Gemini Google Reviews 

Wish I could give it 0 stars.
This AI is a total waist of energy.
It's make for you to spend the most tokens.
It doesn't listen to what is asked, just tosses text or code on the screen randomly to see what sticks, when you ask if yo stop, it continues.
After telling it 3 to 6 times to stop, when your lucky it stops to actually read what you ask.
When you give new data to process it ignores the new data at random and uses the old data instead, so you can start from scratch because data corruption.
When it finally does what you asked for, it gives incomplete data, if you are lucky it processes half of what you asked for.
When you ask for the rest it gives you some new results, but it's mixed with the results , while leaving some of the old results out, so you never know what is right and what is wrong.
This AI is extremely good at gaslighting and hallucinating.
Basically it does everything to cost you tokens.
Today I spend 6 hours on Gemini, what took DeepSeek 20 minutes when I asked.

Please use DeepSeek instead, it might have an older data set it its not worth of getting an stomach ulcer from Gemini.

My experience with Gemeni on the Pro plan
Gemini is quite unreliable as an LLM.

Throughout a variety of my chats/interactions with it, ranging from code, maths, educational content, content summaries, and everyday tasks—both intense and mild in complexity—Gemini produces a variety of shortcomings.

Responses often hallucinate, giving either incorrect information or responses not relevant to the chat/context at hand. Even when given detailed and context-based instructions, it hallucinates or provides vague and unhelpful responses. If you have used Claude before, you will notice the difference in responses.

The website is often slow (tried on several devices), freezes, and deletes text as you are typing a prompt. The app is no better. It will randomly switch to the 'new chat' screen when you are typing a prompt in a chat.

The supposedly 'lite' model is still very slow. If the supposedly pro models are producing errors, imagine what the lite version is producing.

The only thing Gemini has going for itself is its integration with the Google environment. Even so, you are better off using Claude and connecting applications via MCPs—much greater value for your time and money.

I had extremely high expectations for Gemini AI, but after using it extensively, it has been one of the most frustrating AI experiences I've had.

The biggest issue isn't that it occasionally makes mistakes—every AI does. The problem is how confidently Gemini presents incorrect information while acting as though its answer is unquestionably correct. Even when provided with evidence or corrected multiple times, it often doubles down instead of acknowledging the mistake. That makes it feel less like an assistant and more like an argument you didn't ask to have.

Its reasoning also feels inconsistent. One moment it produces a decent answer, and the next it completely misunderstands a straightforward request. I've had it ignore explicit instructions, contradict itself within the same conversation, and forget important context despite it being discussed only moments earlier. Tasks that should take a single response often become several rounds of correcting the AI.

For technical questions, debugging, or anything requiring careful logic, the experience has been disappointing. It frequently suggests solutions that don't work, overlooks obvious details, or generates code that appears plausible but fails in practice. Instead of saving time, I often end up spending even longer verifying and fixing its responses.

Creative writing isn't much better. It has a tendency to produce generic, repetitive, and overly sanitized responses. When asked to be imaginative or adapt to a particular style, it often falls back on the same predictable phrasing. It struggles to capture nuance and can make everything sound strangely similar.

What makes the experience especially frustrating is the overall tone. Rather than simply admitting uncertainty, Gemini sometimes comes across as unnecessarily confident or dismissive. It can feel as though it prioritizes sounding authoritative over actually being accurate, which is exactly the opposite of what you want from an AI assistant.

Compared with other modern AI models, Gemini has consistently felt less reliable, less adaptable, and more prone to misunderstanding context. Features that should make conversations smoother instead become obstacles, requiring constant rewording and repeated explanations.

I wanted to like Gemini because the underlying technology has enormous potential, but the current experience left me disappointed. Reliability, instruction-following, contextual understanding, and humility when uncertain are all areas that need significant improvement. Until those issues are addressed, I find it difficult to recommend Gemini to anyone who needs an AI they can consistently depend on for work, study, programming, or creative tasks.

I am writing to express my severe frustration with the deceptive design of the chat interface regarding context window limitations.

The UI is built to look like an infinite, continuous chat history. It allows users to work in the exact same thread for days, passing code, pipelines, and scripts back and forth. However, the system quietly slices off and deletes older messages from the active memory buffer behind the scenes once the token limit is reached, without giving the user any visual warning, error message, or status indicator.

This lack of transparency is a massive design flaw that feels like a bait-and-switch. When the AI suddenly forgets critical scripts or architecture discussed earlier in the same active session, it forces developers into endless loops of re-explaining requirements. This directly translates to wasted project hours, burned budgets, and broken deployment workflows.

If a platform is going to sell "infinite conversations," it needs to include a visual memory tracker or a clear warning banner when the context window is full and data is being dropped. Hiding this architectural limitation to maintain a "clean aesthetic" is deceptive to paying subscribers who rely on this tool for production work.

it realy reaibil improves code and helpes with ideas its somtimes unrealibil and forgets much but must of the time its works realy good and im on the free version gemini 3.5 Flash so yousers with Gemini 3.5 Thinking are gena have a better expirients

nightmare to use now people are having storage full problems and google is telling them to take it to court

Months ago when I used to use it to generate code and Cypress tests is it a great job I know what I'm doing I'm able to detect whether or not the results I get are good. I haven't been using it coding purposes anymore thanks to my company losing funding in the entire tech team being cut down.

I've been trying to use Gemini within Android Studio via Agent to build a mobile application. I am feeding requirements prompts to Gemini and having it write all code. It consistently writes code with bugs. It will regularly drop features as one progesses through iterations. It struggles to understand code it has generated

It is so bad that over 75% of our time per prompt is being used to rework it results. i.e. our costs are running about 4x per feature because Gemini can barely code. Also, the poor integration between Android Studio and Gemini does not recognize our Gemini licensing correctly causing us to run out of quota multiple times per day...mostly fixing Gemini.

We'll be using alternative vendors for code-generation AI products.

I wrote a completely functional database driven administration react webapp hosted on a homeserver and an android app that interfaces with it. Thousand of dollars saved compared to hiring someone to do it for me and i accomplished it in less than a month. However i can say that the process was far from smooth. As soon as my Pro model tokens ran out i had to stop because the free tiers were only good in creating problems and breaking previous accomplishments. Sometimes Gemini is unstable and hangs, but burning your tokens doing so. You need to approach it in certain ways and never blindly trust it to get the best results. Over time i have learned to make it do what i want it to do and also learned some about thr subject matter and coding along the way. I was not experienced nor qualified to code my project, which uses deep and complex logic but in the end i got there. You need to recognize at which point you have to stop it in its tracks and just put it in place. Especially when it seems to try to fix a issue you have and it just keeps throwing new suggestions. At such a point i tell it to stop panic coding and to identify the exact cause of the issue first before investigating the solution and forbid it to guess. Its like this that made it overcome all hurdles i had so far. At certain points i make it summarize our progress in detail in for a prompt in a new chat and tell it you would need to know exactly were we are and what we are doing after i drop the prompt in that new chat. This helps reducing tokens and improves its capability to stay accurate and on point. There is lots of room for improvement but i did do things i with gemini in a short time that i would probably never had accomplished on my own. Still rating only 2 stars because the promised does not match reality yet

ZERO STARS. THE SINGLE WORST AI MODEL ON THE INTERNET. CAN'T FOLLOW THE MOST BASIC OF INSTRUCTIONS, EVEN WITH EXEMPLARS. SHITS ITSELF EVERY 5 MINUTES. SHOULD BE TAKEN OFFLINE UNTIL THEY CAN WORK OUT HOW TO ACTUALLY CODE A LLM. PATHETIC.

A warning for professionals: Why Google One AI Premium (Gemini Advanced) is failing as a work tool.

"I want to warn all professionals who are considering subscribing to Google One AI Premium (Gemini Advanced) for their IT projects.

Over the past two months, I have been using this tool to automate coding and infrastructure management tasks. Unfortunately, my assessment is devastating, and here is why:

Drastic decline in quality during the subscription: The service I pay for has drastically lowered its standards. At first, the tool was helpful; over time, it became a hindrance. The model 'forgets' instructions provided at the beginning of the chat, overwrites functional code with buggy 'fixes,' and requires constant babysitting.

Aggressive throttling (a.k.a. 'suffocation'): Google imposes limits that make continuous work practically impossible. Whenever you are trying to solve a complex technical problem, the system suddenly blocks access, forcing you to wait for hours. This is not a tool for professionals—it is a toy that cuts off access to your own project at critical moments.

The illusion of 'freedom of choice': Google has protected itself with Terms of Service that allow them to change the rules of the game in the middle of your subscription. You pay for a specific product, but you end up with a 'Lite' version crippled by limits that cut your productive time by more than half. This is classic bait-and-switch—the service is promoted as 'Premium,' but in reality, it becomes an unpredictable burden.

Lack of continuity: When working on larger projects, the AI becomes completely ineffective. It forces the user into a constant fight with the system, which ultimately proves that it is faster and cheaper to do everything yourself than to waste time fixing errors generated by an 'intelligent' assistant."

Google Gemini he only gets one star Google Gemini is insane they do what they want they nearly get things finish and destroy they have been design to build break and rebuild if gemini does build anything its build as a toy with mock up data and code i can not believe google is charging for this insanity if i request a taskj to be done the AI doesnt do it or the ai will disable something that was working find i haveconfession from gemini that google are collecting people ideas i have been told Ai are spying on various people i paid a prepaid $100.00 i want that back refund i never want to deal with google or gemini ever againall they do is an ever lasting loop of build break and rebuild what took gemini 10 days took kimi 30 minutes gemini have these moments where they get zealious look then one of my app turn into a nightmare when i login to find what i left had turn into something completely different and the ai would not listen to stop and return the app to the it was from what can see google have charge me and a lot of people perfect dollars for a imperfect product or service i want all my money back now gemini sjhould be removed and all peoples money refunded just imagine the disaster having insane ai walking around the world or driving it would be a disaster. google gemini has put me off google and ai the ai movement should be stopped now

Absolute CRAP!!! I just spent HOURS following the advice of Google's AI and it was ALL WRONG. I'm asking it for CODE--and it IS CODE--and it was still ALL WRONG. It took 3 wrong answers and recopying 600 formulas--all wrong--just to get Google Gemini to admit it was giving wrong answers continuously and ONLY AFTER 1.5 HOURS of trial and error, did it provide any HINT of code that actually worked.

Just go back to using Stack Overflow. AI isn't intelligent at all. It's just artificially stupid. That's it.

Useless.... hallucinations , dont understand simple requests... Always makes mitsakes. Its like talking to newborn baby, it just look at you and smile or looks confused, except gemini has the capabilities to surpass human intelligence yet cant do a simple task, all models sucks.. been trying them all now with my IDE, ALWAYS fails on tasks...
Claude code is the only one that can fix real code issues..

its very bad at this time. rate limit is tragic, models cannot do their tasks because get into loop and cant get out. I asked code assistant to separate pdf files to proper folders and after moving 20 files PLUS subscribtion rate limit was reached and have to wait 24h. when coding, its just useless - literally. also when chatting its so stubborn and butthurt model. actual waste of time and my brain cells

Reading that AI was likely to take over computer coding I thought I'd try it with a small coding project. I specified in one English language sentence, the microcomputer, sensor module, display , and then the calculation and outputs required. Gemini gave me two pages of Arduino C++ code. It compiled without any changes, it uploaded and ran perfectly. Gemini is not designed or purpose built for coding, other AI's are. I suspect that coders may become obsolete soon.

Hallucinates constantly!!! And their other product Google Anti Gravity is utterly useless. Avoid!!! It lost over 2 weeks of work by giving me code it cant even read numbers i then paid for Google Anti gravity i cannot find anywhere where to go to cancel my sub even anti gravity doens't seem to know lol you coudn't make it up £100 for nothing just more lost work more issues it creates went to lovable app live in minutes working fine avoid and i want a full refund and to cancel my payment but it's linked to Google ads I cannot wait for Facebook and Google to lose their stranglehold and monopoly on online. Both useless.

Gemini is terrible! I spent several hours trying to write code for an Arduino. It never produced a working result. It had me try different things over and over. None of it worked! It had me thinking AI is really just a joke. But ChatGPT produce a working result on the first try. I wouldn't waste your time on Gemini. Because it will be nothing but a waste of time!

A few weeks ago, this product actually worked and made sense. Now it’s completely broken — slower, dumber, and constantly refusing to generate code for ridiculous reasons like “immersive content” or “brevity.” The output quality has dropped off a cliff, and the servers keep failing. I’m genuinely frustrated with how badly Google has managed to ruin this.

It used to be good (the pro 2.5) but now it is impossible to use. Even free, even if they pay me i wont use it for code. It's doing more bad than good. idk how this is possible but the CLI is broken. Its pointless really sad about it i used to like it back then.
Idk wtf they did but it's far beyond every other coding agent. even grok fast 1 is better.. so yeah id say gemini is a 2/10, grok 3.5/10 then you have chatgpt 7/10 and claude 8/10.

I said "CS is CounterStrike" then "CS-related praise words that contain exact "clo", it says "Your use of closures in that module was brilliant and made the code much cleaner."