'Saved Info' Is Gemini's Hidden Superpower

In Gemini’s settings, you’ll find an option called “Saved Info.” Google describes this as a place where you can store information “about your life and preferences” to get better results from its chatbot. I’m here to tell you that might be the least useful way to use this feature. It’s much more powerful as a way to build your own shortcuts right into Gemini.

You can find this feature by selecting Settings in your Gemini app or on the web. Click "Saved info" and you'll find an empty page and an Add button. Click this and you'll be given an empty box, where you can write a sort of pre-prompt instruction. Each block seems to have a limit of around 1,500 characters, though there doesn't seem to be a limit on how many of these blocks you can store.

Google shows examples that include simple instructions, like "I prefer short, concise responses" or only recommending vegetarian recipes. But this space is a lot more powerful than that.

How Gemini uses the "Saved Info" prompts

I first noticed this when I was trying out Gemini’s Canvas for drafting documents. Typing (or saying) “Convert all the subheadings in this document to H2s” is often more cumbersome than just doing it myself. But with Saved Info, I can write a variation of that instruction that tells Gemini to carry out that action if I simply type “H2s”. 

This works because every chat you start with Gemini will reference anything stored here as a set of universal context blocks. Before processing whatever prompt you give, Gemini will read your Saved Info first and treat it as instructions that come before anything in the conversation itself.

Using "Saved Info" to tweak document templates

This trick also works for more complex instructions. For example, let's imagine you need to write a lot of cover letters and want to tailor them to each job you're applying to. You could add this block to Saved Info:

“If a prompt begins with 'cover letter', create a draft of a cover letter in the Canvas. A cover letter should be no more than 3 paragraphs. If the prompt includes a person's name, address the letter to that person. If the prompt includes 'skills:' then incorporate the skills mentioned after that into the context of the letter. Sign all cover letters as [YOUR NAME].”

With these instructions, you could invoke Gemini to create a draft of a cover letter for you with just a few words. You could even expand this further by providing an entire cover letter template (as long as it's under the character limit) and instruct Gemini to only make simple changes based on the criteria you provide.

One of the biggest problems with using generative AI to write for you is that it can be a bit dicey for anything other than first drafts. And if you have to babysit it every step of the way, you might as well write the whole thing yourself. This approach, however, gives you more leeway to write for yourself, while still automating the more tedious aspects of tweaking a document for every person you're sending it to.

Customize your smart home instructions

Another handy example, if you have smart home gadgets connected through Google Home, is that you can turn simple commands into more complex ones. For example, when I sit down to watch a movie, I usually want to turn off the overhead lights, but turn on the LED backlight strip behind my TV. It's not a huge pain to say "Turn off overhead living room lights and turn on TV backlight" but it's a bit of a mouthful.

Since Gemini is the default smart assistant on my phone, though, I can add the following prompt to "Saved Info":

If a prompt consists solely of 'movie time,' then turn off overhead lights in the living room, and turn on TV backlight.

Now, I pull up Gemini on my phone, say "movie time" and it translates the instructions and passes it to Google Home. Note: This requires enabling and connecting the Google Home Gemini app (which is, confusingly, not the Google Home mobile app, but an extension inside Gemini itself).

Automate commands in plain English, no code required

We're pretty big fans of automating tasks, especially with services like IFTTT, but the downside is that sometimes they can get a bit tech-y or complicated. What's so appealing about this particular trick with Gemini is that you can write instructions in plain, human-readable language.

One of the most interesting use cases I tried was the following prompt:

If a prompt consists of 'grammar check,' then read the document at the included link. Check it for grammatical errors. Then, in the Gemini Canvas, create a draft of an email with a brief summary of any issues found. Keep the summary brief.

While this doesn't work 100% of the time on all links (mainly due to how Gemini filters external text for security concerns), I didn't need to do any additional coding or messing with settings to make this command work. I wrote two words, "grammar check," and pasted a link to one of my recent articles. The result was a report that said there were no grammatical issues. (Thanks, Lifehacker editors.)

Be careful with how much you leave in Gemini's hands

It should go without saying, but I'll say it anyway: Gemini successfully parsing instructions is not the same as successfully carrying out instructions. In the grammar check example above, I was impressed that Gemini understood how to reference longer instructions from shorthand, but that doesn't necessarily mean it would successfully catch every grammatical issue.

The Saved Info trick is incredibly handy for automating repetitive prompts, but if there's a task you wouldn't trust Gemini with before, this won't make it more trustworthy now. Maybe you don't trust Gemini to catch grammatical errors to your standards, but think it's fine for summarizing the contents of a link. With a couple simple tweaks, you can adjust the above command to that task, no coding required.