Links in a printed book, course handout or report: references that still work in ten years

A web page you cite in a blog post can be fixed the day the link breaks. A web page you cite in a printed book, a course handout, an annual report or a museum label can't. The paper will sit on a shelf for ten or twenty years, and every address on it will be typed by hand, by someone who has no way to ask you what you meant.

This guide is about writing those links so they keep working, or at least so the reader can still find what you were pointing at when they stop. It's for authors, teachers, small publishers, charities writing reports, and anyone who prints a reference list.

How fast cited links go bad

The studies all point the same way.

- In 2024 the Pew Research Center checked a sample of pages from each year between 2013 and 2023. 38% of the pages that existed in 2013 could no longer be reached, and a quarter of all pages from the whole decade were gone. Most of them weren't on dead websites: a single page had been removed from a site that was otherwise fine. Pew also found that 54% of Wikipedia articles have at least one dead link in their references.
- A 2014 Harvard study by Jonathan Zittrain, Kendra Albert and Lawrence Lessig found that 50% of the links in US Supreme Court opinions, and more than 70% of those in the Harvard Law Review and other law journals, no longer led to the material that had been cited.
- In science, a 2014 study of articles published between 1997 and 2012 found reference rot in one in five of all articles, and in seven out of ten of those that cited web pages at all.

Then there's the quieter failure. A 2016 follow-up by the same research group checked links that still worked and compared the page with an archived copy from the time it was cited. For more than three out of four, the content had changed. The address still opens, but the paragraph you quoted is gone, the figures have been updated, or the page now sells something. Researchers call that content drift, and it's harder to spot than a 404, because the reader thinks the link works.

Pew saw a third kind of change too: 32% of links on news sites, and 16% on government sites, now redirect somewhere other than where they first pointed. Sometimes that's the same article at a new address. Sometimes it's the home page.

So assume that by the time your book is ten years old, about a third of its web references will be dead, and many of the rest will show something different from what you saw.

What to print for each web reference

A link alone is the weakest possible reference. If it breaks, the reader has nothing. Print enough that a reader can find the page without it:

- Who and what: the author or organisation, the title of the page or report, and the site's name. A reader can search for that years later.
- When: the publication date if the page has one, and the date you read it. The date you read it is what a reader needs to find the right archived copy.
- The full original address, even if it's long. It's the key every web archive uses. A reader who has only your short link has nothing to look up if the short link dies.
- A DOI if the source has one. Journal articles, many reports and some datasets have a DOI, printed as https://doi.org/10.…. The publisher keeps it pointing at the current copy, so cite it instead of the publisher's own page address.
- An archived copy: the address of a snapshot taken when you cited the page (see below).

Citation styles differ on details. Chicago, for example, prefers a DOI to a URL and asks for an access date only when the page has no date of its own. If a publisher gives you a style, follow it, and add the archived copy in whatever way it allows; most styles accept one.

Save a copy when you cite

The single most useful habit is to save an archived copy of every page the day you cite it. Later, when the page changes or dies, the reader can still see what you saw.

- The Wayback Machine (web.archive.org) is free. Paste the address into its Save Page Now form, wait a few seconds, and you get a dated snapshot address like https://web.archive.org/web/20261018093015/https://example.org/report. Anyone can open it.
- archive.today (archive.ph) does the same, and sometimes captures pages the Wayback Machine can't.
- Perma.cc, run by Harvard's Library Innovation Lab, makes copies meant to last for citation. It's what courts and law journals use. Individuals get a few free links and then pay; people at a member library or university often get it free through their institution.

Archives aren't permanent either. In October 2024 the Internet Archive was taken offline by an attack, and the Wayback Machine came back read-only, without saving new pages, after about five days. Some pages refuse to be archived, and anything behind a sign-in or a paywall usually can't be. For a reference that matters, save it in two archives, and keep your own PDF of the page with your notes.

A copy of a page you saved yourself can't be printed in your reference list, but it lets you answer the reader who writes to you in 2034.

Typing a link from paper

Readers of print type addresses by hand, which changes what a good link looks like.

- Long addresses get mistyped. https://www.example.org/sites/default/files/2026-03/Annual%20Report%20FINAL%20v3.pdf is a full line of text with three places to get wrong. Readers of the reference list can cope; readers of a handout in a classroom mostly won't bother.
- Line breaks hurt. Never add a hyphen where an address breaks across lines, because the reader can't tell whether it belongs. Break before a slash or a dot instead, and check the proof, since typesetting software likes to hyphenate.
- Tracking tags make it worse. Strip ?utm_source=... and similar from anything you print as a reference.
- Ambiguous letters. In a typeface where l, I and 1 or O and 0 look alike, a short code like bit.ly/3lI0O is a guessing game.

This is where short links earn their place in print, in running text, handouts and slides. b59.link/ch4-pew is easy to type, can carry a name that tells the reader what it is, and can be repointed: when the original page dies, you point the short link at the archived copy and the printed book works again. A short link also counts its clicks, which tells you which references readers actually follow.

But a short link is one more thing that can break. If the shortening service closes, every short link in your book can die at once. Google switched off goo.gl links that had shown no recent activity, which is exactly the kind of link that sits in an old book, as we covered in when a link shortener shuts down. And a short link hides its destination, so a reader can't tell from the page what it points at.

The arrangement that holds up best uses both:

- In the reference list: the full reference with the original address (or DOI), the date you read it and the archived copy. This is the permanent record and needs nothing from you to work.
- In the text, handout or slides: a short, readable link for convenience, either on a short-link service you trust or on your own domain (yourbook.com/r/4), so you can repoint it.
- A companion page for the book: one page on your site, such as yourbook.com/links, listing every reference with its current address. Print that one address in the front matter. It's the one link you most need to keep alive, so put it on a domain you own and renew.

QR codes in print

A QR code next to a reference in a handout or a report saves the reader from typing at all. The same rules apply: the code should hold a short link you can repoint, not the long original address, so it stays small and keeps working when the page moves. Print the short link in text beside it for readers who can't scan. Our printing guide covers the size and the quiet zone; a code under about 2 cm wide is asking for trouble on uncoated paper.

Check the links once a year

Links in print don't fail all at once. They go one by one, and nobody tells you. A yearly check catches them while you can still do something:

1. Keep the reference list in a plain list or spreadsheet, one address per line, alongside its archived copy and short link.
2. Once a year, run the addresses through a link checker. On b59.link the link list checker takes up to 25 addresses at a time and sorts them into works, worth a look, dead, no answer and refused, with a file you can download. Nothing is stored.
3. Open everything marked "worth a look" by hand. That's where redirects to a home page and content drift show up: the address answers, but not with what you cited.
4. For each dead or drifted reference, repoint its short link to the archived copy, update the companion page, and note it for the errata or the next printing.
5. Keep the check's file with the date, so next year you can see what changed.

For a book with hundreds of references this takes an afternoon a year. For a handout with ten links, it takes five minutes before each term starts.

Before it goes to the printer

A short checklist for the proof stage:

- Every web reference has an author or organisation, a title, a date read, and the full address or DOI.
- Every web reference has an archived copy, saved on the day it was cited or at the latest at this stage.
- No tracking tags, no session IDs, no addresses that only work while signed in.
- No hyphens added inside addresses at line breaks.
- Short links and QR codes point at pages you've opened on a phone in the last week.
- The companion page exists and its address is printed in the front matter.
- Someone has a reminder in their calendar for the yearly check.

Sources: Pew Research Center, When Online Content Disappears (May 2024). Zittrain J, Albert K, Lessig L, Perma: Scoping and Addressing the Problem of Link and Reference Rot in Legal Citations, 127 Harv. L. Rev. F. 176 (2014). Klein M, Van de Sompel H, et al., Scholarly Context Not Found: One in Five Articles Suffers from Reference Rot, PLOS ONE 2014. Jones SM, Van de Sompel H, et al., Scholarly Context Adrift: Three out of Four URI References Lead to Changed Content, PLOS ONE 2016. Perma.cc's FAQ on accounts. The Internet Archive's October 2024 outage and read-only return as reported by Computing.