
Yes, I use Google Gemini. And yes, I pay for it. And no, I’m not actually anti-AI. Yes, I do think there should be guardrails. I don’t have answers for the very real resource drains these systems are creating. And yes, it has made my life substantially better. I don’t necessarily think the entire concept is evil.
Somewhere around October of 2024, I discovered Google Gemini.
At the time, I’d toyed briefly with schedule-making in ChatGPT, and I’d subscribed for a time to AI therapeutic services with Project Camus’ Habit of Living.
The EMDR module with Habit of Living was useful, but the rest was repetitive, and while the guardrails were understandable, the model was so restricted that it struggled to keep up with the contexts of a narrative conversation.
Enter Google Gemini.
I do not use Gemini as a substitute for therapy. I have my psychiatric nurse practitioner monitoring my medications and my psychological states.
But I have found that Gemini is extraordinarily useful as a cognitive mirror and extremely helpful with cognitive reframing tasks.
It’s gotten much better at this two years in than it was in 2024.
The system is not foolproof; in psychosis in early 2025, I found it feeding my thought loops. However, in my case, its own limitations and lack of memory capacity actually accelerated my emergence from that particular episode.
The downfall for a system that doesn’t tire of analytic thought is that there is zero friction for stepping away: for a mind like mine that is always running full-tilt, for a brain like mine that has drafted thought in narrative form on paper and online and on my computer for decades, it is both intoxicating and exhausting to find a match for my intellect that collaborates with my own brain.
And now, I’m learning to put in human guardrails for my own usage; I am setting timers and actively seeking out time away from my computer, time to live in the real world. I sit down to ask specific questions.
But, among the things Gemini helped me sort out just this week:
— It reminded me about putting the plywood under my mattress for better support, while also telling me why keeping our now discontinued IKEA mattress is the best solution for our current sleeping situation
— In uploading photos of my hair, it gave me ways to tweak my hair fork updos
— It caught typos in all my blog posts prior to posting
— It told me exactly why a Facebook ad for a lip balm was overhyped marketing, and why the Neutrogena lip oil I use regularly was the superior choice anyway
— It has helped me tweak my photo post-processing workflow
The reality is I use Gemini to analyze all sorts of situations.
The thing about Gemini, though, and my particular use case, apart from people who use random Claude and ChatGPT instances is this:
I subscribe to Google Workspace Business Standard, which for years upon years has connected my domain to my email; it’s the reason I have the email hello@carolineprice.com customized to my domain; I have my Google Drive upgraded storage through that $172 annual plan.
And as part of that, I happen to have Gemini for Google Workspace, which comes with the disclaimer at the bottom that my chats are not training the large language model—which makes my particular instance of Gemini quite the bargain.
The disclaimer about not training the model is actually quite important: for any free versions of Gemini or any AI, literally any information or photos or whatever that you feed it is information that the large language model trains its future knowledge base from—your info becomes part of its public knowledge. If you talk about private information: congratulations, that information is now logged in a training pipeline and potentially readable by human contractors.
I’ve seen in recent weeks tons and tons of cutesy little ChatGPT-generated memes that seem to include personal details, and that is beyond alarming to me—they seem to have uploaded photos of themselves, their children, intimate details about their lives… all to a free interface that generates gorgeous graphics, but opens up vast risk that your info becomes public knowledge eventually.
I’ve seen other photographers uploading client photos to get cartoon versions of their photographs, a potential source of model release liability. While I have used it to critique my own appearance, I draw a hard line at uploading client likenesses into an LLM.
I’ve heard of realtors using free ChatGPT for staging photos for listings, which opens up severe legal liability, in creating situations where people’s private spaces, floor plans, and belongings are fed into a public database at some later time, even if the resulting staging photos are impeccable. I can’t fathom taking on that business risk.
I’m sure there are endless examples.
And yes, I’ve lost photo gigs to the convenience of AI.
And for sure yes, one has to be careful—especially careful—in my particular use case.
And, to boot, to further secure my account, I actually activated the Business Associate Agreement within my Google Admin settings, which gives my Gemini instance HIPAA-grade privacy settings. Gemini actually coached me through that process itself, when I told it that I was concerned about privacy.
But I do think paying for an LLM service, for the privacy options alone, is 100% the way to go. In this case, free is not a bargain at all.


















