Article directories

By Bill Hartzer · Hartzer Consulting · 25+ years in search, domains and expert witness work

What article directories are

An article directory is a website that accepts articles from anybody, publishes them with no meaningful review, and attaches an author box at the end containing the writer's name, a short biography and one or two links. The links were the point. The article was the entry fee.

The model had a plausible surface. Publishers who needed content could take articles from the directory and republish them on their own sites, keeping the author box intact; writers got distribution and links; the directory sold advertising against the traffic. In practice the republishing rarely happened outside of automated scraper sites, the writing was produced purely to carry the links, and the anchor text in the bio box was a commercial phrase rather than a name.

This was, for several years, one of the two or three dominant link tactics in the industry. A single article could be reworded mechanically into hundreds of variants - the practice known as spinning - and submitted across a list of directories in an afternoon. The whole apparatus was demolished by the Panda update, which targeted thin and duplicated content at scale and removed the directories' search visibility, and therefore their traffic, and therefore any reason for anybody to be on them.

The grade here is obsolete, not high risk, and I want to be precise about that. There is no penalty waiting for a site with old directory links in its profile. The category is inert. It is documented here because legacy profiles are full of it, because occasional vendors still sell it to people who have not looked at the market in fifteen years, and because it is the single clearest illustration of a principle this course keeps returning to.

How it worked, step by step

In the past tense, because the pipeline no longer connects to anything.

1. Write one article. Five hundred words on a subject adjacent to the target site, produced to a formula.
2. Spin it. Software substituted synonyms to generate hundreds of nominally distinct versions, so that each directory received unique-looking text. The output was frequently unreadable, and it did not matter, because the audience was a crawler.
3. Write the author box. Two sentences and one or two links, anchored on the phrase being targeted rather than on a name or a brand.
4. Submit at volume. Software posted to directory lists that circulated among practitioners. Some directories had a review queue; most approved automatically.
5. Wait for republication. The theoretical multiplier - other sites picking up the article and carrying the bio box. In reality this happened mostly on scraper sites.

Look at what is absent at every stage: an editor. No one at any point decided that this article was worth publishing to their readers. The directory's business model required accepting everything, because rejecting submissions would have reduced its inventory to nothing. That absence is the whole diagnosis, and I will come back to it.

What it costs in time and effort

Historically almost nothing, which was the appeal. The cost was software, a directory list, and the writing of one article - amortized across hundreds of placements. Cost per link approached zero, and the industry behaved accordingly.

Where the category still has a price today, it is bundled: article submission packages appear as line items inside larger offers, or as fodder for tiered schemes, quoted by the hundred. Nothing about the economics has changed except that the return went from small to nil.

The genuine costs now fall on whoever inherits the mess. A legacy profile carrying two thousand directory links is slower to audit, produces inflated referring domain counts that need explaining, and invites the reflexive suggestion that somebody disavow them all - itself a cost, and speculatively at that scale a risk for no gain.

When it works and when it does not

It does not work, and there are four independent reasons, any one of which would have been sufficient.

Duplication. The model was duplicate content by design. One article across three hundred sites is one document, and search engines resolve it to a canonical source. The other 299 copies are not 299 endorsements; they are replication, which is exactly the mechanism that also neutralizes press release syndication.

No editorial gate. This is the deepest reason and it connects to the principle of relevance. A link is worth something because a publisher, who has a reputation to protect and readers to serve, decided that pointing at you was worth doing. Relevance is not a property of a page's subject matter alone - it is a property of somebody's judgment about their audience. A directory that accepted everything made no judgment at all. There was nothing for a search engine to count, because nobody chose anything.

The sites lost their audience. Panda removed their search visibility, the traffic went, the republishing partners stopped, and the remaining sites became link repositories with no readers. Many simply went offline; the survivors sit unmaintained.

The links are worthless by measurement. Trust Flow of zero across most of the category. And the wider point applies here as everywhere in this section: on my own site, 69.3% of 22,260 referring domains sit at Trust Flow 0, on a profile where no link was ever bought. Zero-trust links accumulate for free, in volume, without anybody's help. Manufacturing more of them adds nothing that the web was not already giving away.

Is there a surviving legitimate version? Yes, and it is not this. Writing for a publication with an editor who can say no is guest posting - a different tactic with its own page, its own risks, and an editorial gate that is the entire source of its value.

Common mistakes

• Confusing article directories with guest posting. The distinction is the editor. If the site would publish anything you sent, it is a directory whatever it calls itself, and a modern content-submission platform with automatic approval is the same product with a new interface.
• Spinning content and expecting anything from it. Mechanically reworded text was detectable long before language models made the whole practice pointless.
• Keyword-anchored bio boxes. The most identifiable artifact in the entire era. A profile whose top anchors are service phrases attached to author boxes is a profile with a date stamp on it.
• Mass-disavowing legacy directory links. They are inert. A speculative disavow at that scale is more likely to catch something you valued than to fix a problem you do not have.

A worked example

This is the clearest single piece of evidence I have on the whole category, and it came out of a routine analysis rather than a hunt.

Running a manufactured link profile through a related sites view - a co-citation report that shows which other sites share referring domains with the one you are examining, and therefore what company it keeps - returned this:

• 1st-in-articles.com
• 1stoparticles.com
• 25000articles.com
• 2articles.com
• 365articles.com
• momarticles.com

Six domains, one naming formula, several of them at Trust Flow 0. Read the names themselves, because they are a fossil record. The leading numerals and the 1st prefix are directory-era thinking, from a period when practitioners believed a name starting with a digit or the word first would sort higher in listings. The 25000 and 365 are volume claims - the selling point was quantity of articles, which is the selling point of a warehouse rather than a publication.

Now the analytical point. A related sites report is showing you sites that share inbound sources with your subject. When that report comes back as a naming series, it is telling you that the profile was assembled from a single submission pipeline, one operator, or one list circulated among many. A site with a genuinely earned profile shares referring domains with its actual industry peers, because the same publishers cover the same sector. This one shares them with a farm.

That is the entire category in one screenshot: near-identical names, no differentiation, no audience, no trust, all showing up together because they were always one thing wearing several domains.

How to measure it

Nothing to measure on the performance side. Two things worth doing on a profile that contains them.

Segment and exclude. Identify the directory portion - the naming series makes this unusually easy - and break it out of your headline referring domain count. Report the number above a trust threshold alongside the total, so nobody is ever in a position to quote an inflated figure back at you.

Read them as dating evidence rather than as links. A large directory cohort tells you when a site's link building happened and who was doing it. If you have taken over a profile, that is genuinely useful context: it tells you which era's assumptions are baked into the site, and it usually predicts what else you will find - keyword-anchored bio boxes, spun content still sitting on the site itself, and a set of pages built for phrases nobody searches any more. The links are worthless; the archaeology is not.

What not to do is spend a week building a disavow file. Unless a manual action specifically names these links, leave them. They are decaying on their own - the median referring domain on my own profile stops linking after about 1,080 days, and Trust Flow 0 domains stop at a median of 859 - and directories lapse faster than most.

The verdict

Article directories are obsolete. Not dangerous, not penalized, not worth cleaning up - simply gone. Panda removed the traffic that made the sites worth being on, the duplication meant the links were never independent endorsements, and the absence of any editorial gate meant there was no judgment for a search engine to count in the first place.

The reason this page exists is the principle underneath it, which outlives the tactic by a wide margin. A link is worth something because somebody with a reputation chose to give it. Every mechanism that removes the choosing - open submission, automated approval, self-service publication, a directory that takes everything - removes the value at the same time, no matter how many domains the result is spread across.

If you want the modern version of what article directories promised, it is writing something good enough that an editor with standards wants it. That is slower, it is refusable, and it is the only part of the old model that was ever load-bearing.