Perplexity started it in November of 2024 with Buy with Pro. Then Open AI with ChatGPT followed in April 2025. These two large language models (LLMs) are rapidly evolving from friendly chatbots that could write a great prospect email or blog post into helpful AI shopping assistants capable of turning a 150-word search prompt into a purchase. And, all without the user ever visiting a store or hitting a buy button. It’s a transformation that could fundamentally reshape retail.
As OpenAI, Perplexity, and others race to capture this trillion-dollar opportunity, the future of how consumers search and buy hinges on two critical questions: how these platforms will make money, and how their algorithms will decide which products to show consumers (or buy on their behalf).
The answers will determine whether these chatbots deliver on their promise of personalized commerce in their truest and most authentic sense — or become a more sophisticated version of today’s pay-to-play search and commerce platforms.
Where visibility goes to the highest bidder rather than the best match.
Retail’s Search and Discovery Roots
The world’s first department store opened in 1852 in Paris. Le Bon Marche was a big deal in its day. For the first time, consumers could stroll through a single store and see a collection of products that once required separate trips to separate stores — all without the pressure to buy. Fixed prices eliminated haggling, and shopping became both social and transactional.
It would take another century before the idea of assembling multiple stores in a single enclosed structure called a mall would open in Edina, Minnesota. The year was 1956. Thirty-nine years later, Amazon would take the mall online. As all of you know, what started as an online bookstore in 1995 would become the world’s largest online retailer, accounting for more than half of the world’s online sales today.
Shopping agents are positioned to become invisible sales conversion engines.
Roughly ten years after Google’s launch in 1998, the rise of vertical platforms like Houzz, 1st Dibs and Etsy gave consumers access to product search and discovery alternatives, and a more direct connection between consumer intent and the purchase of specific goods without leaving those sites. In the mid to late 2010s, social commerce sites like Instagram, Pinterest and TikTok brought creators, influencers and consumers together to find, discover and buy from brands new and emerging.
Today, less than three years after the launch of Open AI’s ChatGPT, shopping agents are positioned to become invisible sales conversion engines. Their strength as a conversational interface, capable of understanding complex requests, makes them well-suited to complete purchases without users ever visiting a physical or digital store or leaving the conversation.
Provided they have a little — well, actually, a lot — of help from their payments and commerce friends.
An emerging Agentic AI commerce ecosystem now stands at the ready to help advance their ambitions. Card networks, payment processors, FinTechs and AdTech startups use tokenized credentials and established payment infrastructure to collapse the traditional browse-select-purchase journey into a natural conversation capable of making a sale.
Existing technology foundations and massive infusions of capital fuel this momentum. The enthusiasm across the payments and commerce ecosystem to support a new way of matching consumers and merchants creates mutually reinforcing tailwinds.
From Keywords to Detailed Prompts
Now, granted, these are very, very early days.
Google processes 8.5 billion daily searches (heading toward 13 billion some analysts report) and delivers three quarters of Google’s massive sales engine. It is integrating AI-overviews into search results. Amazon’s Alexa serves tens of millions Prime Members whose intention is purely transactional on their platform. A new souped-up subscription Alexa is touted as handling complex searches and completing them, all tied to the Amazon ecosystem of third-party sellers and 600 million connected devices that are Alexa-enabled.
In this future, the very concept of “shopping” becomes a contextual, embedded part of a conversation with a chatbot.
Large platforms like Expedia and Booking.com and Airbnb are developing their own agents to help travelers navigate and book complex travel-related requests without leaving their platforms.
But the speed at which the GenAI chatbots have amassed an audience shows the potential for how these models could upend the retail and commerce status quo by changing where consumers start their searches and end by making a purchase without a lot of steps or friction in between.
Where the very concept of “shopping” becomes a contextual, embedded part of a conversation with a chatbot.
When the Chatbots Do the Shopping
Those conversations are much more detailed and have the potential to become much more transactional, just like any face-to-face conversation with a knowledgeable sales associate. For instance, instead of asking a Google search to recommend a highly-rated paint brand, an AI chatbot query might start by asking what top designers might recommend as the right color for a living room that gets the morning sun and has yellow and green upholstery. That response could become the starting point for a purchase that is context first, product and brand recommendation second.
It’s a threat that extends far beyond the traditional search and discovery platforms like Google and Amazon. Retailers could find themselves caught in the proverbial chatbot crossfire. As AI agents increasingly handle the search and presentation of results (or completed sales), traditional retailers risk becoming invisible in the commerce ecosystem altogether.
Retailers could find themselves caught in the proverbial chatbot crossfire.
In the good old days of organic search, retailers and businesses optimized their products to index on Google. Elaborate schemes to play nice with Google were the purview of many digital, SEO and paid media teams.
In the world of GenAI, how to get indexed remains a mystery box. Everyone sees the LLMs pinging their sites to train their models, but the algorithm for training the model and being presented as an AI overview on Google or part of the conversational search return is unclear.
And therein lies the complexity and uncertainty of this exciting and potentially transformative new world. No one — retailers, consumers, data networks or issuers — really understand how these AI shopping agents will work, how they make their recommendations, or the business model that will support making them.
In this world, payment credentials might emerge as the real winner as embedded offers, financing, rewards and other data-driven incentives become an invisible part of the transaction.
The Business Model Question: Innovation or History Repeating?
The venture capital pouring into these platforms signals expectations for massive adoption and ROI. OpenAI’s $300 billion valuation, with Anthropic, Perplexity and others following suit, represents more than just Silicon Valley optimism. These investors seem to be betting on a fundamental restructuring of how commerce works.
The question isn’t whether these platforms will monetize shopping, but how they’ll do it. History offers both cautionary tales and hopeful precedents.
In 1998, Google launched with a mission to “organize the world’s information.” It had to make money and naturally turned to ads. At first, this was limited to a few clearly marked ad-supported searches on the right side of the page that were easily distinguished from organic results.
Eventually there were “sponsored results” at the top of the page, shopping carousels, and featured snippets from paying partners. Much of the entire first page more or less became premium real estate for sale to the highest bidder.
Amazon’s journey followed a similar path.
As LLMs platforms race to capture retail market share, they face a fundamental choice between short-term monetization and long-term trust building.
Amazon was launched as a customer-obsessive platform, promising unbiased product search and honest reviews. But as the platform matured, sponsored products crept into search results, then dominated them. The company that disrupted retail by eliminating middlemen became the ultimate intermediary, making billions in advertising revenue from brands willing to pay for priority placement in front of Amazon’s high spending Prime Members. Amazon’s Q1 2025 results posted $13.9 million in ad revenue from Amazon’s retail network.
Amazon and Google’s transformations from organic search to sponsored posts happened gradually. And after consumers had invested years in honing their searching and shopping habits on those platforms. They’ve gotten used to scrolling past sponsored content and wondering how reliable those five-star products are. Because for now, at least, there’s no suitable alternative, at scale, to fill the gap.
As LLMs platforms race to capture retail market share, they face a fundamental choice between short-term monetization and long-term trust building. The temptation to follow the proven path of advertising-based models will be strong, especially as investors pressure for returns.
GenAI represents a new form of commerce orchestration across marketplaces, social signals and retail inventory through a simple and single conversational interface. The unique nature of conversational AI suggests that different approaches might not just be possible but necessary to compete. LLM platforms can build on GenAI’s growing sense of user trust, its potential for creating a distinctive new shopping and buying utility and its ability to monetize these new forms of value.
The Algorithm Question: Black Box or Breakthrough?
How these models present recommendations and act on behalf of users presents new challenges because of their current lack of clarity — and new opportunities because of what they could become for the entire commerce ecosystem. And that could reshape how retailers and the ecosystem adapt their products and platforms to drive sales.
For retailers and brands, that now means competing for both customer and AI attention. Retailers will need to ensure their inventory, pricing and product information are optimized for AI crawling and decision-making algorithms. And that their brands are beloved enough to become part of the user prompt. That’s a challenge without knowing how the LLM is making choices.
The conversational nature of these interactions adds another layer of complexity.
Unlike Google’s PageRank algorithm, which follows explicit, auditable rules, these LLMs generate responses through complex neural networks that even their creators can’t fully explain. It’s why no one really understands how and why they continue to hallucinate — and do so most convincingly. This opacity could create the most precise consumer-product-price matching ever experienced — or deliver subtle forms of influence that would be impossible to detect or regulate.
The conversational nature of these interactions adds another layer of complexity.
Traditional search engines respond to explicit keyword-based queries with discrete results. AI assistants engage in dynamic dialogues. A search query about planning a vacation can easily become an AI-agent’s task to make hotel bookings, restaurant reservations and must-see tourist attractions. Each touchpoint represents an opportunity for commercial influence — unless it is clear how these models will monetize those interactions. And how brands and retailers must adapt their strategies to remain a relevant part of what those AI-agents surface.
That said, when properly designed, these systems could understand consumer needs with a level of nuance and accuracy that goes well beyond simple product matching based on organic search and clicks. An AI agent might recognize that someone living in Boston buying patio furniture in November might find end-of-season sales appealing. And even suggest a slightly more expensive set that will last three times longer because of its weather-proof coating, ultimately saving money.
What’s Next
What we have learned over observing the last 173 years of retail transformation is that whoever controls search and discovery controls the commerce experience for consumers and retailers.
And that the best commerce experiences combine authority with relevant context, efficiency with ease of use and transacting with trust.
The rise of AI shopping assistants creates new and interesting possibilities for how these platforms deliver on that potential, and how they monetize these transformative experiences. Today, the LLMs who have dipped their toes into the shopping pool say the service will be free.
The winners in this new era may be those who recognize that when conversations drive commerce, trust itself becomes the product.
That won’t last for long. But instead of simply recreating the sponsored search model with a conversational interface, these AI agents have the potential to do something much more innovative. The winners in this new era may be those who recognize that when conversations drive commerce, trust itself becomes the product. And that monetizing trust comes wrapped around a different business model.
Technology exists to create either future.
The choice lies with those building these systems and the consumers who will ultimately determine whether they are, indeed, better than the current retail status quo.