How AI is transforming online shopping. A consumer’s perspective of the browsing and buying experience – pros, cons, and feedback about using major LLM chatbots for eCommerce.
Jul 04, 2025
Ever since the convergence of search and eCommerce some 25 years ago, consumers have become native users of technology to make purchasing decisions. What began as simple keyword queries has evolved into a demand for instant results, personalised recommendations, and seamless buying experience. Today, AI search and conversational commerce carries that same promise, only they’re delivering shoppable convenience and speed at what feels like an exponential rate of improvement.
The fundamental shift can be pinpointed to ChatGPT’s public release in late 2022, sparking the rapid evolution of artificial intelligence tools and large language models (LLMs) for the general population. With ChatGPT weekly active users skyrocketing past 400 million as of February 2025, the widespread adoption of AI for search has organically taken users to the next logical use case: shopping.
Searching and buying online is practically a natural behaviour now, but the advancement of technology has similarly evolved consumer behaviours to make decisions and take actions with increasing immediacy. As a result, many marketers are arguing that the funnel is collapsing as the buying journey from awareness to purchase compresses into a shoppable singularity.
In recent months, three of the biggest players in AI – ChatGPT, Perplexity, and Gemini – have rolled out new shopping features to address this desire for discovery and delivery in one fell swoop. Let’s put them to the test and see how they can help an average consumer, such as myself, search and shop.
Only the free, signed out version of each AI was used to simulate a guest user experience.
The search prompts were identical across the platforms:
Locating a product with specific details: “Where can I buy the adidas Samba in white and black in size 9 US in Melbourne?”
A product recommendation based on given criteria: “Where can I buy noise-cancelling headphones with a budget of $500 in Melbourne?”
OpenAI launched ChatGPT publicly in November 2022, quickly becoming embraced by the public as the predominant generative AI platform. Since then, there have been 5 major core updates, including GPT-4 and the multimodal GPT-4o. In 2025, OpenAI introduced shopping features, allowing users to search, compare, and buy products directly through ChatGPT Shopping. It features personalised recommendations based on user queries and third-party metadata, prominently displayed product card carousels, and integrated product links.
ChatGPT claims its shoppable product listings are shown independently, helping eliminate a layer of bias, a subtle but important detail that boosts trust in results.
Searching for the adidas Samba loaded product cards that included callouts like “Alternative Model” for the Samba ADV (a skateboarding version). The images are pulled from the respective retailers’ product listings, creating a visually engaging way to showcase options, although it did occasionally result in misaligned placement as seen below.
ChatGPT Shopping was most impressive seeking products when given a query with specific criteria. This was best demonstrated in my search for headphones, yielding a wide range of options and retailers. Each option’s summary highlighted top picks, where to buy, and a brief explanation of its key selling points. When selecting a product card, the item would load on a side bar with additional information, including reviews and shipping information, providing an easy way to compare options.
Like many LLMs, ChatGPT can be prone to inaccuracies: it incorrectly identified a retailer’s location. Additionally, users may not always receive the same response twice for a given query, which can be frustrating.
The buy button that accompanies the side bar links out to the retailer’s product page. While functional, this took me out of the ChatGPT experience, opening up opportunities for distractions and detours before I reached the checkout.
The listed retailers that followed the product cards contained outbound links to buy. For searches with lots of potential options, it was mildly annoying having to scroll back up the chat to access the product cards with the links to purchase. For some, this would compound the friction point detailed above.
Perplexity launched its conversational search engine also in late 2022, initially using AI models including OpenAI’s GPT before developing the in-house Sonar model. It provides answers with citations through a simple search bar interface instead of traditional link lists. In 2024, Perplexity added shopping tools like product cards, one-click checkout, and the visual “Snap to Shop” feature, with some options currently available only to Perplexity Pro users in the US.
Perplexity was particularly good at nudging me with related questions that aim to provide additional context around my search.
It was also very detailed in its responses, providing numerous local options where available, plus a summary of each retailer, including address, store features, and a brief description of the company.
Perplexity interpreted my query about headphones a bit more literally, listing out retailers as opposed to providing product recommendations. Interestingly, it did name some boutique and specialist audio stores over some of the big-box names.
From a user interface perspective, the query being sticky at the top of the chatbox as I scrolled down the response was useful for reference. The information being grouped into tabs made for a somewhat more procedural experience, resembling a site navigation rather than a conversational feel typical of ChatGPT.
Perplexity tended to recommend retailer information such as brick-and-mortar locations rather than direct product purchase links. These responses were not conducive to a shoppable, eCommerce-first experience.
By extension, the interface wasn’t really that intuitive for actually making purchases. Like ChatGPT, links opened in new browser tabs, but instead of directing to the product page, they landed on the retailer homepages. If I was ready to buy something, I would have to navigate to the right product or open a new tab to go about the traditional online shopping journey – negating the actual need to engage with the LLM.
Google pioneered search engines and entered the cultural lexicon over 25 years ago, but AI offerings from newer players have threatened its market share. After a lukewarm launch with Bard in early 2023, it wasn’t until almost a year after ChatGPT’s debut that Google released Gemini, a considerably better AI interface. After regular improvements and 5 major version updates, the company announced incoming shopping features, including advanced AI search and agentic checkout, at their 2025 I/O event in May.
Gemini was fairly generous with Samba results, with links out to the stores that carried them. Like ChatGPT, Gemini included recommendations about what to consider when buying the product.
When there were many options available, Gemini presented them in an organised manner, grouping retailers by size and location. It even called out Amazon specifically for its competitive prices – a factor that definitely influences many consumers.
The basic version of Gemini had the least visually impactful interface, with all responses purely in text with a handful of outbound links. The recommendations for the headphones also implied Google’s search-first approach, interpreting the query about headphones in Melbourne in a literal sense, offering retailer locations (much like Perplexity). Overall, the incentive and mechanisms to purchase weren’t that present.
In short, the most basic version of Gemini is not an accurate representation of its agentic AI shopping power and would leave most shoppers wanting more. However, not all hope is lost. Google’s suite of enhanced shopping features should emerge on their AI Mode in the near future, which is not yet available to the general public. That may then yield a better shoppable AI experience.
Of the 3 major AI tools I tested for shopping, ChatGPT got me the closest to actually making a purchase. This was largely thanks to the familiar chat layout and generous linkouts to products. Perplexity wasn’t far behind, but it was probably more adept at giving me information about retailers and products than landing me on a checkout page. Gemini in its current incarnation still has room for improvement, which doesn’t look to be too far away.
Until the broader technical parts are figured out – integration, aggregation, payments, et al – average AI users wanting to use their preferred platform for shopping may only find it useful in its established function as a chatbot and research tool.
That being said, AI-enhanced shopping is improving at such a rapid rate that some of my grievances will likely be rectified sooner rather than later, especially as currently experimental features become available publicly.
Despite the potential of agentic AI to be used for shoppable experiences, my experience with these big 3 LLMs finds that the agency is in the hands of the user. While the decision making should be ultimately left to the consumer as they are the ones using the products, there still exists a gap in closing the sale – that is, bringing the checkout to the forefront. The answers at my fingertips are great, but how about a buy now button as well?
