The Links-on-Page Column and What It Supports

Most link exports carry a column counting the links on the referring page — labelled links on page, external links, or outbound links depending on the vendor. It gets used as a divisor: a link from a page with 8 links is worth more than one from a page with 400, so rank the list by that.

The intuition has a real ancestor and no measurable present. Worth separating the two, because the column does have a good use and it isn’t this one.

Where the divisor idea comes from

The original published PageRank formulation divided a page’s score across the links leaving it. That’s a documented property of an algorithm from an academic paper, not a guess — and it is also the reason the folklore is so durable, because it was once literally the arithmetic.

What replaced it is unpublished. Whether a modern ranking system treats outbound-link count as a divisor, a weak input, a feature among thousands, or nothing at all is not disclosed, and anyone who tells you the exponent is inventing it. This site’s standing position applies: no claims about search-engine weights.

So the honest version is: the number is real, its historical significance is real, and the conversion rate from “400 links on the page” to “your link is worth 1/400th of something” does not exist in any form you can measure.

What the column actually counts

Before using it as a signal, find out what it’s a count of. Vendors differ, and the differences are large:

External only, or all anchors? A count that includes internal links is dominated by navigation. A 40-item nav plus a footer puts every page on the site at 60-plus before any content exists.

Unique targets or instances? A page linking to the same URL three times may count as one or three.

Are nofollowed links included? Some counts include every anchor regardless of attributes, some exclude nofollowed ones, and attribute parsing has its own caveats — see how tools count nofollow, UGC, and sponsored links.

Rendered or raw? Links injected by JavaScript exist for a crawler that renders and not for one that doesn’t.

When was it counted? The figure is from the last crawl of that page, on a schedule you can’t see. The page may have been repaginated, pruned, or turned into a listing since — the same timestamp problem as in what first seen and last seen actually mean.

Five decisions, all defensible, none of them the same across vendors. Which means the column is not comparable between tools even before you decide what it implies.

The job it does well

Sorting a large referring-page list to find pages that are structurally lists of links.

That’s the pattern worth catching, and it’s a shape rather than a threshold. A page with an outbound count in the hundreds, a title that’s a category name, and a dozen of your competitors alongside you is a directory listing. A page with an outbound count in the hundreds because it’s a long, heavily-cited reference article is not. The count alone can’t tell them apart; the count plus opening five of them can, in about two minutes.

Used this way the column is a triage aid: it moves the pages most likely to be aggregations to the top of a list you were going to sample anyway. That’s a genuine efficiency, and it costs nothing.

Three ways it gets misused

As a quality score. “Links from pages with fewer than 20 outbound links” as a filter on a prospect list. It will discard well-cited resource pages and keep thin pages that happen to link out rarely. The number isn’t measuring editorial care.

As a ranking key for reporting. Sorting your existing profile by outbound count and presenting the top of the list as your best links. It presents an artefact of page structure as a finding.

In arithmetic. Computing a “diluted authority” figure by dividing a page score by its outbound count. The output has no unit and no referent, and it will be quoted downstream as if it did. Related in spirit to the arithmetic problems in page-level authority metrics: a score that’s ordinal and compressed doesn’t survive being divided.

A hypothetical illustration of how badly the ordering can fail: three referring pages with outbound counts of 6, 52, and 310. The first is an abandoned personal page with no inbound links of its own. The second is a well-cited industry explainer. The third is a conference programme listing every speaker. Sorted by the column, the ordering is exactly backwards from any reading you’d defend after opening them. (Illustrative figures, not measurements.)

What to look at instead

If the question is “is this a link worth having,” the observable inputs are:

• Whether the referring page has its own inbound links, which is what page-level scores estimate.
• Whether the page is fetchable and free of exclusion directives — the check in whether the linking page is indexed.
• What the anchor and surrounding context say, as data rather than as a target ratio, per anchor text distribution as a dataset.
• Whether the link sends anyone. Referral sessions are measured directly rather than estimated, which makes them the strongest evidence in the whole stack — the point of how to tell whether a link did anything.

None of those is a divisor either. They’re just observations that survive being written down.

Reporting practice

If the column appears in a report, name what it counts and where it came from: “outbound links per referring page, [vendor] definition (external only, unique targets), last crawl dates ranging 3–41 days.” If you can’t state the definition, don’t put the number in front of anyone.

Better: don’t report it as a metric at all. Use it as a sorting key during analysis, then report what you found by opening the pages — “9 of the top 30 referring pages by outbound count are directory listings” is a finding, and it’s the kind of composition statement that holds up, per building a link report you can defend.

What nobody outside the vendor knows

Each vendor’s exact counting rule and whether it has changed, whether rendering is applied, and how stale any individual row’s count is. And on the other side: whether outbound-link count enters modern ranking systems at all, in any form. Two closed boxes, one column between them.

The column is fine. It’s a description of a page’s structure, cheaply available, useful for finding aggregations. It becomes a problem the moment it’s treated as a price.