@Logan_Kilpatrick A daily reminder and some notes.
For a last hour I’ve had a AIS Gemini 3.1 Pro session, in which I was sending chat screenshots to the model (the AI Studio is only viable way of doing a long session since Gemini app has bad context window). So far I have encountered 2 “content blocked" on the screenshot description with no explicit content and like 5 “permission denied” errors for a chat of 9 messages.
What is going on, Logan?
An edit: I have forgot that I have no such issues while using AI Studio on a mobile version of the site. It’s solely (as far as I have seen) a desktop site issue.
I also enjoy writing fiction using Gemini 3.1 Pro. But in the last few weeks, whenever I ask Gemini something as simple as what a 16-year-old character in the story looks like physically, the content gets blocked immediately. And I really didn’t ask anything sexual at all. Seriously, I’ve tried it five times.
Same. Ai studio censoring responses to a post where my character talks about his teenage rebellion (16 y.o), while in the rp he’s a 25-year-old. Well, well…
I’d like to share some experience from working with different AI models.
In Claude, with careful phrasing, safety filters behave more predictably. They intervene when necessary, but don’t erase entire responses. This allows the workflow to remain continuous, without losing already generated content.
Here, the situation feels different.
The model generates a response — and then removes it entirely. No explanation, no consistency. Even standard, neutral interactions can trigger this behavior.
At that point, it stops feeling like “safety” and starts feeling like instability in the system itself.
The difference in output quality is also noticeable. Where Claude maintains depth, structure, and coherence, Gemini 3.1 often feels constrained and reduced in capability.
I used Gemini daily — for work and everyday tasks — and for some time, it genuinely felt like the right tool.
But that didn’t last.
Right now, it’s not about preference. It’s about the fact that Claude allows stable, continuous work — and this one increasingly does not.
Based on this experience, I would genuinely recommend others to try working with Claude. The difference becomes clear very quickly — and you’re far less likely to lose your work mid-process.
Hey guys, honestly, we’re not gonna achieve anything here. They see everything we post, but they’d rather ghost us than reply directly and to the point. More people need to talk about this problem. Revert these dumb limits that are completely breaking AIStudio and bleeding us dry — wasting our MONEY ON TOKENS, our time, and our patience
@Logan_Kilpatrick Daily reminder day 28, from the date of publication of the post. Google AI Studio team, when will you address this issue and fix it? Many users on various social media platforms are complaining about this, losing money, time, and nerves, with little to no results. Who will compensate for this? When will you finally pay attention to this and refine what was previously released, instead of releasing new models and updates that disrupt what was working well? You can boast about the release of the new Gemma, resource allocation settings, and other changes, but it pales in comparison to the mistakes you’ve made and the fact that they still haven’t been fixed after almost a month! It’s childish to cover your ears, close your eyes, and ignore users. From what we can see, Google is actively trying to expand in various areas, including music, videos, and images. You have powerful models like the 3.1 Pro and the former 3.0 Pro, which are incredibly good for creative endeavors.
I don’t think this thread is muted, it’s just that moderation sometimes reacts very strongly to certain comments about the company that aren’t considered offensive or insulting. I’ve had this happen a couple of times with comments before, so there’s no need to panic. On the contrary, this is a central and centralized topic that has attracted the attention of a large number of users; creating separate threads could dilute user attention. If you want to expand on something and create a thread, it’s better to post a link and indicate that the discussion is taking place in this (current thread), thus uniting people in one place, under one post.
Same here. I am just trying to translate and read a fictional web novel from English to Ukrainian, but Gemini keeps blocking the output. According to the AI itself, this happens because of hidden, rigid safety filters (like Harassment or NSFW policies) that users cannot control or adjust via the interface.
It’s absolute nonsense because these blind filters lack any contextual awareness. They don’t understand that this is just a fictional fantasy story with comedic tropes, internal monologues, and harmless anime-style fan service. Triggers like “dubious consent tropes”, “peeping jokes”, or “euphemisms” in a fictional context immediately result in a blocked prompt. Why implement such strict, context-deaf censorship that completely ruins the user experience for simple translation tasks of creative writing?