What is conversational UX design?

Conversational UX design is the practice of designing interfaces that people interact with through natural language rather than by tapping buttons and navigating menus. It covers voice assistants like Alexa and Siri, text-based chatbots, and the AI chat interfaces built on large language models. The core skill is designing a dialogue: anticipating what a user will say, handling the many ways they might say it, and guiding the conversation toward a useful outcome without a visual layout to lean on.

What is the difference between VUI and GUI?

A GUI (graphical user interface) is visual and spatial. The user sees the available options laid out on a screen and chooses between them. A VUI (voice user interface) is linear and invisible. Nothing is on screen, so the user cannot see what is possible, cannot skim, and cannot easily go back. That single difference drives most of what makes voice design hard: you have to make capabilities discoverable through the conversation itself, keep the cognitive load low because everything must be held in memory, and design carefully for the moments when the system mishears or misunderstands.

Is voice and conversational design a good career path in UX?

It is a growing specialism rather than a separate career. Most designers add conversational skills on top of a solid UX foundation rather than starting there. Demand has broadened as AI chat interfaces have become mainstream: products now ship chat and voice features that need someone who can design a dialogue, write the system's responses, and plan for failure states. The underlying UX process (research, prototyping, testing) is the same, so the fastest route in is to build strong UX fundamentals first and then specialise.

What tools do conversational UX designers use?

For flows and dialogue mapping, most designers start in a whiteboard tool like Miro or FigJam, or a dedicated flow tool like Voiceflow or Botmock. Prototyping conversation is often done in a simple script or spreadsheet before any build, because the dialogue is the design. For chatbots, platforms such as Voiceflow, Landbot, and the builders inside customer-service tools let you prototype flows without engineering. For AI chat, prompt design and testing the model's responses become part of the work.

How do you test a conversational interface?

The most valuable early method is a Wizard of Oz test, where a person plays the role of the system and responds to a real user in real time. It surfaces the phrases people actually use and the points where the conversation breaks down, long before anything is built. After that, standard usability testing applies: give participants realistic tasks, watch where they get stuck, and pay particular attention to recovery when the system gets something wrong. Reading real transcripts once a product is live is the richest source of improvement.

Voice and Conversational UX Design: A Practical Guide

Ask someone to picture a user interface and they will describe a screen: buttons, menus, a layout. But a growing share of the interfaces people use every day have no screen at all, or use the screen only as a backdrop to a conversation. You speak to a voice assistant. You type to a chatbot. You hold an open-ended exchange with an AI chat tool. Designing those experiences is its own discipline, and it is one that has moved from niche to mainstream faster than almost any other area of UX.

This post explains what voice and conversational UX design is, how it differs from designing for the screen, the principles that make a conversation work, and how the skill fits into a modern UX career.

Want to design for AI-driven and conversational products? UX Academy (myuxacademy.com)'s AI UX Design course covers designing for chat, voice, and AI-powered interfaces on top of solid UX fundamentals. Cohort 1 starts 21 Sep 2026 -- reserve your place with a £99 deposit. Or try the free masterclass first.

What voice and conversational UX design is

Conversational UX design is the practice of designing interfaces where the primary interaction is natural language. It splits into a few overlapping areas:

Voice user interfaces (VUIs) -- voice assistants such as Alexa, Google Assistant, and Siri, plus voice features inside apps and in-car systems.
Chatbots -- text-based assistants on websites, in messaging apps, and inside products, from simple rule-based flows to more capable AI-driven ones.
AI chat interfaces -- the open-ended, large-language-model-powered tools that have become part of daily life for millions of people.

What unites them is that the design is a dialogue rather than a layout. You are not arranging elements on a canvas. You are anticipating what a person will say, planning how the system responds, and shaping the back-and-forth toward a useful outcome. The interface reveals itself through the conversation, not before it.

VUI vs GUI: why voice is a different problem

The clearest way to understand conversational design is to contrast it with the graphical interfaces most designers start with. For a refresher on how screen-based UX works end to end, our UX design process guide walks through the standard stages.

A GUI is visual and spatial. Everything available is on the screen at once. The user can see the options, skim them, ignore the ones that do not apply, and choose. Going back is a button. Discovering a feature is a matter of looking.

A VUI is linear and invisible. Nothing is on screen. The user cannot see what the system can do, cannot skim, and cannot glance back at the last step. Every option has to be spoken, held in memory, and acted on before it disappears.

That single difference drives most of the hard parts of voice design:

Discoverability. In a GUI, capabilities are visible. In a VUI, the user has no menu to scan, so they often do not know what they can ask. The design has to teach capability through the conversation without lecturing.
Cognitive load. A screen offloads memory: the options are right there. A voice interface asks the user to hold everything in their head. Long lists of spoken choices fail because nobody remembers option four of six.
Error and ambiguity. People phrase the same request in dozens of ways, and speech recognition mishears. A GUI button is either pressed or not. A spoken request is a probability. Designing for the times the system gets it wrong is not an edge case in voice work; it is most of the work.

Chatbots and AI chat sit between the two. They have a screen, so they can show buttons, quick replies, and history, but the core interaction is still language, so they inherit the ambiguity and the need to design for failure.

Principles for designing conversations

A few principles hold across voice assistants, chatbots, and AI chat.

Design the dialogue, not the screen

The deliverable in conversational design is a flow: what the system says, the range of things the user might say back, and how each branch is handled. Most designers map this in a whiteboard tool or a flow tool before anything is built, because the conversation is the design. If the dialogue does not read well out loud, no visual polish will save it.

Set expectations early

Because capabilities are invisible, the opening moments matter. A good conversational interface signals what it can help with quickly and concretely, rather than an open "How can I help?" that leaves the user guessing. Naming one or two real things the system does is more useful than a friendly but empty greeting.

Keep turns short and give one thing at a time

Long spoken responses lose people. So do walls of chatbot text. Break information into short turns, confirm progress, and avoid asking the user to hold several things in memory at once. In voice especially, offering more than about three options in a single turn tends to fail.

Confirm, but do not nag

For anything consequential (a booking, a payment, a deletion), confirm before acting. For low-stakes steps, constant "Did you mean X?" checks make the experience feel slow and untrusting. Matching the amount of confirmation to the cost of getting it wrong is a judgement call that good conversational designers develop.

Design the failure states first

The user will be misheard. They will ask for something out of scope. They will go quiet. A conversational interface is judged largely on how gracefully it recovers: acknowledging the problem, offering a way forward, and never trapping the user in a loop. Weak chatbots fail here more than anywhere else.

Write in the product's voice

In conversational design, the words are the interface. Tone, brevity, and clarity are design decisions, not a copywriting afterthought. This overlaps heavily with UX writing -- our UX writing guide covers the principles that apply directly to bot and assistant responses.

A simple process for a conversational project

The underlying UX process does not change for conversation; the artefacts do.

Research. Learn how people talk about the task in their own words. Real language, collected from interviews or existing transcripts, is the raw material for the whole design.
Map the happy path. Sketch the ideal conversation from start to successful outcome. Keep it short and natural.
Branch and handle failure. Add the realistic detours: misheard input, out-of-scope requests, silence, corrections. This is where most of the design effort goes.
Prototype cheaply. Test the dialogue with a Wizard of Oz session -- a person playing the system -- before any build. It reveals the phrases users actually use.
Test and read transcripts. Once live, real transcripts are the richest feedback you will get. Read them regularly and feed the gaps back into the flows.

Tools of the trade

Conversational designers tend to reach for a small set of tools. Flow and dialogue mapping usually starts in Miro or FigJam, or in a dedicated tool such as Voiceflow for building and prototyping voice and chat flows without engineering. A plain script or spreadsheet is often enough to draft and pressure-test the dialogue itself. For chatbots, builders like Voiceflow and Landbot let you prototype working flows quickly. For AI chat, prompt design and evaluating the model's responses become part of the toolkit -- a fast-moving area covered more broadly in our look at AI in UX design and the AI tools UX designers use.

How it fits a modern UX career

Conversational and voice design is best understood as a specialism layered on top of strong UX fundamentals, not a separate track you start from scratch. The research, prototyping, and testing skills are the same ones every UX designer needs; what changes is the medium.

That layering matters for anyone planning a route into the field. The market has shifted toward AI-literate designers who can work across screens, chat, and voice, and who understand how a conversation is structured rather than only how a screen is arranged. Building solid UX foundations first, then adding conversational skill, is both the realistic sequence and the one employers expect.

On UX Academy's AI UX Design course, you learn to design for chat, voice, and AI-powered products on top of a proper UX grounding -- research, flows, prototyping, and testing applied to conversational interfaces. The instructors work in the industry and teach the process live. No prior design experience required.