Losing Gems is an upgrade. I know that puts me at odds with roughly everyone who has posted about this in the past week, and I know the slash-command interface Google is moving to looks like a step backward for anyone who just wants to type a question and get an answer. I still think the trade is worth it, and I want to explain why before I get to the parts that genuinely worry me.
Here is what is actually happening. Google is retiring Gems, the custom AI assistant feature inside the Gemini app, and replacing it with a format called skills. A notice in the app tells users that starting November 17, 2026, Google will automatically begin migrating existing Gems to skills. You keep using your Gems until they migrate. No action required on your end. Skills were introduced alongside Gemini Spark, and they get invoked by typing a forward slash in a chat thread and picking one from the list.
Why the slash is not the insult people think it is
The main complaint I have seen is that slash commands feel less consumer-friendly than just talking to an all-in-one agent, the way you would with something like Meta’s Muse. That is a fair read of the interface. It is also, I would argue, the wrong thing to optimize for if you actually use these tools for work.
Gems had a discovery problem. You built one, you named it something clever, and then it lived in a side panel you forgot existed. Every custom assistant I have built across every platform has followed the same arc: enthusiastic setup, two weeks of use, permanent abandonment. The friction was never in the chatting. It was in remembering the thing existed and then navigating to it.
A forward slash inside the thread you are already in solves exactly that. It is the same pattern Slack, Notion, Linear, and Discord landed on independently, because it turns out that typing one character to summon a menu is faster than any amount of deliberate navigation. If you have ever used a command palette, you already know this. The interface is not less friendly. It is less obvious, which is a different problem and a smaller one.
What actually concerns me
The migration is automatic, which sounds reassuring until you think about what automatic migration usually means in practice. Custom Gems carried instructions, uploaded files, and tone calibration. Skills are a different format with a different invocation model. Google has said it will move your Gems over. It has not, as far as I can tell from what has been shared, spelled out what survives that move intact and what gets flattened.
That is the question I would want answered before November 17. If you have a Gem you depend on, the responsible move is to copy its instructions into a plain text file now. Not because I expect Google to lose them, but because verifying a migration is a lot easier when you have the original to compare against.
The second concern is access. Skills arrived with Gemini Spark, and there is a real question about whether free users end up on the same side of the wall they were on before. Migrating people’s custom assistants into a format that requires an upgrade would be a bad look, and I have no evidence that is the plan. I also have no evidence it is not.
The naming treadmill is the real cost
This is the part where the criticism lands squarely. Google has a habit of renaming and merging features instead of letting a product name sit still long enough to mean something. Gems existed, users learned the word, wrote tutorials around it, built workflows on it, and now the word is going away. Skills will presumably be the word until it is not.
That churn has a cost that never shows up in any changelog. It is paid by the person who wrote documentation for their team, the reviewer whose old posts are now wrong, and the casual user who concludes the whole thing is too unstable to bother learning. Better plumbing does not compensate for teaching people a vocabulary and then swapping it out.
What I would do
- Export the instructions from any Gem you actually use, before November 17, 2026.
- After migration, run one real task through the new skill and compare the output to what you remember getting before.
- Give the slash interface a week before judging it. Command palettes feel awkward for about three days and then feel mandatory.
- Watch the access tiers closely. That is where this either stays a format change or turns into something else.
My honest read: the mechanism is better, the branding discipline is worse, and the migration is the part worth watching. Google gets credit for moving custom assistants to a pattern people already have muscle memory for. It loses that credit if the handoff drops anything, or if the price of a skill turns out to be higher than the price of a Gem.
đź•’ Published: