Conversational Commerce: How AI Chatbots Are Changing Online Shopping

That's the question every retailer is quietly answering right now, whether they've framed it that way or not. Online retail has moved through a few genuine inflection points over the years, such as the shift to mobile and the rise of one-click checkout, and conversational commerce is arguably another: not a feature bolted onto a website, but a change in the basic grammar of how people shop.

For most of e-commerce's history, the customer has had to adapt to the interface. Conversational commerce reverses that arrangement:

It sounds simple. In practice, getting it right has taken the better part of a decade of trial and error, and far better language models than the chatbots most people remember.

At its core, conversational commerce is the practice of buying and selling through natural dialogue (text, voice or chat) rather than through static pages and forms. The term dates back to around 2015, but what it means in practice has changed almost beyond recognition since then.

The early version was little more than decision trees dressed up as conversation: "Press 1 for sizing, press 2 for returns." Functional, but nobody mistook it for a real exchange. What has changed is the underlying intelligence, built on natural language processing. Modern AI chatbots can hold context across a multi-turn conversation, interpret a vague or oddly phrased question, and respond with something that resembles judgement rather than a scripted branch.

It is about whether the system behind that window can genuinely understand what a customer is asking for.

Vendor-supplied conversion figures deserve a degree of scepticism, because many are drawn from a single best-case deployment or from the vendor's own customers rather than a representative sample. Headline numbers do circulate. One vendor, Rep AI, reported in 2025 that 12.3% of shoppers who engaged with its AI chat made a purchase, against 3.1% of those who did not.

Because shoppers who start a chat may already be more motivated to buy, a gap like this shows a strong association, not proof that the chatbot caused the sale. Whatever the true size of the effect, though, the mechanism is plausible.

A well-implemented chatbot doesn't simply answer questions. It can remove the specific moment of friction where a shopper would otherwise close the tab and compare elsewhere. A traditional product page has to anticipate every possible question in advance and hope the answer is already on the page. A conversation isn't bound by that constraint: it answers the question actually being asked, at the moment the doubt appears. That structural difference, more than the technology itself, is what separates conversational selling from a faster version of the old browse-and-buy model.

Three moments in the buying journey are where a real-time conversation tends to help. Each is a point where a shopper would otherwise hesitate, search elsewhere or simply leave.

Answering Product Questions Before They Become Objections

Every unanswered question is a reason to delay. A chatbot can resolve it on the spot (sizing, materials, compatibility) before doubt has a chance to turn into abandonment.

Guiding Customers Through Complex Choices

Some products can't be sold with a filter menu. A chatbot can ask what a customer actually needs, then narrow dozens of options down to two or three that fit, doing in seconds what a sales assistant would do in-store.

Recovering Abandoned Carts Through Conversation

Many abandoned carts are not lost causes but unanswered hesitations. A well-timed message that addresses the specific reason someone paused can work better than a generic discount code.

Not every customer welcomes a chat window, and dismissing that resistance would undercut the credibility of everything said above. Three reasons are commonly cited:

None of this argues against conversational commerce as a strategy. It argues against treating a chatbot as a universal replacement for every other path to purchase, and for offering conversation as an option rather than the only door in.

Text-based conversation solved the "ask instead of search" problem. Voice removes the keyboard entirely, and for certain moments that matters more than it might first appear:

Where it currently falls short is complexity. Comparing five products by their specifications is straightforward in text and awkward out loud: voice struggles the moment a decision needs visual comparison. For now, voice looks less like a replacement for chat-based conversational commerce and more like a narrower channel suited to specific, low-ambiguity purchases, with text remaining the default for anything that requires genuine deliberation.

Conversational commerce doesn't operate in isolation. Its effectiveness depends heavily on what the system already knows before the conversation even starts. A chatbot that recognises a returning customer's past purchases can skip the small talk and get straight to a relevant recommendation, drawing on the same underlying logic covered in our guide on AI-Powered Personalization in E-Commerce.

Pricing adds a further layer worth understanding. Chatbots may well field questions like "can you do any better on this price?", a scenario that touches directly on how dynamic pricing systems operate behind the scenes, and a topic covered in more depth in our dedicated guide on AI Dynamic Pricing: How Retailers Decide What You Pay, in Real Time.

Deploying a chatbot is a technical decision. Getting value from one is a judgement decision, and the difference between the two explains why some implementations underperform despite using capable technology.

A few questions are worth answering before rollout, not after:

Getting these decisions right requires the same underlying skill as the rest of conversational commerce strategy: understanding what the technology is genuinely good at, and where human oversight still earns its place.

Reading about conversational commerce and knowing how to deploy it profitably are two different skill sets, and the gap between them is where many implementations lose momentum.

At London Optimum Training & Consultancy (LOTC), these courses are relevant starting points for e-commerce teams making these decisions. Training is available in London, Barcelona, Paris and Dubai.