Why Links Are Blue: A 1993 Changelog Line, a Purple Twin and Google's 41 Shades

Open any browser with its default styles and you will see it. Unvisited links are blue. Visited ones are purple. Nobody taught you this, and no style guide told you to expect it. Yet if a link showed up in orange, you would hesitate before clicking.

That reflex traces back to a single line in a software changelog from 1993. The story behind it has no tidy origin, and the colour has since been tested, fought over and monetised by some of the biggest companies on the web.

One line in a changelog

NCSA Mosaic was the first widely used graphical web browser. Its developers, Marc Andreessen and Eric Bina, released it in January 1993. The release notes for version 0.13, dated April 12, 1993, contain the sentence that matters:

Changed default anchor representations: blue and single solid underline for unvisited, dark purple and single dashed underline for visited.

That is the whole announcement. No rationale, no study, no design memo. The notes sit between fixes for core dumps and a note about a built-in finger interface.

Notice the word "changed." It implies Mosaic had used some other treatment before, and the team swapped it for blue and purple. Nobody wrote down why.

What links looked like before

Blue was not the obvious choice, because the early web had no obvious choice. Tim Berners-Lee's original browser showed black text with underlines. ViolaWWW, a 1992 browser, drew black links on a gray background. Going back further, the 1983 hypertext system HyperTIES used cyan links on black.

Then Mosaic shipped on several operating systems, and its look spread. Netscape Navigator arrived in 1994 and Internet Explorer in 1995. Both showed blue links on gray. Once three of the most used browsers agreed, the convention stopped being a choice and became a habit.

So why blue?

Nobody knows for sure. Several theories compete, and each is plausible.

• Legibility. Berners-Lee has speculated that blue is the darkest colour and so threatens legibility least. Blue text still reads as text.
• Operating system habits. Elise Blanchard, writing for Mozilla, points to Windows 3.1, which had recently used blue to mark selected items. Blue already meant "this responds to you."
• X Windows. Commenters on a TidBITS piece suggested blue was the standard highlight colour on the X Windows setups used at CERN.
• Taste. TidBITS notes it could be as simple as a key programmer liking blue. Some decisions are arbitrary.

One popular explanation deserves suspicion: that blue was picked for colour-blind users. Blanchard's analysis argues against it. By her figures, the contrast between blue and black body text was only about 2.3 to 1, well below what modern accessibility standards would accept.

Run the numbers yourself. Today's default link blue is #0000EE. Against black text, the WCAG contrast formula gives about 2.2 to 1. That is not enough to tell links from text by colour alone, which is why Mosaic paired the colour with an underline. The underline did the accessibility work. The colour did the branding.

The purple twin

Blue got the credit. Purple did the quiet work.

The HTML standard still sets the defaults Mosaic started: #0000EE for links, #551A8B for visited links, and #FF0000 for active ones, the instant you press them.

Look at how those two relate. Against white, the blue scores about 9.4 to 1 and the purple about 11 to 1. Both read easily. But against each other, they manage only about 1.2 to 1. A visited link and an unvisited one are nearly the same lightness. What separates them is hue, and hue alone.

That means someone who cannot distinguish blue from purple loses the "you have been here" signal entirely. It is a real weakness in the original scheme, and a good reason to vary lightness as well as hue.

Bowman, Mayer and 41 blues

The blue link sat quietly in the background of the web until Google turned it into a debate.

On March 20, 2009, the designer Douglas Bowman posted a farewell titled "Goodbye, Google." He was leaving for Twitter, and his reasons were about process. He wrote that a team at Google couldn't decide between two blues, so it tested 41 shades between them to see which performed better. And he added a second example: "I had a recent debate over whether a border should be 3, 4 or 5 pixels wide, and was asked to prove my case."

He closed with a line that still gets quoted: "I can't operate in an environment like that. I've grown tired of debating such minuscule design decisions."

Marissa Mayer, then a Google executive, became the face of the approach. She is widely reported to have said that adjusting the colour of ad links added about $200 million a year in revenue. A slightly purpler blue is said to have beaten slightly greener ones.

Is the $200 million real?

Treat the number with care. It circulates through blog posts and conference talks, and the primary evidence is thin.

Critics have raised fair objections. Google cannot control the screens people use, and the same hex code looks different on different monitors. The reported difference between the best and worst click rates was tiny. And $200 million, set against the scale of Google's ad business, is a small slice of revenue.

You do not have to resolve that argument to take the point. A small, measurable edge, multiplied by billions of impressions, can be worth real money. A single study of one audience on one page still does not tell you which blue suits your site.

What to take from it

Mosaic's team picked blue for reasons we cannot recover. Google's team picked a blue by testing 41 of them. Both ended up with a blue that works, which says something about how little separates "good enough" from "optimal."

Here is the practical version:

1. Keep the convention unless you have a reason. Readers arrive already trained. A blue, underlined word says "click" before they read it.
2. Never rely on hue alone. The underline is what makes links accessible. Drop it and you owe your readers another cue.
3. Separate visited from unvisited by lightness. Hue differences vanish for some readers, as the 1.2 to 1 ratio between default blue and purple shows.
4. Test your own audience. Google's number may not transfer to you, but your own click data will.
5. Check contrast on real backgrounds. Default blue holds up on white at about 9.4 to 1. On a light gray such as #C0C0C0, the kind of background early browsers used, it falls to about 5.2 to 1. Your background changes the answer.

Blue links are a habit with a murky birth. Respect the habit, then check the numbers.