Brand & vision

Origin

arXiv was built in 1991 to break the hold commercial publishing had on science, and thirty-five years later it is still free, still run for researchers, and still the place physics, mathematics and computer science go first.

Brand statement

Quoted in full. This is the text under discussion, not a summary of it.

arXiv is the light side of the internet. We were built in 1991 to break the hold commercial publishing had on science. It was too slow, too expensive, too much gatekeeping. We aim to be the opposite and offer the most useful website on the internet, best-in-class for speed, readability, accessibility, and efficiency. The rebellion succeeded, but we still play by its rules: always free, answerable only to researchers, zero ego, and no dark patterns. The light side, indeed.

This is the text under discussion, not a summary of it.

Where arXiv sits in the field

Two audits measured arXiv against the research-publishing ecosystem — eighteen platforms for brand colour, nineteen article pages for what those pages actually do to a reader. The conclusions, in brief.

The field is one colour

12 of 18 platforms use a dark blue as their primary brand colour, and eight of those twelve sit in a band so narrow that Springer Nature, IEEE and Taylor & Francis are separated more by lightness than by hue.

arXiv takes the same hue and goes light: Open Blue is thirty points lighter than the lightest platform in the field, and paired with a warm near-black where everyone else pairs with black.

Nobody else leaves readers alone

0 advertising or tracking networks on an arXiv paper page. Only arXiv and the two overlay journals can say that. Every commercial platform measured loads between eleven and thirty-eight third-party domains.

IEEE sets twenty-eight cookies on an article body it will not show without a subscription. Springer runs ten tracking networks on a page that costs USD 39.95 to read.

The paper comes first

108px from the top of the page to the paper title on arXiv — the highest of any platform measured. The field average is more than three times that, and two platforms push the title beyond 600px.

A first-time visitor to arXiv is contacted by two outside domains, both code libraries, and given one cookie.

Full research: competitor brand colour audit · clicks-to-content audit

What this looks like on the page

The abstract page on the left is what arXiv serves today. On the right is the Phase 1 proposal, which merges the abstract and the full paper into one page. The right-hand panel is the live mockup, not a screenshot, so it stays current as the work does.

The arXiv abstract page as it looks today: a dense two-column layout with the paper title near the top and links to the PDF and HTML versions in a side panel.

Today — the abstract page. The reader reaches the paper itself in one more click.

Proposed — abstract and full paper on one page. Open the full mockup

Each change below is there because a brand principle asked for it, not because it looked better.

Change	What it does for the reader
The paper is on the page	Today the abstract page sends the reader to a separate address for the full text. Nearly every open-access platform measured already serves the paper on arrival; arXiv is the exception. The merged page removes that step. Principle: the interface gets out of the way.
One click to cite	Copying a citation today means opening a panel and selecting the text by hand. The proposal puts a copy button next to the citation. No platform measured does better. Principle: speed and efficiency are different promises.
Still no advertising, still no tracking	The redesign adds no advertising surface, no engagement metrics, no view counts, and no third-party tracking. This is the part of the brand that is easiest to lose by accident and hardest to win back. Principle: no metrics, no gamification, no dark patterns.
Structure a machine can read	The current abstract page has seven top-level headings where there should be one, which is the worst heading structure in the sample. The proposal has one, with real landmarks underneath it — better for screen readers, and better for every tool that reads arXiv rather than browses it. Principle: readable by people and machines, equally.
Light where the field is dark	Open Blue and a warm near-black, against a field of corporate navy. The visual lightness is the positioning, not a decoration on top of it. Principle: lighter than the establishment.

One honest note: the proposal currently places the paper title lower on the page than the abstract page does, because it carries a prominent link back to the abstract while both pages exist. If the merged page becomes the default, that link retires and the title sits higher than on any platform measured, arXiv included.

Voice

Honest, plainspoken, accessible, and slightly mischievous.

Honest — We say what is true, including the unflattering parts: open issues, known limitations, “this is a prototype”. No marketing gloss, no overclaiming.
Plainspoken — Write as real people, for real people. Our users live around the world with varying familiarity with English, so we avoid slang, contractions and jargon. When we are unsure, Simplified Technical English is the guide.
Accessible — Literally (WCAG, screen readers, low bandwidth, JavaScript off) and tonally. Nobody should feel they need a secret handshake to use arXiv.
Slightly mischievous — A dry wit, a wink, the smileybones. Never cute for its own sake, and never at the reader's expense. Just enough humour to signal there are humans here.

The voice applies to everything that has words: interface copy, empty states, error messages, docs, alt text, commit messages. When in doubt, read it aloud. If it sounds like a brochure or a legal notice, rewrite it. Text that carries a task — error messages, form help, instructions — follows the stricter rules in STYLE.md.

What this means for the design

The brand is not decoration. Each principle names the policy or pattern it drives, so the line from brand to pixel is traceable.

1. The interface gets out of the way

Zero ego means the design never competes with the science. Chrome recedes as the reader goes deeper; navigation stays permanently short and resists link-creep (DESIGN-POLICIES, nav cap). The most useful website is the one you stop noticing.

2. No metrics, no gamification, no dark patterns

Answering only to researchers is literal: arXiv shows no view counts, download counts or citation counts on public pages, and does not rank or promote papers. We do not manufacture urgency, we do not farm attention, and toasts stay near zero. If a feature's job is to drive engagement rather than serve a research task, it does not belong.

3. Lighter than the establishment

The light-side metaphor is in the palette: warm Repository Brown and light tints where commercial publishers go navy-and-corporate. Open Blue and Access Lime are bright, not sombre. The visual lightness is the positioning — see color-mapping.md.

4. Readable by people and machines, equally

Best-in-class readability includes machines: clean semantic HTML, real headings and landmarks, content present before JavaScript runs. Co-equal HTML and PDF serve the same goal. Open strategic question: whether to move from co-equal to HTML-first. A brand decision waiting on leadership, not a design-system one.

5. Accessibility is a brand value, not a checkbox

“The light side” and “open to all” mean WCAG 2.1 AA is the legal floor (the Accessible Canada Act) and 2.2 AA is what we build toward: spoken math, keyboard-reachable everything, forced-colors and reduced-motion support, usability with JavaScript off. The accessibility rules in DESIGN-POLICIES are this principle made concrete.

6. Honest, plainspoken copy — with a wink

Interface text is direct and never overclaims. The mischief lives in small, optional places — the smileybones, a dry empty state — never in anything load-bearing, and never in anything a stressed researcher has to parse.

7. Built to last by a small team

arXiv is daily infrastructure maintained by a very small team, so designs must be low-maintenance: no patterns needing manual upkeep, no per-paper hand-tuning, citation formats generated from metadata. Durable and boring-where-it-counts beats clever and fragile. This too is zero ego.

8. Speed and efficiency are different promises

Two distinct pillars, not synonyms. Speed is how fast a paper moves from submission to announced — a throughput commitment. Efficiency is how few clicks a researcher needs to get what they came for. Measured against seventeen platforms on 2026-08-12, arXiv does not lead here: nearly every open-access publisher puts the full paper on the landing page, while arXiv sends the reader to a separate /html/ page. The directive that follows: count the clicks to every core task — download, cite, find the DOI, switch version — and minimise them. The common cases belong on the surface, not in menus.