Author: admin

Google Translate is getting an AI upgrade designed to make real-time conversations and language learning easier. The update is rolling out this week in the Translate app on iOS and Android.

The first tool, for live translations, lets you have a back-and-forth conversation with someone by surfacing audio and text translations as you speak, so you can easily follow along. Advanced Gemini models allow for support of more than 70 languages, including Arabic, French, Hindi, Korean and Spanish. Google says its voice and speech recognition models are trained to isolate sounds, so the live translation feature should also work in noisy environments like an airport or a cafe.

I gave the Arabic translation a spin, and it did a good job picking up on even unconventional Iraqi slang. Judging from my intermediate knowledge of Spanish, it also translated my rambling well. I also like that Translate offers both text and audio translations, so you can go with whichever medium you prefer or revisit what was said via the on-screen text.  

A second feature is designed to help you practice a new language. You can choose whether you currently have a basic, intermediate or advanced understanding of the language you’re learning and then set a goal. For example, I noted that I wanted to practice Spanish so I could get around the city and have casual conversations (hearkening to my days at Mobile World Congress in Barcelona when I ambitiously dusted off my high school Spanish). 

Translate will then surface recommended scenarios like asking for the nearest bus stop, greeting a neighbor or chatting about your hobbies. You can then choose either a listening or speaking exercise. In the listening sessions, you’ll tap the words you hear, and in the speaking one, you can practice having a back-and-forth conversation. The language learning feature is currently available for English speakers practicing Spanish and French, and for Spanish, French and Portuguese speakers practicing English.

“These updates are made possible by advancements in AI and machine learning,” Google Product Manager Matt Sheets said in a blog post. “As we continue to push the boundaries of language processing and understanding, we are able to serve a wider range of languages and improve the quality and speed of translations. And with our Gemini models in Translate, we’ve been able to take huge strides in translation quality, multimodal translation, and text-to-speech capabilities.”

This comes as Google adds other language-specific features like Voice Translate on the Pixel 10 series. That feature can also translate what someone is saying in real time, but while chatting on the phone, and goes the extra mile of mimicking the sound of their voice, instead of superimposing a robotic one. Gemini Live can also have a back-and-forth conversation in a handful of languages. And last year, Google Translate added 110 new languages thanks to AI advancements.

How to access Google Translate’s new tools 

The live translate and language learning capabilities in Google Translate are rolling out now on iOS and Android. Live translations are currently limited to users in the US, India and Mexico. 

To access the new features, open up the Google Translate app, which you can download in the Apple App Store or Google Play. You’ll see a tab for “Live translate” on the bottom left, and another for “Practice,” the language learning tool, on the right.  

When you go into Live translate, you can pick which languages you want to translate and then start talking. The translated audio and text will automatically appear.

Tapping into Practice will prompt you to set up the language, level and goal you’re looking to meet, and then lead you to tailored exercises.

New research from BrightEdge shows that Google AI Overviews, AI Mode, and ChatGPT recommend different brands nearly 62% of the time. BrightEdge concludes that each AI search platform is interpreting the data in different ways, suggesting different ways of thinking about each AI platform.

Methodology And Results

BrightEdge’s analysis was conducted with its AI Catalyst tool, using tens of thousands of the same queries across ChatGPT, Google AI Overviews (AIO), and Google AI Mode. The research documented a 61.9% overall disagreement rate, with only 33.5% of queries showing the exact same brands in all three AI platforms.

Google AI Overviews averaged 6.02 brand mentions per query, compared to ChatGPT’s 2.37. Commercial intent search queries containing phrases like “buy,” “where,” or “deals” generated brand mentions 65% of the time across all platforms, suggesting that these kinds of high-intent keyword phrases continue to be reliable for ecommerce, just like in traditional search engines. Understandably, e-commerce and finance verticals achieved 40% or more brand-mention coverage across all three AI platforms.

Three Platforms Diverge

Not all was agreement between the three AI platforms in the study. Many identical queries led to very different brand recommendations depending on the AI platform.

BrightEdge shares that:

• ChatGPT cites trusted brands even when it’s not grounding on search data, indicating that it’s relying on LLM training data.
• Google AI Overviews cites brands 2.5 times more than ChatGPT.
• Google AI Mode cites brands less often than both ChatGPT and AIO.

The research indicates that ChatGPT favors trusted brands, Google AIO emphasizes breadth of coverage with more brand mentions per query, and Google AI Mode selectively recommends brands.

Next we untangle why these patterns exist.

Differences Exist

BrightEdge asserts that this split across the three platforms is not random. I agree that there are differences, but I disagree that “authority” has anything to do with it and offer an alternate explanation later on.

These are the conclusions that they draw from the data:

• “The Brand Authority Play:
  ChatGPT’s reliance on training data means established brands with strong historical presence can capture mentions without needing fresh citations. This creates an “authority dividend” that many brands don’t realize they’re already earning—or could be earning with the right positioning.
• The Volume Opportunity:
  Google AI Overview’s hunger for brand mentions means there are 6+ available slots per relevant query, with clear citation paths showing exactly how to earn visibility. While competitors focus on traditional SEO, innovative brands are reverse-engineering these citation networks.
• The Quality Threshold:
  Google AI Mode’s selectivity means fewer brands make the cut, but those that do benefit from heavy citation backing that reinforces their authority across the web.”

Not Authority – It’s About Training Data

BrightEdge refers to “authority signals” within ChatGPT’s underlying LLM. My opinion differs in regard to an LLM’s generated output, not retrieval-augmented responses that pull in live citations. I don’t think there are any signals in the sense of ranking-related signals. In my opinion, the LLM is simply reaching for the entity (brand) related to a topic.

What looks like “authority” to someone with their SEO glasses on is more likely about frequency, prominence, and contextual embedding strength.

• Frequency:
  How often the brand appears in the training data.
• Prominence:
  How central the brand is in those contexts (headline vs. footnote).
• Contextual Embedding Strength:
  How tightly the brand is associated with certain topics based on the model’s training data.

If a brand appears widely in appropriate contexts within the training data, then, in my opinion, it is more likely to be generated as a brand mention by the LLM, because this reflects patterns in the training data and not authority.

That said, I agree with BrightEdge that being authoritative is important, and that quality shouldn’t be minimized.

Patterns Emerge

The research data suggests that there are unique patterns across all three platforms that can behave as brand citation triggers. One pattern all three share is that keyword phrases with a high commercial intent generate brand mentions in nearly two-thirds of cases. Industries like e-commerce and finance achieve higher brand coverage, which, in my opinion, reflects the ability of all three platforms to accurately understand the strong commercial intents for keywords inherent to those two verticals.

A little sunshine in a partly cloudy publishing environment is the finding that comparison queries for “best” products generate 43% brand citations across all three AI platforms, again reflecting the ability of those platforms to understand user query contexts.

Citation Network Effect

BrightEdge has an interesting insight about creating presence in all three platforms that it calls a citation network effect. BrightEdge asserts that earning citations in one platform could influence visibility in the others.

They share:

“A well-crafted piece… could:
Earn authority mentions on ChatGPT through brand recognition

Generate 6+ competitive mentions on Google AI Overview through comprehensive coverage

Secure selective, heavily-cited placement on Google AI Mode through third-party validation

The citation network effect means that earning mentions on one platform often creates the validation needed for another. “

Optimizing For Traditional Search Remains

Nevertheless, I agree with BrightEdge that there’s a strategic opportunity in creating content that works across all three environments, and I would make it explicit that SEO, optimizing for traditional search, is the keystone upon which the entire strategy is crafted.

Traditional SEO is still the way to build visibility in AI search. BrightEdge’s data indicates that this is directly effective for AIO and has a more indirect effect for AI Mode and ChatGPT.

ChatGPT can cite brand names directly from training data and from live data. It also cites brands directly from the LLM, which suggests that generating strong brand visibility tied to specific products and services may be helpful, as that is what eventually makes it into the AI training data.

BrightEdge’s conclusion about the data leans heavily into the idea that AI is creating opportunities for businesses that build brand awareness in the topics they want to be surfaced in.
They share:

“We’re witnessing the emergence of AI-native brand discovery. With this fundamental shift, brand visibility is determined not by search rankings but by AI recommendation algorithms with distinct personalities and preferences.

The brands winning this transition aren’t necessarily the ones with the biggest SEO budgets or the most content. They’re the ones recognizing that AI disagreement creates more paths to visibility, not fewer.

As AI becomes the primary discovery mechanism across industries, understanding these platform-specific triggers isn’t optional—it’s the difference between capturing comprehensive brand visibility and watching competitors claim the opportunities you didn’t know existed.

The 62% disagreement gap isn’t breaking the system. It’s creating one—and smart brands are already learning to work it.”

BrightEdge’s report:

ChatGPT vs Google AI: 62% Brand Recommendation Disagreement

Featured Image by Shutterstock/MMD Creative

When it started, YouTube was considered a fun internet distraction. These days it’s seen as an irresistible force. The video-sharing platform’s importance to the revenue and roadmap of its parent has never been more significant. In 2024, ad sales hit $36.1 billion—more than 10% of the $350 billion Alphabet reported for the period. That’s almost as much as Disney, Paramount, NBCUniversal and Fox generated from ads combined.

But YouTube isn’t just a money-maker. The 20-year-old company sits atop a unique archive of 20 billion videos that can be used to train generative artificial intelligence applications, including those able to produce moving images. In June, CNBC reported that Alphabet’s other well-known brand—Google—has directly benefitted from this, though it says its AI endeavors are meant to benefit content creators rather than (as many of them suspect) replace them. In this Bloomberg Originals mini-documentary, we explain how YouTube evolved from scrappy startup to streaming behemoth, and how TikTok is battling it for the loyalty of young users.

Microsoft announced its first homegrown AI models on Thursday: MAI-Voice-1 AI and MAI-1-preview. The company says its new MAI-Voice-1 speech model can generate a minute’s worth of audio in under one second on just one GPU, while MAI-1-preview “offers a glimpse of future offerings inside Copilot.”

You can try MA1-Voice-1 out for yourself on Copilot Labs, where you can enter what you want the AI model to say, as well as change its voice and style of speaking. In addition to this model, Microsoft introduced MAI-1-preview, which it says it trained on around 15,000 Nvidia H100 GPUs. It’s built for users in need of an AI model capable of following instructions and “providing helpful responses to everyday queries.”

Microsoft AI chief Mustafa Suleyman said during an episode of Decoder last year that the company’s internal AI models aren’t focused on enterprise use cases. “My logic is that we have to create something that works extremely well for the consumer and really optimize for our use case,” Suleyman said. “So, we have vast amounts of very predictive and very useful data on the ad side, on consumer telemetry, and so on. My focus is on building models that really work for the consumer companion.”

The company plans on rolling out MAI-1-preview for certain text use cases in its Copilot AI assistant, which currently relies on OpenAI’s large language models. It has also started publicly testing its MAI-1-preview model on the AI benchmarking platform LMArena.

“We have big ambitions for where we go next,” Microsoft writes in the blog post. “Not only will we pursue further advances here, but we believe that orchestrating a range of specialized models serving different user intents and use cases will unlock immense value.”