Referring domains, IPs and subnets, explained

Link-authority tools all count who links to you — but what they count decides how easy the metric is to game. The ladder, from cheapest to hardest to fake:

Backlinks (total links) — every link from every page. One spammy page with 10,000

footer links = 10,000 backlinks. Nearly meaningless raw.

Referring domains — distinct domains that link to you. Much better: 10,000 links

from one blog count once. Still fakeable: register 500 cheap domains, host them on one server, link away.

Referring IPs — distinct IP addresses behind those linking domains. Now those 500

domains on one server count once. Costs real infrastructure to fake.

Referring subnets — distinct /24 networks (C-classes) among those IPs. Majestic's

flagship metric. Faking it means buying address space across many networks — the most expensive link signal to manufacture.

How webrank computes them

We resolve the DNS of our entire domain universe (~293M domains) and join it with the Common Crawl inter-domain link graph: for any domain, the linking domains' A-records give referring IPs, and their /24 prefixes give referring subnets. Subnet diversity feeds the pagerank authority score — log(referring_subnets) rewards breadth across networks, not raw link volume.

These metrics are live on every domain page, and since algorithm v4 subnet diversity is part of the pagerank composite itself (derived weight 0.96 — see methodology). Counts deepen as the DNS pass over the full universe completes (top-ranked domains first — see coverage).

Reading the numbers

referring domains high, subnets low → links concentrated in few networks: PBN / link-farm

signature.

subnets high → organically earned links; hard to buy.
Compare like with like: a local business with 40 referring subnets can be healthier than a

global brand with 400 — scale by niche, not absolutes.