What ChatGPT can Teach Ecommerce Companies About the Future of Product Discovery

You’ve probably played with artificial intelligence (AI) chatbot ChatGPT already and felt like you were touching magic. It feels like AI on a level we’ve never seen before, and its existence is massively exciting to all of us working in AI because it opens up so many new possibilities.

As lots of SaaS companies race to make something — anything — that incorporates ChatGPT within it, it’s the underlying technologies that ChatGPT is built on that are most exciting. And these technologies, when applied strategically (and not just for the sake of saying “me too!”), are going to change the future of ecommerce product discovery.

There are three disruptors tied to ChatGPT in the market today that have the biggest potential for ecommerce.

1. Large Language Models (LLMs) and transformers will change search habits.

LLMs and transformers — which help computers understand the context of long-form text — enable AI technologies to truly understand a query instead of just comparing it to keywords.


One of the reasons this technology promises to be so disruptive is that as it becomes more widespread, it has the potential to change how we search — at least in some situations. What shoppers have been used to up until now is searching with keywords and sounding almost like “cave people” doing so: Think of the queries “rug for kitchen” or “nespresso pods vertuo.” People don’t write like this because it’s the most convenient way for them to search, but rather because we’ve become accustomed to using just keywords and avoiding any additional words that might confuse a traditional search engine.

With transformers and LLMs, these terse and stilted searches become less necessary. Shoppers will be able to add expressiveness and nuance to their queries, explaining what they want in full sentences and expecting a reasonable response back. When shoppers have a more complicated need, like getting a non-skid kitchen rug that’s absorbent and ideal for hardwood floors — or finding pants that won’t be too warm for a New York summer, or finding bread that has the fewest carbs in it — being able to explain themselves fully really does help.

2. Companies can turn search into a larger (personalized) conversation.

The second major innovation that ChatGPT is popularizing is understanding that every interaction a person has with it is part of a larger dialogue. Remembering what a person has told it in the past allows the AI to tailor future responses with that knowledge in mind. It’s personalization that’s not just an add-on but based entirely on AI and built into the core of what ChatGPT is. The way it takes in the context of the user’s journey to return more relevant results feels genuinely useful.

AI-based product search technology today follows this same principle to deliver cohesive results and experiences that build on each other and make sense across the shopper journey. For example, by using clickstream data (information collected while a user browses a website) rather than keyword matching as a primary determiner of what products to show people, ecommerce companies can present search results to shoppers that map to their individual histories as well as in-the-moment wants and needs.

The reason this type of AI-driven personalization is so critical is that — contrary to the “relevant” results that traditional keyword-based search engines optimize for — this type of personalization acknowledges that the same query can have different answers for different people, as well as different answers at different times for the same person.

Understanding the context of a query is a big part of what makes its results attractive to a user. For example, if I inform ChatGPT that I live in Seattle and ask for shirts to buy, I get one set of recommendations. If I tell it I’m moving to New York, my recommendations will be different. It’s the same query, asking for shirts — but the AI is taking into account what it knows about me, rather than just returning the same basically “relevant” results (shirts) that a traditional keyword-based engine would spit out, regardless of the context. In ecommerce, context is critically important to how great (or not-so-great) an experience feels.

3. More specialized solutions will arise in the retail tech market.

ChatGPT also is paving the way for more widespread use of specialized solutions for product search and beyond — illustrating their benefits to companies and consumers alike.

Traditionally, ecommerce companies used “one-size-fits-all” search and discovery engines that were not built around the fact that domain-specific clickstream data exists. If results looked basically relevant, or merchandisers could boost and bury them to make them look basically relevant, this would be considered good enough.

Now that engines tailored to specific domains (such as ecommerce) and clickstream data are becoming available, retailers and brands can tap into a combination of higher revenue and merchandising controls. The advantage to using ecommerce-specific AI is that machine learning algorithms can learn this domain really, really well. For example, understanding that add-to-carts are a critical signal in ecommerce — and that revenue is a critical goal — massively helps an engine learn to optimize for revenue.

If an engine focuses on ecommerce, this type of optimization is an obvious thing to do. But for a one-size-fits-all search engine that’s supposed to work for everything — from searching for support documentation to searching forums to searching media and also promising to do ecommerce — this is not possible, because revenue and add-to-carts only exist in one of those situations. What’s more, it also prevents merchandisers from attaining the critical ecommerce-specific insights they need to drive strategy. When both a purchase and an add-to-cart are considered a conversion by the AI, how do you know what strategies are working to drive revenue?

This isn’t a problem ChatGPT aims to solve, and that’s OK. It’s one of the things I admire most about ChatGPT developers OpenAI: The company openly admits that each of its models, like ChatGPT and Codex, are meant for a specific use case. It’s not a one-size-fits-all because there are advantages to being domain-specific.

Making Ecommerce Magic

In the end, a lot of the magic we’ll see in the ecommerce space will likely be a marriage of all three disruptive trends and technologies that ChatGPT is bringing to the fore. This combination allows ecommerce companies to better understand an individual’s context (and how and when that context may change), be much more personalized in that understanding and provide greater customization from the merchandising side.

And for vendors and retailers alike that are exploring the future of search and discovery for enterprise ecommerce, all of these advancements in AI make it a really thrilling space to work within. As retailers face the need to provide valuable, engaging customer experiences, the gap between cookie-cutter (and even off-base) results and hyperpersonalized, optimized experiences will only grow over time. By illustrating what’s already possible today and what will be possible tomorrow — as well as how accessible that is — ChatGPT has given companies a lot to chat about…and also to execute on!

Eli Finkelshteyn is Co-founder and CEO of Constructor, an AI-powered product search and discovery platform tailor-made for ecommerce. An accomplished data engineer and entrepreneur, Finkelshteyn is committed to building innovative, data-driven search experiences that are both optimized for retailers’ key KPIs and personalized to customers’ needs.

Feature Your Byline

Submit an Executive ViewPoints.

Featured Event

Join the retail community as we come together for three days of strategic sessions, meaningful off-site networking events and interactive learning experiences.


Access The Media Kit


Access Our Editorial Calendar

If you are downloading this on behalf of a client, please provide the company name and website information below: