names. Write the SOP as “create a persistent workspace with the brand guidelines attached” rather than “create a Project,” and attach this table as an appendix. A process documented by your ChatGPT-first content team then works for your Claude-first product team without rework. It also protects you from renames, since vendors rebrand features constantly but the underlying categories have stayed stable for a couple of years now.
Also check where an “equivalent” feature is genuinely weaker before you treat the platforms as interchangeable. Sharing is the clearest example. ChatGPT has a public GPT Store; Claude and Gemini limit sharing to your organization or direct links. If distributing a branded assistant to customers is part of the plan, the platforms are not actually substitutes for that job.
