I switched to a new Android voice assistant, and Google Gemini now feels outdated

When Google announced that he would replace the assistant from the twins, I was preparing for the fact that, as I expected, it would be inevitable: a rich functioned experience, overshadowed by the disappointing period of the transition. However, almost a year later, the transition actually became smoother than I expected. Installation of reminders and management of Smart Home devices now works, and I was not completely lost without the old assistant Google. But while the Gemini was not a disaster that I was afraid of, I can’t say that this is a clear update.

Despite all Google’s promises about deeper reasoning and better answers, I feel that there are very few, that the twins can do, which the assistant could not. Therefore, when I found out that the startup of AI intensively intensified to offer an alternative, I was immediately curious to see how he was compared to the old and new assistants to Google. And after a week of use, I can say with confidence that in the near future I will not return to the twins.

As I mentioned earlier, numerous Gemini extensions mean that now he can control my smart home and make notes. Nevertheless, this fails to answer the questions as the outgoing assistant Google with a bang. My main grip with the twins is connected with how much it relies on the main large language model (and his personal knowledge) to respond to my requests instead of using his built -in search engine.

An indicative example, I usually ask twins about promotions and get variable results that upset me to the end. Invitation, how “how is the prices of Apple shares?” After over the past few months, a vague response over the financial indicators of the company lost, but, most importantly, it did not provide a price at all.

Only when I specifically asked: “What is the price of Apple shares?” This twins finally gave me the final answer. And even then this connected me with three other websites and gave lectures on the volatility of the prices for shares.

Meanwhile, perplexity responds to such requests with a rich information box, which even contains an interactive diagram. Look at yourself.

This is not one -time, I came across the same problem, trying to transform currencies many times. Ask the twins how many US dollars in the euro, and you will most likely get a rejection of responsibility for fluctuations in the exchange rate than the actual number. If the stars are leveled, you can get an answer, but instead it can be faster to use Google’s search.

On the other hand, bewilderment does exactly what I want from the assistant: he receives real -time exchange rates from several sources and clearly represents them, although without beautiful graphs this time. More importantly, he performs this task sequentially each time.

Then I asked both assistant: “Where can I see the reprint of the interstellar this weekend?” Here Gemini did exactly what I expected from the decade of Google Assistant: he refused to search for an answer. This is a rather awkward result, if not more.

Meanwhile, embarrassment completed a quick search and analyzed eight sources to offer the release date and recommendations on how to book tickets. From here I just used the Google circle to search and independently move along one of the web reservation sites. Is the puzzling reaction perfect? Not at all, but this is a big step forward compared to the twins.

While most competing digital assistants on Android stop when answering questions, this is not so bewilderment. Like Gemini, you can ask him to perform several tasks related to the phone, such as sending a text message or email, call contact and open the application. He can also play music through Spotify and set reminders.

Impressive, embarrassment can understand vague instructions, such as “remind me to see the white lotus in the next season when it is released.” Here you expect that the assistant will first be distorted by the date of release of the show and establish a reminder accordingly, and this is exactly what makes bewilderment.

When I tried the same thing with the twins, I was horrified, finding that she simply set the task with all the text, instead of following the instructions at the end. This is the type of inconsistency that makes the experience of AI Google so disappointing as to live and potentially dangerous if you rely on this for something important.

Embarrassment can also perform several tasks at the same time, similarly to the functionality of the Multi-Extension Gemini, which Samsung showed at the recent Galaxy S25 launch event. For example, he can find nearby Thai restaurants and remind a reminder to book a table later.

On this note, bewilderment has several cool tricks in the sleeve. On the one hand, you can ask him to book the table, and he will bring you to the restaurant Opentable page to configure the reservation. In the same way, you can book Uber’s trip using only my voice – I checked it with the help of a quick “call me to the airport”, and she worked as expected. Perplexity also has an amazingly good pen in localized sentences. He managed to give me instructions to the nearest coffee shop (in this non-chain), and booking the cab in the place was just another voice team.

Finally, bewilderment can also read the contents of your screen when necessary. This means that you can indicate it to do things such as summing up the contents of your screen, and then send the result as an electronic letter to a friend. Or, since I am inclined to use it most, translate a foreign language. Gemini offered this function for some time, so I am glad that I did not lose it when I switched to an assistant in embarrassment.

Confusion may have an advantage in some areas, but this is not entirely equal to the Gemini. Unlike bewilderment, the Google assistant can read your emails, access to files from the disk and more smoothly integrate with services such as tasks and assistants. The closest bewilderment can get to read your mailbox – this is the summation of your notifications, which is useful, but far from so powerful.

Another key difference is how you call on each assistant. Gemini support the word for activation without hands, while bewilderment requires holding the power button or using a diagonal gesture. I personally do not mind this, because I usually keep the phone when I need help, but, nevertheless, this is a noticeable restriction. At the moment, perplexity is not enough for any kind of management of a smart home, although this may not be a transaction breaker if you already have smart speakers.

The only missing part that can be more difficult to ignore is the integration of the calendar. At the moment, perplexity cannot be connected with the calendar, which means that you cannot check or add events. However, over time, this is an easily solved problem, since many other applications allow you to bind your Google account with the calendar and access to the disk.

With these shortcomings should you switch to bewilderment? It all comes down to what you need from the digital assistant of your phone. If your priority is reliably obtaining answers to questions, regardless of how vaguely formulated, bewilderment is an excellent choice without a doubt. But if you need your assistant for direct interaction with your digital life, Gemini is currently holding an advantage.

At the moment, I made my switch, and I do not regret it. Gemini can significantly improve in the future, but bewilderment does everything I need, and then some. More than anything, I am just glad that there is real competition in this space. In the end, only a few years ago, the stagnant assistant Google was the only viable option.