Short answer: measure internal linking by graph properties and user journey, not by a "topical authority score". Google recommends crawlable links and contextual anchor text and says that links help page discovery. For professional services, useful metrics are orphan rate, contextual-link coverage, wrong-target rate, journey completion, collision rate and maintenance cost. Ranking, traffic and AI citations must be analyzed separately because they can have many confounders.

Define the population

Choose the cluster or section audited: for example all pages for a service, methodology, case studies and related articles.

Excludes expired campaigns, intentionally noindex pages, and URLs that are not part of the journey.

The baseline

Save:

  • list of URLs;
  • canonical intent;
  • owner;
  • incoming contextual links;
  • outgoing contextual links;
  • click depth;
  • status/indexability;
  • language/region;
  • collisions;
  • Search performance;
  • relevant conversion paths.

Version the snapshot.

Metric 1: orphan rate

Numerator: eligible pages without incoming internal crawlable links. Denominator: eligible pages.

Before fixing, check if the page is worth keeping. An orphan can be a symptom of a page with no role.

Metric 2: near-orphan rate

Define the internal threshold, for example pages with very few contextual links and no route from the hub. Keep the threshold the same between rounds.

There is no universal Google threshold.

For important pages, check if incoming links come from contexts that continue the same issue or decision.

Footer links and related-content blocks can be reported separately.

Metric 4: wrong-target rate

On a sample of links, the evaluator classifies useful',neutral', wrong target',collision risk', `wrong language'.

This is more informative than raw link count.

Metric 5: anchor clarity

Define a heading for descriptive, generic, and stuffing anchors. Does not require exact-match; requires sufficient context for the reader.

Metric 6: journey completion

Choose representative routes:

  • problem → diagnosis;
  • diagnosis → methodology;
  • methodology → evidence;
  • evidence → service, when relevant.

Measure dead ends and click paths without forcing every journey to end commercially.

Metric 7: collision rate

Number of pages claiming the same main intent. Internal linking does not compensate for a cluster with ambiguous ownership.

After consolidations, this rate should decrease.

Metric 8: concentration risk

A recommendation engine can send disproportionately many links to a hub. It measures distribution and triggers reviews when a page becomes a target for almost any topic.

It does not artificially equalize the graph.

A target may return 200, but be editorially outdated. It uses review and owner metadata to detect these cases.

Metric 10: maintenance cost

It measures how many links need to be revised after consolidations, rebrands or service architecture changes. A very dense graph can generate debt.

Search results

Track indexing, queries and landing pages. Use event log for content updates, backlinks, PR and algorithm.

An increase after linking changes is correlation until the comparative design supports more.

AI observations

If you monitor source citations, keep them separate. A more coherent internal graph may produce no detectable change in AI systems.

This is no reason to falsify an authority score.

The denominators

Orphan rate uses eligible pages. Wrong-target rate uses rated links. Journey completion uses eligible sessions/journeys. Do not mix the denominators.

Observation window

Graph metrics can be checked after deploy. User behavior and Search outcomes have different windows. Keep each series with its period.

False attribution risks

  • simultaneously rewritten content;
  • updated service pages;
  • backlinks;
  • campaigns;
  • global navigation changes;
  • seasonality;
  • Search algorithm.

Action thresholds

P0: wrong target to wrong pages/noindex/other language for critical paths. P1: important orphan pages. P2: anchors/weak context. P3: marginal opportunities.

Acceptance criteria

The measurement system is valid when the population, snapshot, headings, denominators and event log are versioned and each finding can be mapped to concrete URLs.

How to measure your path to conversion without over-attributing

A user can read a diagnostic article, then methodology, case study and service page. Internal linking can facilitate the path, but it is not the only cause of conversion. Keep page sequence and time to contact as observations, not ranking proof.

It also compares sessions that go directly to service pages. If the editorial route produces better qualified leads, this is a useful commercial result even if the total traffic remains stable.

Measuring conflicts

In addition to clicks, track if several pages are competing for the same next step. A graph can have many links and still unclear ownership.

After the consolidations, check if the old links point to the new owner and if there are no remaining routes to retired pages.

Stop condition

When priority pages have clear contextual paths, broken and wrong-target links are below the threshold, and new recommendations become predominantly redundant, put the system into monitoring. Don't aim for link density as a goal in itself.

Threshold for action

Not every click path variation requires modification. Prioritize links that point to deprecated pages, the wrong language, or an incorrect intent owner. Minor differences in CTR can remain under observation until there is a repeatable pattern.

Claim ledger

  • FACT/EVIDENCE: Google recommends crawlable links and contextual anchor text.
  • FACT/EVIDENCE: links help discover pages.
  • PRACTITIONER GUIDANCE: graph metrics and journey metrics must be separated from Search outcomes.
  • NOT PROVEN: a universal topical authority score derived from internal linking.

Conclusion

The impact of internal linking is most clearly seen in the graph and user paths. If important pages become findable, bad links decrease and journeys are more coherent, you have a real result. Ranking and AI citations can complement the analysis, but should not retroactively define success.

Sources reviewed

Related insights