DMOZ (Open Directory Project)

DMOZ was a directory, not a search engine

This is the central correction, and everything else about DMOZ follows from it. The Open Directory Project never ran a crawler, never indexed the text of a page, and never ranked anything by relevance to a query. It was a catalogue: volunteers decided which sites deserved a listing in a given subject category, wrote a one-line human description for each, and filed them into a hierarchy — Arts, Business, Computers, Games, Health, Home, News, Recreation, Reference, Science, Shopping, Society, Sports, plus World for material not in English.

What you got from DMOZ was not a ranked list of pages matching a query. It was a curated shelf of sites on a subject, each with a description a person had written. The name is an abbreviation of directory.mozilla.org, the address Netscape hosted it at, which is itself a reminder of what it was.

The relationship with search engines ran the other way from what most people assume. DMOZ's data was licensed out to search engines, not bought in from them, and that export is the reason it matters to search history at all.

How a listing happened

No crawler, no algorithm, no index of page content. The pipeline had three steps:

1. A site owner, or an editor, suggested a URL into a specific category.
2. A volunteer editor holding permissions over that category reviewed it, accepted or rejected it, and wrote or edited the title and description.
3. The resulting hierarchy was published on dmoz.org and — critically — exported as weekly RDF dumps that anyone could download and reuse.

Two structural consequences fell out of that design. First, throughput was capped by the number of volunteers who happened to be active in a category, which meant commercial categories developed submission backlogs measured in months and sometimes years. Second, nobody could buy their way past the queue: editors were unpaid and answerable to no submitter by design, which was the deliberate contrast with the Yahoo! Directory's paid-review model of the same era.

The licence changed once and it mattered enormously. The original Open Directory License required a specific DMOZ attribution table on every page displaying the data. In 2011 that was quietly replaced with a Creative Commons Attribution licence — genuinely free and GPL-compatible. That change is the reason the data legally survived the shutdown six years later.

What DMOZ powered

DMOZ supplied the directory layer for a substantial part of the early web industry:

• Google Directory at directory.google.com was DMOZ's hierarchy and descriptions, re-sorted by Google's PageRank. Google shut it down in July 2011, removing DMOZ's largest distribution channel.
• Netscape Search and AOL Search used ODP data, which is unsurprising given who owned it.
• Alexa, Lycos, HotBot and a long tail of portals and ISP start pages carried ODP listings at various points.
• Search engines used ODP descriptions as fallback snippets when a page had no usable meta description of its own.

That last one produced a piece of syntax worth remembering: the NOODP robots meta tag. Google, MSN/Live and Yahoo all honoured <meta name="robots" content="noodp"> as an instruction not to use the ODP description in a search result snippet. When DMOZ closed, the tag became dead syntax overnight, though it lingered in page source for years afterwards.

How big it got, and when it started shrinking

DMOZ was never measured as a search engine by StatCounter or comScore, because it was not one. The meaningful scale figures are its own dated counts:

• 5 October 1999: 1 million URLs.
• April 2000: around 1.6 million URLs, passing the Yahoo! Directory.
• 14 August 2000: 2 million URLs.
• 18 November 2001: 3 million listings.
• 3 December 2003: 4 million URLs.
• April 2013: 5.17 million sites in 1.017 million categories, with 97,584 contributing editors cumulatively.
• 31 October 2015: 3.996 million sites in 1.026 million categories.

Read the last two together. The 2015 figure is lower than the 2013 figure: the directory was pruning dead links faster than volunteers were adding live ones. The editor numbers tell the same story from the other side — roughly 4,500 editors at the Netscape acquisition, about 7,330 active editors in August 2006 against 75,151 cumulative contributors by 31 March 2007. A large historic roster and a small working crew is what decline looks like in a volunteer project.

Ownership, money, and the road to closure

DMOZ was founded on 5 June 1998 as GnuHoo by Rich Skrenta, Bob Truel, Chris Tolles, Bryn Dole and Jeremy Wenokur, then at Sun Microsystems. The name drew immediate objections from the GNU project and from Yahoo over the “-hoo”, and it was renamed NewHoo. Netscape acquired it in October 1998, renamed it the Open Directory Project and hosted it at directory.mozilla.org. AOL acquired Netscape in 1999, and Verizon acquired AOL in 2015 — so Verizon was the ultimate owner when it closed.

There was no business model for the directory itself. Editors were unpaid, there was no advertising on dmoz.org, and a listing could not be bought. The value to Netscape and then AOL was strategic: a free, reusable, competitor-denying catalogue of the web at a moment when Yahoo's paid-review directory was the commercial benchmark, plus content to fill AOL and Netscape search products.

Three things then removed that value. A catastrophic server failure from 20 October to 18 December 2006 left editors unable to work for nearly two months, and public confidence never fully recovered. Google Directory closed in July 2011. And directories stopped mattering commercially. By the time Verizon inherited it, nobody had a reason to pay the hosting bill.

A notice appeared on dmoz.org in early March 2017 announcing the closure and giving 14 March 2017 as the date; the site actually became a single landing page on 17 March 2017. Both dates circulate, which is why both are given here.

What happened to the data after 2017

This is the part most write-ups skip or get wrong. Because of the 2011 relicensing, the final RDF dump was legally free to redistribute, and several distinct things grew out of it. As of 19 August 2026:

• Curlie is the official successor, announced in September 2017 and run by former ODP editors. It kept DMOZ's hierarchy and, unlike the archives, continues to accept new listings and recruit editors. Its home-page counters read 3,000,833 sites, 1,031,646 categories, 92,264 editors and 92 languages when checked on 19 August 2026; those are live counters and will drift.
• odp.org is a deliberately frozen static snapshot built from the final DMOZ RDF file by former editors. It states plainly that it is not adding new sites and is not affiliated with DMOZ.org or AOL, and reports roughly 3.86 million listings across about 1 million categories in 90 languages — DMOZ's state at closure. It is ad-free, and it is the better citation for DMOZ as it was; Curlie is DMOZ as it continued.
• dmoztools.net was the widely cited non-profit mirror that hosted the archive and the RDF dumps after 2017. As of 19 August 2026 that domain no longer resolves in DNS. A large number of articles, and Wikipedia references, still point at it. Those links are dead, and this is the single most common stale fact about DMOZ on the web.
• dmoz.org itself still resolves, to an address consistent with registrar or cloud parking when checked on 19 August 2026, but serves no directory. Unrelated commercial “DMOZ directory” sites have appeared on similar domains trading on the name; none is the Open Directory Project.
• The raw RDF dumps — content.rdf.u8 and structure.rdf.u8 — remain in circulation through Curlie, mirrors and academic datasets, and are still used as a labelled taxonomy corpus in text-classification research.

What it was good at, and what it was not

DMOZ did two things extremely well. It produced human-written, non-promotional descriptions of sites at a scale nobody has matched since. And it gave that data away under a licence permissive enough that the corpus outlived the institution by nearly a decade. If you want to know what the web looked like to informed humans in 2005, the ODP archive is a better artefact than any crawl of the period.

It was bad at precisely what it needed to be good at: throughput, accountability and freshness. Submissions sat unreviewed for months or years. Editors were volunteers answerable to nobody, and the recurring allegation — that some favoured their own sites, blocked competitors, or sat on submissions indefinitely — was openly and constantly complained about in webmaster forums, though systematic evidence of corruption is thin. Whole categories went stale when an editor lost interest. The falling listing count from 2013 to 2015 shows the project losing ground to link rot.

Human curation does not scale to a web adding millions of pages a day. DMOZ is the definitive demonstration of that, not a counterexample to it.

Common misconceptions

“DMOZ was a search engine.” It was a directory. It never crawled, never indexed page text, and never ranked results by relevance. It supplied data to search engines.

“DMOZ used Google's results.” Backwards. Google Directory was built on DMOZ's data until Google closed it in July 2011.

“You could pay for a DMOZ listing.” No. Listings were free and decided by volunteer editors — the explicit contrast with the Yahoo! Directory's paid-review model. Anyone taking money for a DMOZ listing in the 2000s was selling something they did not control.

“DMOZ shut down in 2016.” No: 17 March 2017, with the announced date given as 14 March 2017.

“DMOZ's data is lost.” It is not, and the 2011 move to Creative Commons Attribution is why. Curlie continues it, odp.org preserves a static snapshot, and the RDF dumps circulate freely.

“Curlie is DMOZ.” Curlie uses DMOZ's hierarchy and much of its data and is run by former ODP editors, but it is a separate, actively edited project. Its listing count of about 3.0 million as of 19 August 2026 is materially different from DMOZ's roughly 3.86 million at closure.

Frequently asked questions

Sources

• en.wikipedia.org/wiki/DMOZ
• searchengineland.com/rip-dmoz-open-directory-project-closing-270291
• odp.org/about.html
• curlie.org/