Software & Apps

How to Choose Accent Colors for a Light Website

Choose accent colors that stay readable on a light website: a worked ivory, copper and sage palette, with every text pair checked for contrast.

Picking an accent color for a light website looks like a taste decision, and taste is part of it. But most accent colors fail for a more boring reason: they are chosen as a swatch and never tested where they will actually live — as a link on an ivory background or white text on a button. A color that looks warm and confident in a palette preview can turn unreadable the moment it has to carry meaning at 16 pixels.

The method is short: start from what the accent has to do, pick a small palette around one main accent, and check every text pair for contrast. The worked example uses a real sample — an ivory ground with a copper accent and a sage supporting color — with every ratio measured, including the two pairs that failed.

What accent colors do

On a light website, the background does most of the visual work. It covers most of the screen, sets the temperature of the whole page, and stays quiet. The accent has the opposite job: it is the small, loud part of the palette, marking the things a visitor should notice or touch — links, primary buttons, and active states. Because it is used sparingly, it can be far more saturated than anything else on the page without overwhelming it.

An accent also has to do two jobs that pull in different directions: stand out from the ground so the eye lands on it, and stay readable wherever it carries text. A mid-tone copper stands out beautifully against ivory as a button fill, but the same copper as small link text on the same ivory can fall short of the contrast a reader needs. The same color can pass in one role and fail in another — which is why you choose an accent for a role, not in the abstract. Name the roles first: page background, surface color, main text, muted text, and one main accent, plus optionally one supporting accent. Five or six colors, each with a job, is enough for most small sites.

Select a palette

Start with the ground, not the accent. The background sets the temperature everything else is judged against: a warm ivory makes the same copper feel earthy and calm, while a cool blue-white makes it feel sharper and more commercial. Pure white is clean and clinical; a warm off-white such as #FAF6EF feels softer and more editorial; a light gray feels neutral and product-like.

Next, describe how the site should feel in two or three words. "Calm and precise" points toward a narrow, muted palette with plenty of breathing room. "Warm and handmade" points toward cream grounds, browns, and clay. "Bold and urgent" points toward saturated contrast and a hot accent. Use these words as the test later, when a candidate accent looks good in isolation but wrong for the site.

Then pick one main accent, from something that already exists: a brand color from the logo, or the color of the most important action on the page. If there is no brand color yet, use the color of the thing visitors must do most — usually the primary button. One anchor is enough; two competing anchors pull a design in two directions.

Build the supporting colors from that anchor using the color wheel. An analogous set — the colors next to the anchor — is the calmest choice and the best default for a website. A complementary color, opposite the anchor, gives maximum pop for a call to action, but at equal saturation and lightness it can vibrate against the anchor, so it usually needs desaturating slightly. A monochromatic set, the same hue at different lightness levels, is the safest and also the flattest. FBSubNet's free Analogous Color Palette Generator does this step in the browser: enter a base HEX color and it returns five coordinated colors. For the copper below, #B4652A, the neighbors run from deep rose (#B42A50) through copper into olive gold (#A2B42A). You would not use all five; pick one as the accent and let the others suggest supporting shades.

Add a supporting accent only if it earns its place — for example, copper for primary actions and sage for secondary highlights. It should be clearly different from the main accent in hue or lightness, or visitors will read the two as a mistake. Finally, write the palette down as tokens with roles: page background #FAF6EF, surface #F1ECE1, body text #2B2B26, muted text #6B6B60, primary accent #B4652A, supporting accent #7D8C6F, and a darker copper #8A4A1F for text on the light ground. Stored as named CSS variables, the roles stay attached to the colors in the stylesheet.

The worked example: ivory, copper, and sage

The sample palette is deliberately ordinary: an ivory ground (#FAF6EF), a warm copper accent (#B4652A), a muted sage (#7D8C6F) supporting color, and dark charcoal (#2B2B26) body text — the kind of combination a small business or portfolio might actually choose. Warm palettes are where contrast problems hide, because warm mid-tones look darker to the eye than they measure.

First, the pair the whole page depends on: body text on the ground. Dark charcoal #2B2B26 on ivory #FAF6EF measures 13.20:1, far above the 4.5:1 minimum for normal text and past the stricter 7:1 level as well — and gentler than pure black on pure white, which measures the maximum 21:1. Settle this pair first, because every other decision sits on top of it.

The copper accent was then tested in each of its roles, and this is where the sample stopped behaving. Copper text on ivory, #B4652A on #FAF6EF, measures 4.03:1: it clears the 3:1 bar for large text, so copper headings are fine, but it fails the 4.5:1 bar for normal text, so copper links at body size would be harder to read than the surrounding text. The fix is a darker copper for text roles: #8A4A1F on ivory measures 6.33:1. Keep the brighter copper for fills and large display type; use the darker copper for links and small text.

The button role failed too, and more narrowly. White text on a copper button, #FFFFFF on #B4652A, measures 4.34:1 — just short of the 4.5:1 minimum for normal text. A bold or large-size label would still pass the 3:1 large-text bar, but a standard-size label does not. Three fixes work: darken the button fill to #8A4A1F, where white text measures 6.82:1; set the label in bold at a larger size; or use the brighter copper only for large display buttons.

The sage supporting color produced the clearest failure. Sage text on ivory, #7D8C6F on #FAF6EF, measures 3.33:1: it passes for large text only and fails at body size. Pale greens are the classic version of this problem — elegant in a mockup, unreadable on a real screen, because the eye reads lightness more than hue. The fix is a darker sage, #5C6B4E, which measures 5.31:1 on ivory and passes for normal text. The lighter sage still works as a decorative fill or tag background, anywhere it never carries words. Muted text deserves the same check: the sample's #6B6B60 on ivory measures 5.00:1 and passes, but the default gray many designers reach for, #767676 on white, measures about 4.22:1 and fails, and #999999 only reaches about 2.8:1. If a muted color cannot clear 4.5:1, it is not muted text; it is decoration.

PairRatioVerdict
Body text #2B2B26 on ivory #FAF6EF13.20:1Passes AA and AAA
Dark copper link #8A4A1F on ivory6.33:1Passes AA normal text
White text on dark copper button #8A4A1F6.82:1Passes AA normal text
Dark sage #5C6B4E on ivory5.31:1Passes AA normal text
Muted text #6B6B60 on ivory5.00:1Passes AA normal text
Copper #B4652A as body-size link on ivory4.03:1Fails AA normal; passes AA large
White text on copper button #B4652A4.34:1Fails AA normal; passes AA large
Sage #7D8C6F as body text on ivory3.33:1Fails AA normal; decorative use only

The lesson is not that copper and sage are bad choices. A palette is a set of pairs, and each pair has to be checked in the role it will actually play. Two pairs in this ordinary sample failed as first drawn, and both were fixed by darkening one side of the pair. Keep the bright version for large fills and display type; keep a darker version of the same hue for text — one accent for presence, one for words.

Validate accessibility

The contrast ratio compares the relative luminance — roughly, how much light a color reflects — of a foreground and a background, from 1:1, where the two are identical, to 21:1, black on white. The WCAG thresholds give concrete targets: normal body text needs at least 4.5:1 at level AA, and large text — at least 18 points in regular weight, or 14 points in bold — needs at least 3:1. The stricter AAA level asks for 7:1 for normal text. In practice, aiming for AA on all body text is the standard most teams use.

The fastest way to test a pair is FBSubNet's free Two-Color Contrast Checker, which runs entirely in the browser. Enter the foreground and background HEX values and it reports the WCAG contrast ratio plus pass or fail results for AA normal text, AA large text, and AAA normal text. Run every pair the palette will actually produce: body text and links on the ground, muted text, white text on each button fill, and text on each card surface. A pair that is never tested is assumed — and the worked example shows how often that assumption is wrong.

Test on real pages, not just in the checker. A pair that passes on a flat swatch can still fail in context: text over a photograph, a link inside a tinted card, or a button label in a thin font weight. Check the finished page at the sizes and weights it will ship with. Also check non-text parts: WCAG asks for 3:1 for interface components and meaningful graphics, so focus rings, input borders, and meaningful icons need enough contrast too. And color should never be the only way meaning is conveyed — pair it with a label, an underline, or a shape. For the fuller method of turning one base color into a five-color palette, see How to Build a Consistent Website Color Palette, which works through it with a blue base and the same tools used here.

Choose the accent, then prove it

It comes down to a short sequence: pick the ground first, choose one main accent with a real job, add at most one supporting color, assign every color a role, and check each text pair for contrast. The ivory, copper, and sage sample looked ready at first glance and still had two failing pairs, both fixed by darkening one side of the pair. That is the normal outcome, not a sign the palette was wrong.

Work through your own palette the same way: generate the supporting shades with the Analogous Color Palette Generator, convert formats with the HEX / RGB Color Converter when you need RGB or HSL values, and run every pair through the Two-Color Contrast Checker before the colors reach a page. The result is a small palette you can defend pair by pair — and a website where the accent draws the eye exactly where it should, without ever making the words harder to read.