Mascots in webdesign

A drawn character can give a small website a face. The trouble starts when it gets a voice.

Geert Valère Allegaert · October 2026

This site has a mascot. There is an alligator on the homepage, drawn in a few lines, and the footer says the site was made by the alligator himself. That is all he does. He does not wave, does not appear when you hesitate, does not tell you what you might want to do next, and has never said a word to anyone.

That restraint is deliberate, and it comes from one of the best-documented failures in interface design. A mascot is a good way to make a small business recognisable and a little warmer. It becomes a problem the moment it stops being a picture and starts behaving like a person.

The paperclip

In 1997 Microsoft put an animated assistant into Office. Its ancestor was Microsoft Bob, a home program from 1995 full of cartoon guides, built on research showing that people respond to computers socially, the way they respond to other people. Bob failed. The characters moved on into Office, and the default one was a paperclip with eyebrows, officially called Clippit and known to everybody as Clippy.

The paperclip was not even the favourite. In tests before the release, other characters scored higher; it was chosen mainly because a paperclip goes with an office. And it became famous for one behaviour. Type something that looked like the opening of a letter and it appeared, offering to help you write it — no matter how many times you had already said no.

Microsoft turned the assistant off by default in Office XP in 2001, and ran a campaign mocking its own creation to sell the upgrade. Office 2007 removed it altogether.

The most useful account of why people hated it is an undergraduate thesis. In 2003 Luke Swartz, supervised at Stanford by Clifford Nass — one of the researchers whose work had inspired the characters in the first place — interviewed Office users and ran two experiments. The paperclip broke every rule you would apply to a human assistant: it interrupted, it watched, it repeated the same question, and it never learned. Several people called it fine for beginners and patronising for themselves. And whether people liked it depended mostly on what they thought it was for. Those who saw it as a tool found its animations annoying; the few who saw it as a diversion enjoyed them. Same character, same behaviour, opposite verdict.

The paperclip has come back since, as a joke that works. In 2023 an independent developer put Clippy on the Windows desktop again, this time answering questions through ChatGPT, with Bob promised as a second character. It is a good test of the old diagnosis. The face that once promised more than it could do now sits in front of something that can actually hold a conversation. What remains different is the timing: the new Clippy waits until you ask. That was always the part that mattered.

What the research does and does not show

There was a hopeful name for the idea behind all this: the persona effect, the claim that a character makes people respond better to a system. In 2000 Doris Dehn and Susanne van Mulken reviewed the experiments on it. Their conclusion was modest. A character tends to make a system more entertaining. Almost everything else depends on which character, doing what, in which kind of task — and many of the studies changed more than one thing at once, so they could not say which change did the work.

One experiment from that literature makes the pattern clear. People who watched presentations guided by a character rated them more entertaining and found the material less difficult. They did not understand or remember any more of it. A mascot changes how a page feels. There is no good evidence that it changes what people understand or get done.

Two further findings point the other way. Ben Shneiderman, the most persistent critic of human-like interfaces, observed in 1995 that such characters are cute the first time, silly the second and an annoying distraction the third — a wear-out that a single round of user testing will never catch. And in a study by Rickenberg and Reeves, people doing a task while an animated character watched them reported more anxiety and performed worse. A pair of eyes on the screen is not neutral.

Most of this research is about assistants that talk, not about a drawing in a corner, and most of it was done with small groups of students. So the honest summary is narrow: a character can make a page more pleasant, the effect wears thin with repetition, and a character that acts has to behave better than a person would be expected to — because people judge it as one.

Where a mascot earns its place

Recognition. A small business with a common kind of name and a generic logo is hard to remember. A character is easier to remember than a wordmark, and it travels: favicon, invoice, a sticker on a laptop. For a one-person practice, it can also say that there is somebody behind the site, without a stock photo of a handshake.

Moments with nothing to do. An empty list before anything has been added, a page that is loading, a form that has just been sent, a 404. These are the places where a small character costs nothing, because the visitor is not in the middle of a task it can interrupt.

A voice that stays with the company. Mailchimp has had its monkey, Freddie, since the beginning, and its public style guide is clear about how far he goes: the brand may be funny, but clear always comes before entertaining, and for Freddie himself there is one instruction — “Don’t write in his voice.” The character stands for the personality; the words are written by people, in the company’s own voice. That is the most useful rule about mascots I know.

Where it should stay out

Anywhere something has gone wrong that matters. A failed payment, a lost file, a refused application, a medical result. A cartoon that shrugs at someone who has just lost money tells them the organisation does not understand what happened. The same goes for whole sectors: a government service, a hospital, a bank have visitors who are worried before they arrive.

Inside the task. The paperclip’s mistake was not its face, it was its timing. A character that appears unasked, covers part of the page or offers help nobody requested is an interruption with eyes. Everything that has been said about pop-ups applies to it, with the added cost that people take it personally.

As a lever on feelings. The newer danger is the mascot used deliberately. A well-known language app has an owl that sends reminders, looks disappointed when you skip a day and has appeared on the app icon looking ill and tired. The owl is not really the mechanism. What it guards is a streak, a count of days in a row, and what it adds is a face that is sorry when the count breaks. That is where a mascot meets gamification: the score keeps people coming back, and the character makes losing it feel personal. It works, by most accounts, and that is the problem. Researchers who catalogue manipulative design call the general technique nagging. A drawn face makes it stronger, not kinder.

As the only carrier of meaning. A mascot that looks happy when something worked and sad when it did not is telling half the story to a screen reader, and nothing at all to someone who cannot see the face clearly. A decorative mascot gets an empty alt text so it is skipped; a mascot that means something gets a text that says what it means — and the message is in the words on the page as well. Animated characters respect the visitor’s setting for reduced motion.

If you are thinking about a character for your own site, these are the questions I would settle first.

A face, not a voice

The paperclip was never hated for being a paperclip. It was hated for talking when nobody had asked, and for being unable to do what its friendly face promised. Give a mascot a face and it can make a site easier to remember. Give it a voice and you have hired an assistant — and people will judge it the way they judge any assistant who keeps interrupting them.

Read the originals

• Luke Swartz, Why People Hate the Paperclip (Stanford, 2003). The history of the Office Assistant, interviews with its users, and two experiments on how labels change the way a character is judged. The best single text on the subject.
• Mailchimp, Voice and Tone. A company with a famous mascot explaining in a few lines where he stops.
• Doris M. Dehn and Susanne van Mulken, The impact of animated interface agents: a review of empirical research (International Journal of Human-Computer Studies, 2000). What the experiments on the persona effect do and do not show.
• R. Rickenberg and Byron Reeves, The effects of animated characters on anxiety, task performance, and evaluations of user interfaces (CHI 2000). What happens when the character on the screen is watching you.
• Colin M. Gray and others, The Dark (Patterns) Side of UX Design (CHI 2018). The taxonomy in which nagging gets its name.
• AG Connect, Clippy keert terug met de kracht van ChatGPT (2023). The paperclip as a front end for a language model, made by an independent developer, not by Microsoft.