Gemini is a tool I regret not using sooner - it excels in real-time information retrieval, has a massive context window, is built for research-heavy work, and it also seems to understand my prompts a little better than other tools like ChatGPT. While I was already relying on Google’s Gemini-powered AI Mode in Google Search, I eventually started switching over to using the dedicated Gemini interface, and it gave me so much more to work with.
One of the features in Gemini that I’ve come to appreciate a lot these days is Gems. They’re somewhere between custom AI assistants and AI agents that not only help you work faster by speeding up tasks, but they also keep your workflow consistent. Here’s how I make the most of them in my regular daily AI workflow…
What exactly are Gems?
Gemini’s custom AI experts
Gemini Gems are referred to as "custom AI experts". They're essentially customizable little versions of Gemini designed to handle specific tasks or topics. Instead of opening a blank chat and rewriting the same prompt with the same instructions again and again, Gems let you define the role, instructions, and behavior of the AI once. Then you can use the Gem across multiple chats to repeatedly apply those same instructions. Gemini comes loaded with a handful of preset Gems, but it’s best to create your own for your workload and domain.
They’re meant to standardize or automate things that are repetitive. For example, they’re perfect for marketing content creation where a business needs to create batches of brand-aligned email sequences or social media posts. You can also set them up for repetitive technical tasks such as drafting specific code snippets. Or extracting key points from your queries and files, such as performance data and KPI trends, every week.
Whatever it is you’re working with, Gems remove repetitive prompting. Once a Gem is configured, it knows the context, how to behave, and can jump straight into the task without needing an explanation.
How to create and use a Gem
Gemini makes it simple
Gemini designed Gems to be a core part of the Gemini experience, so they’re very easy to access and navigate. Open the hamburger menu in the left panel to reveal your chats, and select Gems at the top. In the Gems page, you’ll see Gems from Labs, which are different from the classic Gems I’m talking about here - they’re mini apps meant for automating complex workflows. Classic Gems are right under them, and there will be a couple of presets made by Google.
Hit New Gem and give your gem a name and description. The Instructions box is where you will give it a prompt that defines its role, behavior, and your expectations - Gemini can actually help you refine these instructions as you write them (click the magic wand icon). I recommend sticking to the four-part prompt formula that’s been widely adopted by the AI prompt engineering community:
- Persona: Give the Gem a role, such as “design instructor”.
- Task: What do you want Gemini to accomplish? Draft emails, write code, proofread?
- Context: Background information, such as the target audience of your newsletter.
- Format: How do you want the output to be structured? Tables, bullet points?
Upload any relevant files if needed - I added some of my design course materials to my “persona and scenario generator” Gem. And in the Preview box on the right, you can test out the Gem with prompts to see how it performs, and make the necessary tweaks. To actually use the Gem, you’re going to go to the Gem Manager again, find your Gem in the list, click the three-dot menu, and select New Chat. This chat will now incorporate your Gem.
No more context rebuilding
And easy task switching
The biggest change Gems made to my workflow is that I don’t start from a blank canvas every time anymore. Normally, the first few prompts in my AI chats are spent setting things up. I explain the role, the format I want, give it context, and only after all of that does the work actually begin. Gems let me skip that step entirely. All I have to do is open a Gem that matches the task and start working immediately.
Another cool thing is that Gems are reusable indefinitely. So I can create as many as I need, and jump to whichever one I need depending on the task at hand. The Gem already handles the context for me, which reduces the cognitive load of task-hopping and helps me keep moving.
A small library of AI helpers
The whole point of Gems is to handle repetitive tasks with more speed and consistency. It’s almost like a system prompt, except you have a collection of them that you can flip between. Gemini also gives me more consistent and predictable responses this way, which means less editing and reprompting on my end. It’s the perfect way to stick to a brand kit, style guide, format, or whatever else you’re working with. And the coolest thing is that they’re available to free Gemini users without limits.