Have you ever wondered why Google is a $2 trillion behemoth while Wikipedia is a nonprofit? The answer is simple: the magic of commercial search. When you search for "how many protons does a cesium atom have?", Google makes nothing. But when you search for "best tennis racket," it starts printing money. This asymmetry defines the entire search economy. Now, with the rise of AI, this balance is being radically altered.
I recently read an in-depth analysis by a16z partners Justine Moore and Alex Rampell. Their insights into how AI is reshaping the e-commerce landscape were deeply striking. They not only analyzed the potential threats facing Google but, more importantly, painted a completely new picture of e-commerce in the AI era. In this picture, the traditional search-compare-buy model is being replaced by intelligent purchasing experiences driven by AI agents. I've spent considerable time pondering their insights, combined with my own observations of the industry, and would like to share some deeper thoughts.
Google's real crisis: not search volume, but value migration
Justine's article made a striking point: Google could lose 95% of its search volume and still see revenue growth, as long as it retains those commercially valuable queries. This idea sounds counterintuitive, but it actually reveals a core secret of the search economy. After further reflection, I realized that this underlies a deeper issue: AI is changing the location of value creation.
Traditionally, Google plays the role of an information intermediary. Users have purchasing intent, Google provides search results and ads, merchants gain traffic, and Google collects advertising fees. This is a relatively simple three-way game. However, the emergence of AI agents has disrupted this balance. When ChatGPT or Perplexity can directly answer the question "What is the best tennis racket?" and provide specific recommendations, why would users still click on Google's ad links?
More importantly, AI isn't just answering questions; it's redefining search itself. Our previous search process involved: asking a question → getting a list of links → clicking to view → comparing information → making a decision. AI agents, on the other hand, follow this process: describing a need → getting recommendations → purchasing directly. The comparison and research steps in between have been significantly reduced or even eliminated. This means traditional search engines have not only lost search volume but also their crucial position in the decision-making process.
Clues can be seen in testimony given by Apple Senior Vice President Eddy Cue at the DOJ antitrust trial in May 2025. He stated that Safari search volume had declined for the first time in more than two decades. This news immediately caused Alphabet's stock price to drop nearly 8% in a single day, wiping out over $150 billion in market capitalization. Although Google's Q2 earnings report showed continued growth in search revenue, suggesting that the current loss is primarily due to low-value queries, the direction of the trend is clear.
I believe Google faces more than a simple competitive threat, but rather a structural challenge to its business model. When AI can directly handle the entire process from intent recognition to purchase decision-making, the traditional "traffic → advertising → conversion" model will become inefficient or even obsolete. What Google needs isn't a better search algorithm, but a fundamentally new business model to adapt to AI-driven consumer behavior.
AI-powered transformation of five purchasing behaviors: from impulsive to thoughtful
In her article, Justine categorizes purchasing behavior into five categories, ranging from impulse buys to major life purchases. Each category will undergo varying degrees of change in the AI era. I find this framework quite accurate, but I'd like to delve deeper into the psychological mechanisms behind each purchase behavior and how AI is reshaping them.
Impulse buying might seem like the area least impacted by AI, as impulse buying implies a lack of rational research. However, I believe this assessment is superficial. The true power of AI lies in predicting and guiding impulses. Imagine that when you see a funny T-shirt on TikTok, AI has already analyzed your browsing history, purchase history, social media activity, and even your emotional state, then precisely recommends the product that best meets your current needs. This isn't a simple algorithmic recommendation; it's a deep understanding and manipulation of human impulse psychology. I believe this personalized impulse guidance has the potential to make impulse buying more frequent and accurate.
The AI transformation of everyday essentials is the easiest to understand and implement. However, I've observed an interesting phenomenon: when AI begins to influence our daily purchasing decisions, our consumption habits may undergo subtle shifts. For example, AI might adjust the timing and quantity of your purchases based on price fluctuations, inventory levels, and even weather forecasts. A clever AI agent might discover a discount on a certain brand of laundry detergent a week before you're running low, purchase it in advance, and recommend you try it. This "smart arbitrage" behavior could potentially allow consumers to unknowingly achieve better value for money while also forcing brands to rethink their pricing and promotional strategies.
Lifestyle purchases are where I believe AI will have the greatest impact. These purchases are characterized by a certain price threshold, personal taste, and a certain level of research. Justine mentioned products like Plush, but I think this is just the tip of the iceberg. The real revolution will come from AI's deep learning of personal style and preferences. Imagine an AI assistant that not only knows your past purchases but also understands your body type, skin tone, lifestyle, social circles, and even your aspirations. It can recommend not just individual products, but entire outfits and even lifestyle upgrades. This level of personalization is unattainable through traditional e-commerce platforms.
Functional purchases are the most complex and challenging to implement with AI. These purchases typically involve significant expenditures and long-term use, and consumers need not only product recommendations but also expert advice. I believe a new category of AI applications will emerge here: AI advisors. These AIs not only possess extensive product knowledge but also engage in in-depth conversations similar to those of human sales experts. They can inquire about your specific needs, usage scenarios, budget constraints, and even your future plans, and then provide highly personalized recommendations. Crucially, these AI advisors are cross-brand and won't favor specific products due to commissions or inventory constraints.
Major life purchases are perhaps the area where AI has the least impact, yet also the most important. Decisions like buying a house, getting married, and getting an education are too significant and personal to be completely delegated to AI. However, AI can play a vital role in information gathering, option comparison, and risk assessment. The AI coach I envision isn't meant to make decisions for you, but to help you make better ones. It can sort through vast amounts of information, identify potential pitfalls, simulate the long-term consequences of different options, and even assist in contract negotiations. I believe the value of this kind of AI coach lies in its neutrality and comprehensiveness, unlike human advisors who may have conflicts of interest.
Amazon and Shopify's Moats: Dual Advantages of Data and Infrastructure
Justine pointed out in her analysis that Amazon and Shopify have stronger defensive capabilities than Google. I completely agree with this view, but I would like to delve deeper into the source and sustainability of this advantage. Amazon's advantage lies not only in its control of the entire supply chain from search to delivery, but more importantly, in its possession of the most valuable behavioral data.
Amazon knows what you bought, when you bought it, how quickly it arrived, whether you returned it, whether you repurchased it, and so on. This data is far more valuable than search history because it directly reflects actual purchasing behavior and satisfaction. When AI agents need to make purchasing decisions for users, this data is the most valuable training material. While Google knows what you searched for, it doesn't know what you ultimately bought, let alone whether you were satisfied with the purchase. This data gap will be further exacerbated in the AI era.
More importantly, the Amazon Prime loyalty program creates a unique economic phenomenon: sunk cost bias. When you've paid for a Prime membership, you tend to buy more on Amazon to recoup your investment. This psychological mechanism is likely to become even more powerful in the age of AI. When an AI agent searches for the best purchase option for you, it may naturally gravitate towards Amazon, knowing that you're a Prime member and receive free shipping and other benefits.
Shopify's defensive logic is completely different, yet equally powerful. It builds its moat not by controlling consumers but by creating network effects through empowering merchants. As more and more D2C (Direct-to-Consumer) brands choose Shopify, the platform becomes increasingly irreplaceable. In the age of AI, the advantages of this decentralization are likely to become even more pronounced. An AI agent may need to simultaneously access information and complete purchases from hundreds of different brand websites. If all of these websites run on Shopify, a standardized API ecosystem will emerge.
I believe Shopify has another underappreciated advantage: its proximity to brand stories. In the AI era, while functional differences in products can be quickly identified and compared by AI, the emotional connection to a brand still requires human experience. Brands on Shopify often have unique stories and cultures. These soft values are difficult for AI to fully quantify, but they are crucial factors influencing consumer decisions.
Four major infrastructure challenges for AI commercialization
At the end of the article, Justine mentioned the four basic conditions needed for AI to realize its full potential in the business field. I think each one is worth exploring in depth because they are not only technical challenges but also opportunities for business model innovation.
First, there's the issue of better data. Current product review systems do suffer from serious problems: fake reviews, polarization, and a lack of context. But I believe the root of the problem lies in a misaligned incentive system. Consumers typically write reviews based on either extreme satisfaction or extreme dissatisfaction, with in-between states rarely recorded. Furthermore, existing review systems fail to capture product usage scenarios, user expectations, and changes over time.
My ideal data system looks like this: AI agents not only collect subjective user reviews but also monitor actual product usage through IoT devices. For example, a smartwatch wouldn't just rely on five-star reviews, but also on how often and how long users actually wear it. Reviews of a coffee machine wouldn't just rely on text feedback, but also on actual usage frequency, cleaning, and maintenance. Only by combining this objective usage data with subjective feedback can a truly valuable product evaluation system be formed.
The challenge of unifying APIs is more political than technical. Each e-commerce platform has its own API structure, data format, and authentication mechanism. These differences are largely intentional, intended to create platform lock-in. However, in the era of AI agents, this fragmentation could become an efficiency bottleneck for the entire industry. I predict the emergence of specialized API aggregation services, similar to the global distribution systems in the travel industry. These services will standardize interfaces across different platforms, allowing AI agents to seamlessly compare and purchase across platforms.
Identity and memory are the most complex challenges, as they involve balancing privacy, accuracy, and adaptability. I believe future AI shopping assistants will need to build a multi-layered preference model. This model should not only record your past purchases but also understand your values, life stage, financial constraints, and more. For example, it needs to know that you prioritize convenience for weekday lunches but prioritize quality and presentation for weekend get-togethers. This kind of context-aware recommendation requires AI to possess near-human social understanding capabilities.
Embedded data capture may hold the greatest potential for innovation. Traditional data collection is passive and delayed: reviews after purchase, feedback after use. However, AI agents can learn preferences in real time. For example, if you spend a long time focused on a certain feature while browsing a product, the AI can infer that you are interested in that feature. If you quickly skip certain color options, the AI can learn your color preferences. This kind of micro-interaction analysis allows the AI to develop a more nuanced understanding of your preferences.
Reshuffle of e-commerce platforms: Who will win?
After considering Justine's analysis, I've formed my own perspective on the future of e-commerce. I believe AI will trigger a new platform reshuffle, but the winning strategy will be different than before.
In the traditional e-commerce era, competition revolved primarily around three dimensions: selection, convenience, and price. Amazon won in selection with its "Everything Store" concept, while also building an advantage in convenience through Prime. However, in the AI era, the importance of these advantages will shift.
When AI agents can automatically compare prices across the entire network and act as agents for purchases, the price advantage of a single platform will be diluted. When AI enables intelligent batch processing and cross-platform fulfillment, the definition of convenience will also change. The true competitive advantage will shift to data quality, AI capabilities, and ecosystem integration.
I predict the emergence of several new types of platform players: AI-native e-commerce platforms, vertical AI agents, and commercial infrastructure providers. AI-native platforms will be designed from the ground up with the needs of AI agents at their core, providing structured product data, standardized APIs, and an AI-friendly user experience. Vertical AI agents will focus on specific categories, such as fashion AI, digital product AI, or home improvement AI, establishing competitive advantages through deep specialization. Commercial infrastructure providers will provide underlying technical services to help traditional e-commerce platforms integrate AI.
I also believe a new business model will emerge: AI agent subscriptions. Instead of shopping directly on e-commerce platforms, consumers may subscribe to one or more AI shopping agents, which will then make all purchasing decisions on their behalf. These agents will charge subscription fees rather than commissions, thus avoiding conflicts of interest and truly embracing the consumer's perspective. This model has the potential to redefine the distribution of the e-commerce value chain.
AI-powered brand marketing: From mass marketing to individual conversations
AI's impact on business isn't limited to purchasing behavior; it will fundamentally reshape the very logic of brand marketing. In the era of AI agents, the effectiveness of traditional mass marketing will decline significantly, as consumers no longer actively search for and compare products, relying instead on AI agents' recommendations.
This means brands need to learn to talk to AI, not humans. AI agents are more rational and data-driven when evaluating products. They won’t be swayed by fancy packaging or emotional advertising, but will focus on objective performance metrics, cost-effectiveness, and customer satisfaction ratings.
But this doesn't mean brand storytelling is becoming less important. On the contrary, I believe authentic brand narratives will become even more crucial, as AI agents deeply analyze brand consistency and credibility. If a brand's messaging is inconsistent across platforms and at different times, AI can easily identify this and downgrade its recommendation weighting.
I predict the emergence of a new marketing role: the AI relationship specialist. These specialists will ensure that every aspect of a brand's product information, pricing strategy, inventory management, and more are correctly understood and evaluated by AI. They'll need to optimize product data, manage API integrations, monitor AI recommendation patterns, and more.
Another significant shift is the ultimate in personalization. When AI agents gain a deep understanding of each consumer, brands can offer personalized products to each individual. This isn't just about personalized recommendations; it's about personalized products themselves. Imagine your AI agent telling a clothing brand your exact size, color preferences, material requirements, and budget. The brand can then create a unique piece just for you. This kind of mass customization becomes economically feasible in the AI era.
The next decade: What are we witnessing?
After thinking deeply about Justine's analysis and my own observations, I feel that what we are witnessing is not only a transformation of the e-commerce industry, but a deeper shift in economic behavior.
Traditional economics assumes consumers are rational actors who proactively gather information, compare options, and make optimal decisions. However, in reality, we all know that human decision-making is fraught with bias, emotion, and cognitive limitations. The emergence of AI agents may make consumers more rational, as AI can process more information, avoid emotional biases, and consistently apply decision-making criteria.
The widespread adoption of this rational consumption model could have profound implications. First, market efficiency would increase significantly, as consumers would be able to more accurately assess product value. Second, product quality would become more important than marketing prowess, as AI agents wouldn't be fooled by flashy advertising. Finally, price transparency would increase, as AI could easily compare prices across the internet.
However, I also worry that this "hyper-rational" consumption may have some negative consequences. The joy of discovery in shopping could be diminished, as AI agents consistently recommend the "optimal" option rather than surprising or delightful choices. While impulsive buying isn't entirely rational, it's also part of the joy of life. If everything is optimized by AI, life could become overly predictable.
From a broader perspective, I believe the application of AI in the business sector will accelerate the digitization of the economy. As more and more business activities are digitally recorded and analyzed, this will provide an unprecedented data foundation for economic planning and policymaking. Governments may be able to more accurately predict economic trends, identify market failures, and design targeted interventions.
I predict that within the next decade, we'll see AI-driven businesses evolve from experimental applications to mainstream practice. Early adopters will gain significant competitive advantages, but as the technology becomes ubiquitous, these advantages will gradually become commoditized. The true long-term winners will be those companies that can redefine customer value in the AI era.
"`
