Social bookmarking

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

What social bookmarking is

Social bookmarking is the practice of saving a web page to a public, shared account so that other people can find it - your browser's bookmarks, but on a website, tagged, and visible to strangers. The category had a genuine moment. In the second half of the 2000s, services where users saved and tagged links were a real way to discover things, they had real audiences, and a page that caught on there received meaningful traffic.

The SEO product is a different animal that borrowed the name. It is a submission service: you supply a URL, a title, a description and some tags, and a vendor submits them to a list of bookmarking sites. Packages are sold by the hundred. The deliverable is a spreadsheet of URLs where your link now appears.

Crucially, the list is not made up of the sites people actually used. The genuine services either shut down, changed hands and changed model, or turned into something else entirely - the survivors are recognizable today as content platforms and social networks rather than bookmark managers. What the submission packages target is a shadow population: hundreds of near-identical sites running the same off-the-shelf bookmarking script, built for no purpose other than to accept SEO submissions.

I have graded this obsolete rather than high risk deliberately. Nobody is being penalized for bookmark submissions in 2026. There is no downside beyond the fee. The category is not dangerous; it is simply over, and describing it as a risk would misstate what a reader is dealing with.

How it works, step by step

The process is worth laying out because each stage is where a check should be applied and never is.

1. A target list is assembled. Sites running public bookmarking software that permit open registration and do not moderate submissions. The list circulates among vendors and changes slowly, because these sites rarely die properly - they linger, unmaintained, for years.
2. Accounts are created in bulk. The same automation used for other bulk link products, against disposable email addresses.
3. The URL is submitted with metadata. Title, description, tags. The description is the field where keyword stuffing traditionally happened, and it still shows up in profiles as a fossil.
4. The bookmark page is published. A thin page containing your title, your description, your link and a list of unrelated submissions from other buyers.
5. An indexing push is attempted. Because a page that is not in the index cannot pass anything, vendors bolt on pinging or submission services to try to get the new pages crawled at all.
6. A report is delivered - the actual product. Accurate about what was done, silent about whether it means anything.

Note what is missing from that sequence: a person. No human being read your page, decided it was worth saving, and saved it. That is the difference between a bookmark and a bookmark submission, and it is the whole story.

What it costs in time and effort

This is at the very bottom of the price range for link products - quoted per hundred, per thousand, or thrown in free as a bonus alongside a larger order. The cost drivers are account creation volume and whether an indexing service is included. Nothing else, because there is nothing else.

Price at that level is diagnostic. In any functioning market, a link costs something because the seller gives something up: a publisher selling a placement risks editorial credibility. A bookmarking clone gives up nothing, has nothing to give up, and no audience to disappoint. There is no seller of value here, only a fee for the labor of filling in forms.

The real costs land after delivery. Report contamination is the main one - a few thousand bookmark submissions in a profile inflates referring domain counts, and every subsequent audit has to strip them out before it can say anything useful. Then there is the false confidence: a team looking at a spreadsheet of 500 delivered links has no urgency about earning five real ones, and that delay is far more expensive than the invoice.

When it works and when it does not

It does not work, and the diagnosis is unusually clean because the three quality checks that matter all return the same answer.

Indexation. A link on a page that is not in the search index passes nothing. Bookmark pages are thin by design - a title, a description, a link, and a sidebar of unrelated submissions - and duplicated across hundreds of clone sites carrying the same submissions. That is precisely the material an index excludes. Run the delivered URLs through an exact-URL query and most are not there.

The link attribute. User-submitted links carry nofollow or the user-generated content attribute wherever the software is maintained. Where they do not, it is because the site is abandoned - which sends you back to the indexation problem.

Traffic and liveness. Nobody visits these sites. Not few - none, in any analytics I have looked at. And the pages do not last: accounts get purged, sites go offline, the script breaks and is never fixed. Against a baseline where 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, bookmark submissions decay faster still.

Then the argument that closes the category. On my own site, 69.3% of 22,260 referring domains are at Trust Flow 0, and not one link was ever bought. The web produces zero-trust links in enormous quantity, unprompted and free, at a volume no submission package can approach. Paying for more of the commodity the web hands out for nothing is the clearest example of buying something that is not scarce.

What does still work is the thing bookmarking used to be a version of. Sharing a genuinely good page where an audience exists - a subject community, a professional network, an aggregator with real readers - sends real people. Those links are nofollowed too, and it does not matter, because the value is the audience and the coverage that sometimes follows. That is distribution, and it shares nothing with a submission package but a distant family resemblance.

Common mistakes

• Accepting the delivery report as the outcome. The spreadsheet is a list of URLs, not a list of links that do anything. Nobody who buys these checks all of them, and the vendors know it.
• Confusing platforms with clones. Sharing on a network where your audience is present is worth doing. Submitting to a hundred script-driven clones is not the same activity performed at scale.
• Using them as tier fodder. A page with no trust has nothing to pass at any depth in the graph.
• Buying a package because it is cheap. Cheap is not the same as good value when the return is exactly zero. Any multiple of zero is zero, and this is the tactic where that arithmetic is literal.

A worked example

Take a delivered package of 200 bookmark URLs and audit it properly. I recommend doing this once, on your own money, because a spreadsheet turning red in front of you is more convincing than a paragraph.

1. Query each URL exactly. Not the domain - the specific bookmark page. Count the ones present in the index. In my experience it is a minority, often a small one.
2. Read the attribute. Fetch the source and check the rel value on your link. Nofollow or ugc on anything maintained.
3. Check Trust Flow on the hosts. Trust Flow is computed from the trustworthiness of the sites linking in, so a wall of zeros is not a tool limitation - it is the metric correctly reporting that nothing of consequence links to these sites.
4. Look at the page itself. What surrounds your bookmark is submissions from other buyers in unrelated verticals - loans beside dog grooming beside a casino. That is what your listing sits inside, and why nobody reads it.
5. Check analytics for referral sessions. Expect none. I have never seen otherwise.
6. Re-run the whole audit in six months. Count how many URLs still resolve. The attrition is the final answer.

Zero for three on the first three checks, on essentially the whole batch, is the standard result. That is the category described in one afternoon.

How to measure it

There is no performance to measure. Measure exclusion and, if you insist on activity in this direction, measure the right substitute.

Exclusion. Segment bookmark hosts out of your link reports. If a referring domain count is inflated by a few hundred clone sites at Trust Flow 0, the number describes a purchase rather than an achievement. Report domains above a trust threshold, with the zero-trust portion broken out beside it so the reporting stays honest rather than merely flattering.

The substitute. If the underlying instinct was to get a page in front of people, measure that instead: referral sessions from platforms where you shared it, engagement on those platforms, and whether any of it produced a mention or a link somebody chose to give you. Those are real numbers from a real activity.

The three-question filter. The lasting value of this page is the test it demonstrates. Before committing to any link opportunity, ask whether the page is indexed, whether the link is followed, and whether the page has any traffic at all. Thirty seconds per prospect. It disqualifies whole categories, this one included, and it will save you more money than any tool you subscribe to.

The verdict

Social bookmarking as an SEO tactic is finished, and it is finished quietly. There is no penalty, no manual action, nothing to clean up, and no reason to disavow anything. The genuine services that once had audiences changed model or closed; the clone sites built to absorb submissions never had a reader in the first place. What remains is a product with a price list and no mechanism.

The honest framing is that this was never really a link tactic. It was an attempt to simulate the thing that made bookmarking interesting - somebody deciding a page was worth keeping - by automating away the deciding. Once you automate away the judgment, there is nothing left for a search engine to count.

If a proposal or an audit still recommends bookmark submissions, treat it as a dating test for the document rather than a plan, and read the rest of it with that in mind.