If you built a custom GPT in ChatGPT or a Gem in Gemini, it has an end date. Google starts turning Gems into skills on 17 November 2026, and OpenAI retires custom GPTs on 11 December 2026. Both companies are moving to the same idea, a skill: a packaged set of instructions and files that the AI loads when a task needs it. For most people this is good news, because a skill can travel between tools in a way a GPT never could. But a few things won't carry over on their own, and the weeks before the deadline are the time to deal with them.
Why everyone is switching to skills
Custom GPTs and Gems were locked to one product. A GPT only worked inside ChatGPT, and a Gem only worked inside Gemini. Skills are different because they follow an open format. A skill is a folder with a file called SKILL.md that holds the instructions, plus any reference files or scripts the task needs. Anthropic published that format as an open standard, called Agent Skills, in December 2025.
On 22 October 2026, Anthropic handed the standard to the Agentic AI Foundation, a Linux Foundation project that also looks after MCP, the protocol AI tools use to connect to other apps. Anthropic says more than 70 agent platforms now support Agent Skills, including ChatGPT, Cursor, GitHub Copilot and VS Code, and that about one million skills have been published. In practical terms, the work you put into a good skill is no longer tied to one company's product. Google says Gemini can already import a skill from another platform by uploading its SKILL.md file.
The dates that matter
For Gemini, the change starts on 17 November 2026 for personal Google accounts belonging to adults. On that day your Gems become read-only, and Google recreates them as skills automatically. On 17 December 2026 the Gems themselves are removed. Workspace business, enterprise and non-profit accounts move in March 2027, and education accounts in June 2027.
For ChatGPT, custom GPTs stop running on 11 December 2026 on every plan. Enterprise workspaces with an approved deferral get until 11 February 2027. Your old conversations with a GPT stay available, but the GPT itself and its page do not. OpenAI says these dates could still change, so check the notices in your own workspace too.
What carries over, and what breaks
Google's automatic conversion keeps your Gem's instructions and its supported knowledge files. It drops files over the 100 MB skill limit, starting with the largest, and file types skills don't support: audio, video, archives such as .zip, GitHub files and old Office formats like .doc, .xls and .ppt. More importantly, Google says most of the default tools available in Gems don't work with skills, including Canvas, Deep Research, Create video, Create music and Guided learning. If your Gem depended on one of those, it will behave differently as a skill. Download any files you want to keep before 17 December.
On the OpenAI side, nothing happens automatically. You have to migrate each GPT yourself, and the result is a ChatGPT plugin, which is a bundle that can hold skills, connected apps and MCP servers. Your GPT's instructions become a skill and its knowledge files become reference files. Custom actions, the part of a GPT that called outside services, do not transfer at all and need to be rebuilt, usually as a connector or an MCP server. The model you picked, your sharing settings and any unpublished drafts don't carry over either. The new plugin starts out private, even if your GPT was shared with your whole company.
How to migrate a custom GPT
Only the person who created a GPT, or an Enterprise workspace admin, can migrate it. Open My GPTs in ChatGPT, choose Migrate to plugin on the GPT, review the details and follow the prompts. Only the latest published version moves, so publish any edits you want to keep first. After the migration, the old GPT becomes read-only until it retires.
Before you click, copy the GPT's full instructions into a document you own and note every action it uses. Then, once the plugin exists, test it on three or four real tasks. A plugin's skill loads when ChatGPT decides it is relevant, so it may not run on every message the way a GPT did. If it doesn't trigger when you expect, make the skill's description more specific about when to use it.
How to move a Gem early
You don't have to wait for Google to convert your Gems, and doing it yourself lets you check the result. In Gemini, open Customize in the sidebar, then Skills, then Create and Create manually. Paste in your Gem's name, description and instructions, and select Create. To add files, download the skill as a .zip from its menu, put your files into the folder, and use Replace skill. You can then call a skill by typing a slash followed by its name, and Gemini will also apply it on its own when it seems relevant. You can have up to 100 skills switched on at once.
What we would do this week
Make a list of every custom GPT and Gem you or your team actually use, and next to each one note whether it relies on actions, large files or tools like Deep Research. Those are the ones that need attention. Everything else will move with a few clicks.
Then write your best assistant once, as a proper skill, and keep the SKILL.md in a folder you control. The same file can work in Claude, ChatGPT, Gemini and a growing list of other tools, which means the next time a product is renamed or retired, you upload a file instead of starting again. Our Learn lessons cover setting up skills and Projects in each of the big three assistants, and our Stacks show where a skill fits in a team's everyday workflow.