Late 2025 brought some of the most significant AI-related announcements in Google’s history. The November update package revealed the direction the company is heading: from agentic solutions, through a new generation of Gemini models, to investments in infrastructure and education. This is the moment when AI stops being an add-on and becomes the foundation of the entire technology ecosystem.
Some of these solutions have already reached the broader market - Gemini 3 and Nano Banana Pro have rolled out to AI Mode in nearly 120 countries, Antigravity has received higher limits for Google AI Pro and Ultra subscribers, and early tests and controversies have shown just how much responsibility comes with the next generation of models.
Gemini 3 - the heart of Google’s next-generation products
The biggest November announcement was the arrival of Gemini 3 - the most powerful, most flexible, and most "agentic" AI system Google has created to date. The model is multimodal: it understands text, images, audio, video, and code, and it can plan and execute multi-step tasks.
Key capabilities of the new generation:
- deeper understanding of user context and intent,
- the ability to make decisions and plan actions,
- even tighter integration with Google products,
- context windows up to 1M tokens and stronger multilingual performance,
- Gemini 3 Deep Think mode for tasks that require very deep reasoning.
Gemini 3 is already available in the Gemini app and in AI Mode within Google Search. In preview, Gemini 3 Pro is accessible across Google’s ecosystem - in the Gemini app, AI Mode in Search, Google AI Studio, Gemini CLI, and other developer tools.
Google has also expanded Gemini 3 availability in AI Mode: the model launched in English across nearly 120 countries and territories, and Google AI Pro and Ultra subscribers can switch in AI Mode to "Thinking with 3 Pro" to use the most powerful, more analytical version of the model.
AI Mode itself has become an interactive space where an answer is not just text, but a dynamic layout with visual modules, tools, and simulations generated live for a specific query.
In practice, this signals a revolution for both users and content creators. SERPs are gradually transforming from lists of links into environments of generated answers - and content must be prepared so AI models can understand it, verify it, and cite it effectively (from structure and clarity to reliable sources).
Nano Banana Pro - studio-quality images (and the first controversies)
Another major announcement was the debut of Nano Banana Pro. This is an image generation and editing tool that puts far more emphasis on consistency, detail, and professional quality than earlier systems. Nano Banana Pro is built on Gemini 3 Pro - it leverages its reasoning and world knowledge, allowing it to create visuals that are strongly grounded in context.
The model can generate:
- realistic scenes with a high level of detail,
- precise product shots and packshots,
- marketing graphics, infographics, and data visualizations,
- designs that look like the work of a photographer or graphic designer (including polished text on images).
In official documentation, Google emphasizes that Nano Banana Pro is great for turning notes and sketches into diagrams, UI mockups, or product prototypes.
For marketing and creative teams, this is a real shift. Generating commercial-quality visuals becomes faster and cheaper, while still allowing brand consistency. What’s more, Nano Banana Pro is gradually rolling into AI Mode in Google Search - announced in November, and in early December Google began expanding access in additional countries for Google AI Pro and Ultra subscribers, in English.
It’s worth noting, however, that the first weeks after launch brought not only excitement but also criticism. Media outlets pointed out that Nano Banana Pro can generate images that reinforce so-called "white saviour bias". In addition, journalistic testing showed that Nano Banana Pro can produce highly realistic yet controversial scenes, while moderation safeguards were still too weak.
All of this makes Nano Banana Pro both a huge opportunity for marketing and creative production and a strong warning signal: such a high level of realism and contextual accuracy requires a very deliberate usage policy, verification, and clear labeling of content.
Antigravity - a platform for building AI agents
Google also announced Antigravity - a platform designed for building AI agents. This is no longer just about generating code or offering hints to developers. The idea behind Antigravity is to create agents that can:
- analyze user goals,
- choose the right tools,
- execute tasks autonomously (including within an IDE, terminal, and browser),
- monitor progress and improve the process.
The platform is designed as an "agent-first development experience": the agent has its own workspace, access to an editor, terminal, browser, and models for controlling the computer (including Gemini 2.5 Computer Use) and image models (including the earlier Nano Banana - Gemini 2.5 Image).
This is a major step toward software that does not merely respond to commands, but can independently manage processes. For businesses, this could unlock massive automation potential - from document creation and project support to customer service and maintenance of internal systems.
AI in Google Maps and Android Auto - the era of voice-first travel
November also introduced the first Gemini features in Google Maps and Android Auto. This is a strong signal that AI is moving into areas previously dominated by simple navigation algorithms.
Google Maps
New features include:
- more conversational navigation,
- natural-language place search ("show me good cafes on the way to..."),
- voice reporting of road issues,
- landmark-based navigation ("turn after restaurant X").
Landmark-based navigation is rolling out gradually on Android and iOS in the US, and Gemini in Google Maps navigation is expected to be available wherever Gemini itself is available.
Android Auto
Gemini also supports drivers while on the road:
- adding stops along the route,
- sending messages,
- accessing email,
- creating playlists,
- helping with creative tasks (for example a quick brainstorming session while driving).
AI becomes a personal assistant, not just an extra layer on navigation - especially when combined with what is happening in parallel in the Gemini app (agentic task planning, list creation, and information organization).
AI for science and climate - SIMA 2, WeatherNext 2, AlphaFold
Google emphasizes that AI is not only developing consumer products, but also advancing science.
SIMA 2
SIMA 2 is a next-generation agent capable of operating in virtual environments, described directly as an important step toward AGI. Through integration with Gemini, the agent:
- understands goals and can "think" about them,
- holds conversations with users,
- learns from experience,
- behaves like an interactive companion in games and simulations, not just a command executor.
WeatherNext 2
WeatherNext 2 is an improved weather forecasting model that generates forecasts up to 8x faster with resolution down to one hour. It is already used by meteorological agencies for decision-making, which matters for risk management related to extreme weather events.
AlphaFold
Google also highlights the impact that solving the protein structure prediction problem had on biology and medicine. Five years have passed since the AlphaFold 2 breakthrough - work that received the 2024 Nobel Prize in Chemistry. AlphaFold remains one of the most important achievements in AI-assisted science.
AI in everyday life - shopping, travel, education
Travel planning
AI Mode introduced interactive "Canvases" that create travel plans tailored to user preferences - from trip ideas and place recommendations to turning plans into real bookings. You describe the travel style, budget, and activities, and the system groups suggestions, enables comparisons, and lets you adjust the plan live.
Shopping with agents
New Gemini features enable:
- describing needs in natural language,
- receiving results with prices, photos, reviews, and availability information,
- price tracking and automated purchasing when conditions are met,
- having an agent call a store to check availability.
NotebookLM and Gemini Live
Educational features were also introduced, such as:
- generating quizzes and flashcards from your own materials (NotebookLM),
- personalized lessons using notes, articles, and recordings,
- natural voice conversations with AI (Gemini Live) with adjustable speaking pace,
- foreign language practice and job interview simulations.
What do these updates mean for 2026?
- AI Mode and Gemini 3 will revolutionize search. Generated answers will replace a large portion of classic results, and SERPs will become interactive panels with visualizations, tools, and agents.
- Content will need to be written "for AI models", not only for people. Quality, clarity, structure, citability, and reliable sources will become requirements for being included in generated answers.
- Agentic behavior will become the standard. Companies will begin using their own mini-agents to handle processes - from development in Antigravity, through internal workflows, to customer support and back-office tasks.
- Pro-level AI-generated images will become the norm. Creative production will be faster and cheaper, but the risks of visual misinformation and bias reinforcement will increase.
- Mobility, shopping, and education will change through AI. Many manual tasks will disappear, and technology use will increasingly rely on talking to an agent rather than clicking through interfaces.
Bibliography:
Google. (2025). The latest AI updates announced in November. Retrieved from: https://blog.google/technology/ai/google-ai-updates-november-2025/
Google. (2025). Introducing Gemini 3. Retrieved from: https://blog.google/products/gemini/gemini-3/
Google. (2025). 15 ways to use Gemini 3. Retrieved from: https://blog.google/products/gemini/gemini-3-examples-demos/
Google DeepMind. (2025). Introducing Nano Banana Pro. Retrieved from: https://blog.google/technology/ai/nano-banana-pro/
Google. (2025). Gemini 3 and Nano Banana Pro in AI Mode - now in more countries. Retrieved from: https://blog.google/products/search/gemini-3-ai-mode-more-countries/
Hern, A. (2025). Google’s Nano Banana Pro image generator accused of reinforcing “white saviour” stereotype. The Guardian. Retrieved from: https://www.theguardian.com/technology/2025/dec/04/google-ai-nano-banana-pro-racialised-white-saviour-images
