What makes a link valuable
Google has not published a scoring formula, and does not need to for the practical judgement to be clear. The factors below are drawn from its documentation on links, its spam policies, and long industry practice; where something is inference rather than documentation it is marked.
- The linking site's standing. A link from a site that is itself well linked, trusted and established carries more weight than one from a new or obscure site. This is the original logic of link analysis and remains documented in outline.
- Topical relevance. A link from a site or page about the same subject is a stronger reference than one from an unrelated site. Industry practice treats this as central; Google's documentation describes links as signals of relevance as well as authority.
- Editorial placement. A link in the body of an article, chosen by the author, is a reference. A link in a footer, sidebar, author box or widget that repeats across thousands of pages is a template artefact. Google documents that it treats site-wide and boilerplate links differently (inference from its guidance on widget links and link spam).
- The linking page's own state. An indexed page with its own traffic and links passes more than an orphaned page nobody visits. A link on a page the engine cannot crawl passes nothing.
- Uniqueness of the referring domain. The first link from a domain is worth more than the hundredth (industry observation; consistent with Google's statements about site-wide links).
- Attributes. A link marked
nofollow,sponsoredorugcis a hint that it should not pass signals; Google documents that it may still use such links for discovery and understanding. - Context. The words around the link, and whether the page's content supports the reference, matter for how the link is understood.
Third-party "authority" scores approximate the first factor from a tool's own crawl. They are fine for sorting prospects and comparing competitors; they are not a Google metric and not a ranking factor.
Anchor text
The anchor is the clickable text of a link. Google documents that it uses anchor textThe visible, clickable words of a link. to understand what the linked page is about, for internal and external links alike (the internal linking guide covers the internal side, which the site controls).
A natural link profile contains a mix:
| Anchor type | Example | Typical share in an earned profile |
|---|---|---|
| Brand | "Example Co" | Large |
| Page title or close variant | "guide to loft insulation" | Common |
| URL | "example.com/guide" | Common |
| Generic | "here", "this article", "read more" | Common |
| Partial match | "insulation advice from Example" | Some |
| Exact match | "loft insulation installer" (the page's commercial target) | Small |
| Image | the image's alt text | Some |
A manufactured profile inverts this: exact-match commercial anchors dominate, because whoever placed the links chose them. Google's link spam policy names "keyword-rich, hidden or low-quality links" in articles and press releases as examples of link spam, and its spam systems are trained on exactly such patterns. The practical rules: when a site controls an anchor (a guest articleAn article written for another publication's readers under the author's name and published because the editor wants it., a partner listing), use the brand or a descriptive phrase that reads naturally in the sentence; when a site does not control it, do not ask for a keyword anchor; and review the profile for anchor distribution as part of the audit below. The guest article tutorial applies the first rule to the bio and body links of a contributed piece.
Nofollow, sponsored and ugc
Google introduced the attributes to let sites qualify links. Its documentation says:
rel="sponsored"for links that are advertisements or paid placements;rel="ugc"for links in user-generated content such as comments and forum posts;rel="nofollow"where the site does not want to vouch for the link or imply endorsement;- all three are treated as hints, meaning the engine may still use them for crawling and for understanding, rather than as directives to ignore the link entirely.
For an inbound link, an attribute means the link is unlikely to pass ranking signals and might still help discovery. For outbound links, the documented obligation is to mark paid links as sponsored; failing to do so is the paying site's and the selling site's violation. Many publications nofollowA link attribute asking search engines not to treat the link as an editorial endorsement. all outbound links by policy, which is why digital PR is judged on coverage and demand, not followed links alone. The sending side, which of your own outbound links need each value and how to keep citations working, is covered in the guide to outbound links and link attributes.
Reviewing a link profile
A link audit answers three questions: where does the site's authority come from, is any of it at risk, and what does the profile say about the next programme. Sources: the Links report in Search Console (GSCGoogle's search performance and indexing tool.), which lists top linking sites, top linked pages and top anchor text from Google's own data; and a backlink tool for history, attributes and competitor comparison.
- Export referring domains, linking pages, target pages and anchors.
- Classify referring domains by relevance (same field, adjacent, unrelated), standing (established, obscure, suspicious), and type (editorial, directory, partner, forum, syndication).
- Read the anchor distribution for the site and for the most-linked pages. Flag exact-match dominance.
- Look for patterns that match the spam policy: clusters of links from sites that exist only to link, article networks with identical anchors, paid placements without attributes, links in footers across many sites, sudden volumes from one country or platform.
- Check which pages receive links against which pages need them; most sites find their authority concentrated on the homepage and a few old pages, with the pages that need to rank under-linked. This is a link building and internal linking finding, not a risk.
- Compare with competitors who rank for the priority queries: which referring domains do they have that the site lacks? This becomes the prospect list.
- Decide on risk per the next section. For most sites the answer is that nothing needs doing.
Reading an anchor distribution
A legal practice's audit shows its most-linked service page with a profile where exact-match anchors for its main commercial query make up a share no editorial process would produce, all from a cluster of general-interest blogs with no connection to law and near-identical article structures. The practice's previous agency had bought them. The remaining profile — local press, the law society directory, a university clinic's resource page, client sites — is healthy and relevant. The decision: document the bought cluster, submit a disavow of those domains because the history is known and manipulated, stop the programme, and redirect effort toward the relevant sources the healthy part of the profile already shows. The service page's rankings, which had been unstable, are watched over the following months rather than expected to jump.
Risk and the disavow tool
Google's spam policies cover link spam; the consequences are algorithmic (systems that ignore or devalue manipulated links, which Google says is the usual outcome) and manual actions, reported in Search Console, which can suppress a site until the issue is fixed and a reconsideration request is accepted.
The disavow tool tells Google to ignore specified links or domains when assessing the site. Google's documentation is explicit that it is an advanced feature, that most sites do not need it, and that it is intended for sites that have a manual action for unnatural links or that know they have a history of paid or manufactured links. The industry habit of "toxic link" clean-ups based on tool scores has no support in that documentation and can remove links that were helping. The decision table:
| Situation | Action |
|---|---|
| Manual action for unnatural links to the site | Remove what can be removed; disavow the rest; document; file reconsideration |
| Known history of bought or manufactured links, no manual action | Disavow the known manipulated set; stop the programme |
| Random spam links the site never sought | Nothing; Google documents that its systems handle these |
| A tool reports "toxic" links with no known history | Review manually; usually nothing |
| Negative SEO suspected (sudden wave of spam links) | Document; usually nothing; disavow only if a manual action follows |
How quality links are earned
Everything above describes judgement; the activity is in the methods guide and the digital PR guide. The common thread is that the links worth having come from sites with their own readers, in content those readers see, with anchors their authors chose. That also describes the reputation evidence that quality raters are asked to look for under experience, expertise, authority and trust (E-E-A-TExperience, Expertise, Authoritativeness, Trust.): what other sites say about this one.
Common mistakes
- Chasing tool scores. A score is a proxy; relevance and editorial placement are the substance.
- Requesting exact-match anchors. The pattern spam systems look for.
- Routine disavowing. Discarding links that help, against Google's own guidance.
- Ignoring where links land. Authority on the homepage does not rank the service pages without internal linking.
- Counting syndicated copies. One story republished across a network is one link's worth of reference.
- Treating nofollow coverage as worthless. Readers, brand demand and discovery still follow.
The off-page hub sets this guide alongside the methods; the off-page SEO expertise page includes a link-scoring tool that applies these criteria to a single link.
