• Towards AGI
  • Posts
  • Is It The End Of Google Search For Product Discovery?

Is It The End Of Google Search For Product Discovery?

1 in 4 Shoppers Choose ChatGPT for Product Discovery

Author

Shen Pandi
August 22, 2025

Here is what’s new in the AI world.

AI news: GenAI Goes Mainstream

Tech Crux: Don’t use Google’s Imagen 4 until you see this!

Open AI: Open-Sources AI Model to Challenge Market Leaders

Hot Tea: IBM and NASA's AI Aims to Predict Solar Storms' Effects

Towards MCP: Pioneering Secure Collaboration in the Age of AI & Privacy

We are excited to announce a powerful new collaboration with Towards MCP, a cutting-edge Model Context Protocol platform that is redefining how businesses integrate intelligence into their operations.

By integrating with Towards MCP, we can now offer our clients an unparalleled ability to unlock insights from their existing data ecosystems, making intelligent data interaction a standard, accessible reality for every enterprise.

Nearly 60% of American Shoppers Now Use GenAI, 1 in 4 Prefer It Over Google

A recent study reveals that generative AI is rapidly transforming how Americans shop online, with adoption climbing and skepticism easing, though significant trust hurdles remain.

Widespread Adoption with Lingering Caution

According to an Omnisend survey of 1,224 U.S. consumers, 59% already use generative AI tools for shopping-related tasks. Perhaps more notably, reluctance to let AI handle transactions has plummeted: just 32% now express hesitation, down from 66% in February 2025.

Despite this growing acceptance, a strong undercurrent of concern persists. 85% of respondents reported worries about privacy, personalization errors, and overall AI fatigue.

ChatGPT Emerges as Preferred Shopping Assistant

When using AI during online shopping, consumers most frequently turn to these tools for:

  • Product research (57%)

  • Personalized recommendations (45%)

  • Finding deals (40%)

Among the various AI platforms available, ChatGPT is the clear favorite, used by 65% of AI-assisted shoppers. In fact, one in four believe it provides better product recommendations than Google.

❝

“Googling often means ads, SEO content, and dozens of open tabs, Generative AI, by contrast, offers a distilled answer, like getting advice from a knowledgeable friend.”

explains Marty Bauer, E-commerce Expert at Omnisend.

Checkout Automation: Acceptance Grows, But Trust Is Key

Readiness to let AI complete purchases has nearly doubled in five months. This shift aligns with industry moves toward AI-driven commerce, including reports that OpenAI is developing checkout features for ChatGPT.

Yet barriers to full trust remain prominent. Shoppers’ top concerns include -

  • Privacy and data security (43%)

  • AI misinterpreting preferences (37%)

  • Irrelevant recommendations (35%)

  • Overuse of AI in e-commerce (26%)

❝

“Convenience can quickly outweigh fear when the experience is smooth, but trust will define AI’s next chapter in online retail.”

says Bauer.

Recommendations for E-commerce Brands

As major players like Amazon, Walmart, and Mastercard advance AI shopping tools, Bauer advises brands to focus on transparency and user control -

  • Give shoppers clear opt-in/out options and be transparent about data usage.

  • Require explicit approval before AI performs actions like adding items to a cart or checking out.

  • Maintain easy access to human support, especially for high-value or personal purchases.

The survey was conducted by Cint in July 2025 with a nationally representative sample of U.S. consumers.

Imagen 4 Is Awesome! But Comes With Some Downsides

From the early days of Dall-E Mini to more recent tools like HiDream, you’ve watched AI image generators evolve from producing blurry, abstract visuals into sophisticated assistants for artists and designers.

Now, with the release of Google’s Imagen 4, it feels like we’ve entered a new era, one where AI doesn’t just assist, but truly collaborates.

Whether you're a digital artist, marketer, educator, or hobbyist, Imagen 4 brings a level of photorealism, stylistic range, and prompt understanding that stands clearly above today’s alternatives like GPT-4o and Reve Halfmoon.

Things I love about it

1) Unmatched Versatility and Quality

You can guide Imagen 4 through almost any visual style, photorealistic portraits, retro pixel art, watercolor landscapes, even MS Paint-like doodles, and it delivers with striking accuracy.

Its ability to render text within images is particularly impressive, a task many other models still struggle with.

2) Superior Prompt Adherence

Your words matter here. Imagen 4 doesn’t just “get the idea”, it follows instructions closely, interpreting nuance and detail in a way that feels almost intuitive. You won’t need to battle with regenerations or workaround prompts as often.

3) Rich, Vibrant Outputs

Unlike GPT-4o’s sometimes muted or washed-out color palette, Imagen 4 produces images with depth, contrast, and visual punch. This isn’t just tech, it’s art.

Things I hate about it

While Imagen 4 is a powerful tool, it’s crucial to understand its limitations and quirks to avoid frustration and integrate it effectively into your workflow. Here’s what you should be aware of:

1) The Mandatory Watermark

Every single image generated by Imagen 4 comes with an embedded SynthID watermark. This isn't an option you can toggle off. If your project requires completely "clean," unmarked files for commercial or aesthetic reasons, you have two choices,

Use an approved post-processing tool to remove it (and ensure this complies with Google's terms of service) or select a different AI model entirely. Always check the usage policy before attempting to alter any generated content.

2) Strict and Sometimes Unpredictable Filters

The model operates behind a robust and strict content safety filter. It will automatically block prompts related to,

  • Public figures and celebrities

  • Graphic violence or hateful imagery

  • Self-harm and sensitive medical topics

  • Sexual content

The challenge is that these filters can sometimes be overly broad. You might find that a perfectly innocuous prompt gets blocked because it accidentally triggers a banned phrase.

Your best strategy here is to creatively rephrase your request while maintaining the same core intent. Don’t waste time trying to outsmart or argue with the filter; it’s a fixed system.

3) The Speed vs. Quality (and Cost) Trade-off

Imagen 4 offers different generation tiers, and choosing the right one is a key workflow decision.

  • Fast: Use this while you’re brainstorming and refining your prompt. It’s quicker and consumes fewer credits, perfect for iteration.

  • Ultra: Switch to this once your prompt is perfected. It’s slower and more expensive, but delivers the highest quality output for your final product.

  • Standard: This is the middle-ground option, a budget-friendly compromise on quality if Ultra is too costly.

You’ll need to manage your credits and time based on which stage of the creative process you’re in.

The Verdict

Imagen 4 isn’t just an upgrade, it’s a leap forward. With its combination of style range, prompt intelligence, and sheer visual quality, it sets a new bar for what generative AI can do.

If you’ve been waiting for an AI image tool that feels less like a gadget and more like a creative partner, Imagen 4 is worth your attention.

ByteDance Open-Sources New AI Model to Rival DeepSeek and Alibaba

Chinese social media giant ByteDance has launched a new open-source artificial intelligence model, Seed-OSS-36B, positioning it as a competitive alternative to offerings from leading Chinese and global AI developers like Alibaba Cloud and DeepSeek.

Key Features of the New Model


The model, released on Hugging Face by ByteDance’s Seed AI research department, comes in three variants and is designed for both general-purpose use and advanced reasoning.

It supports long-context processing and includes developer-friendly features aimed at broadening its adoption.

Despite its relatively compact size of 36 billion parameters, significantly smaller than many competing open-source models, ByteDance claims Seed-OSS-36B delivers performance that matches or exceeds similarly sized models from rivals such as Alibaba Cloud, Google, and OpenAI.

China’s Growing Open-Source AI Influence


The release underscores how Chinese tech firms are increasingly leveraging open-source strategies to compete on the global AI stage. By making AI code publicly accessible, companies like ByteDance enable third-party developers to adapt and enhance the technology, accelerating innovation and adoption.

ByteDance’s move follows earlier, more restricted releases like its proprietary Doubao LLM in May 2023. The new open model also aligns with CEO Liang Rubo’s push for faster response to AI market trends after internal criticism of the company’s pace.

Global Open-Source Rankings


Chinese models have recently dominated open-source AI performance rankings. As of July, platforms like Moonshot AI’s Kimi K2, Alibaba Cloud’s Qwen 3, and DeepSeek’s R1 outperformed offerings from Google and Meta, according to benchmarking by LMArena, a platform backed by UC Berkeley researchers.

ByteDance’s entry into this competitive space signals its ambition to not only keep pace with domestic leaders but also contest the broader international open-source AI ecosystem.

IBM & NASA Collaborate on New Open-Source AI to Forecast Solar Storms

IBM and NASA have jointly released Surya, a new open-source foundation model designed to forecast solar activity and its potential impacts on Earth and space-based technology.

The model is now publicly available on the AI platform Hugging Face, making advanced space-weather prediction tools more accessible to researchers and organizations worldwide.

Key Features and Capabilities

  • Surya is the first heliophysics AI foundation model trained on high-resolution solar observation data.

  • Early testing shows a 16% improvement in accuracy for classifying solar flares compared to existing methods.

  • The model aims to help experts better anticipate solar storms, which can disrupt satellites, damage electronic systems, and even affect airline operations.

A "Weather Forecast for Space"


Juan Bernabe-Moreno, Director of IBM Research Europe, UK and Ireland, compared Surya to terrestrial weather forecasting, emphasizing its role in preparing for hazardous solar events.

❝

He noted that the model represents both a significant technological achievement and a critical tool for safeguarding modern infrastructure against solar disruptions.


This collaboration builds on previous joint initiatives between IBM and NASA, including:

  • The 2023 release of watzonx.ai, a geospatial foundation model based on NASA satellite data.

  • Last year’s launch of Prithvi, an AI model focused on weather and climate projections.

Surya marks another step in using open-source AI and public data to address global environmental and technological challenges.

Legacy silos slow you down. Intelligent migration fuels growth.

Stop Guessing. Start Knowing.


DataManagement.AI turns your chaotic data into your most powerful asset. Unlock real-time insights, slash costs by 50%, and boost team productivity 20x.

See your data differently.

Take the first step!

Your opinion matters!

Hope you loved reading our piece of newsletter as much as we had fun writing it. 

Share your experience and feedback with us below, ‘cause we take your critique very critically. 

How did you like our today's edition?

Login or Subscribe to participate in polls.

Thank you for reading

-Shen & Towards AGI team