Perplexity’s high bar for UX in the age of AI

January 10, 2024

2023 ushered in a new age of computing. The focus of generative AI has so far been largely on the underlying tech. Meanwhile, most AI products tack closely to the chat interfaces launched by the foundational model providers. Like in the early days of PCs, these command line interfaces ask a great deal from the user.

But now it’s 2024. One company has emerged as pushing the frontier of a UX powered by AI: Perplexity. They’re reimagining how we search the web using AI. The buzz on Twitter and a Bezos-funded Series B, shows they’re onto something.

Part of their success is keeping one foot in the past. Since human behavior is slow to change, there are first principles we can tap into. In 1994, Jakob Nielson wrote the 10 Usability Heuristics for User Interface Design. Here’s how Perplexity applies them to great success.

Heuristic 1: Visibility of system status

The design should always keep users informed about what is going on, through appropriate feedback within a reasonable amount of time.

In any product, whenever there’s a perceived performance delay, showing the user the system is still working is critical. Ideally you can cut the time down by improving system performance. Since that’s not always possible, a progress indicator is a baseline expectation. Perplexity does it even better by showing the user exactly what their model is doing as it’s crawling for the information you’ve asked for. By seeing text like “Considering 8 sources,” or “Researched and summarized,” the user gets to see not just that the system’s working, but what it’s doing and how this new thing works. The result is more trust in the technology itself. The user understands the technology better. It’s more familiar. It’s more likely to be used again.

Heuristic 2: The design should speak the users’ language

Use words, phrases, and concepts familiar to the user, rather than internal jargon. Follow real-world conventions, making information appear in a natural and logical order.

Perplexity’s choice of words is easy to understand since they use the mental model of a human conversation. A piece of text like “Ask a follow-up…” is conversational. It’s a prompt to take a step that shows the value of Perplexity: being able to quickly refine your web search results using the context of your first question. This all seems simple, but it’s hard to get these wording choices correct. The result is a more approachable, less intimidating product. It feels natural in a way no one else has previously matched.

Heuristic 3: User control and freedom

Users often perform actions by mistake. They need a clearly marked “emergency exit” to leave the unwanted action without having to go through an extended process.

AI chat products generally encourage users to share the chat outputs, almost as a source of record. As a user, I haven’t found found value in these features, but Perplexity does a good job of making this more valuable by giving the user more control. Users can edit previous prompts to regenerate results. You can delete followup questions and answers. This gives you tools to edit the series of responses into a more curated package that’s more shareable.

Heuristic 4: Consistency and standards

Users should not have to wonder whether different words, situations, or actions mean the same thing. Follow platform and industry conventions.

Perplexity sticks closely to common words used across the web. When there is a new term or UI introduced, they often explain its meaning via a tooltip. Users may not know every detail right away, but the Perplexity team makes it easy to understand anything novel.

Heuristic 5: Error prevention

Good error messages are important, but the best designs carefully prevent problems from occurring in the first place. Either eliminate error-prone conditions, or check for them and present users with a confirmation option before they commit to the action.

This is one of Perplexity’s breakthrough innovations. When a prompt is expected to provide results that aren’t specific enough, Perplexity asks the user to refine their question. You can assume that Perplexity looks at insufficient results (e.g. too broad, not specific enough) as a failure state. In a search product, that is the case, but incredibly it’s not one that other search providers have built as sophisticated of solutions for.

Asking for clarification feels natural since that’s what humans do when talking to each other. Again, this helps build trust.

Heuristic 6: Recognition rather than recall

Minimize the user’s memory load by making elements, actions, and options visible. The user should not have to remember information from one part of the interface to another. Information required to use the design (e.g. field labels or menu items) should be visible or easily retrievable when needed.

This is another key innovation for Perplexity. Other generative AI chat products have been heavily reliant on recall over recognition, to the detriment of a broadly appealing UX. Perplexity recognizes that just like very few people ask questions at the end of big keynote presentations, not all users are great at asking followup questions. So, they predict what follow up question you may ask and show them at the end of every answer.

Beyond the follow up questions, the team has invested in creating a Discover section, which features new, broadly interesting topics on a daily basis. These human curated prompts are an easy way to engage with the product, even when you don’t have a prompt of your own. Often when I tap into one of these Discover items, I find myself asking follow up questions. It’s a fun, brand new way to interact with the news.

Just like the evolution from command line interfaces to graphical user interfaces, Perplexity’s choice to always give users a path forward via recognition is a big step forward in making this technology more approachable to the masses.

Heuristic 7: Flexibility and efficiency of use

Shortcuts — hidden from novice users — may speed up the interaction for the expert user so that the design can cater to both inexperienced and experienced users. Allow users to tailor frequent actions.

Keyboard shortcuts are sprinkled throughout the interface. Command-K is surely a favorite of us early adopter tech workers. These touches show a product is eager to meet the user wherever they are along their adoption curve.

Heuristic 8: Aesthetic and minimalist design

Interfaces should not contain information that is irrelevant or rarely needed. Every extra unit of information in an interface competes with the relevant units of information and diminishes their relative visibility.

The interface is simple, friendly, modern. While others look more engineering oriented and technical, Perplexity continues their feeling of being approachable yet from the future.

Heuristic 9: Help users recognize, diagnose, and recover from error

Error messages should be expressed in plain language (no error codes), precisely indicate the problem, and constructively suggest a solution.

The only ‘error’ I’ve experienced is when you exceed the quota for Co-pilot (the more advanced search) usage on the free plan. In this case, the status is clear. A tooltip tells you why the feature is unavailable. It’s clear what action you need to take: upgrade.

Perplexity does a great job of avoiding errors by gathering more information from the user as needed. The best type of error message is the error message that’s avoided.

Heuristic 10: Help and documentation

It’s best if the system doesn’t need any additional explanation. However, it may be necessary to provide documentation to help users understand how to complete their tasks.

Perplexity checks this box too. As mentioned above, they do a nice job educating the user on this new technology through usage. But if you want to learn more about Perplexity, they have a straightforward Help and FAQ section. It’s simple since they’re clearly pushing to keep the product itself as simple as can be.

It’ll be great to see how Perplexity continues to evolve. If you’re working on an AI-focused product in 2024, follow Perplexity’s lead: always remember there’s something to learn from 1994.