Define scope
Confirm included sections, exclusions, crawl limits, priority destinations, rendering constraints, and optional data.
Methodology
A crawl is the beginning, not the deliverable.
Six clear stages turn website evidence into internal-link recommendations your team can review and implement.
How it works
Six stages separate evidence from implementation.
The process keeps uncertainty visible instead of turning every similarity score into a confident instruction.
Capture evidence
Record public URLs, canonicals, page types, headings, body context, internal edges, anchors, and crawl limitations.
Generate candidates
Order possible relationships using topical, structural, intent, graph, and business-priority signals.
Suppress noise
Remove already-linked, duplicate, reciprocal, irrelevant, or structurally inappropriate candidates where evidence permits.
Review and decide
Inspect exact source context, target purpose, final anchor, placement, repetition, and implementation cost.
Export and verify
Deliver the reviewed plan, assign ownership, publish under your control, and recheck the live pages.
Evidence hierarchy
A shared keyword is not enough.
The strongest route combines visible source context, an accurate destination, reader usefulness, and an implementation point that does not distort the page.
LEVEL 01Observed page evidence
Visible source paragraph, target purpose, existing links, anchors, page type, canonical, and crawl path.
LEVEL 02Relationship evidence
Topic proximity, intent progression, content role, priority-page need, and current graph position.
LEVEL 03Optional supporting data
Agreed exported GSC query/page data may inform review, but it does not replace source-context inspection.
LEVEL 04Business constraints
Priority destinations, team capacity, publishing workflow, risk, and what the client does not want changed.
Decision matrix
The method makes uncertainty visible.
A stable state and a written reason are more useful than pretending the model has certainty.
| State | Meaning | Required evidence | Next action |
|---|---|---|---|
| Approved | Route works as reviewed | Useful context, accurate target, natural anchor, viable placement | Implement and verify |
| Approved after edit | Relationship works after a change | Edited anchor, placement, destination, priority, or note recorded | Implement reviewed version |
| Rejected | Candidate should not be used | Specific rejection reason tied to visible evidence | Do not implement |
| Pending review | Evidence is incomplete or decision is deferred | Missing information or unresolved judgement named | Investigate or leave out of final pack |
Review questions
What makes an internal-link recommendation useful?
Does the source context naturally support the destination?
A topic match outside the visible paragraph is insufficient.
Is the target the best next page for the reader?
A more commercial page is not automatically the correct destination.
Can the anchor describe the target without sounding forced?
The final language must fit the sentence and represent the destination accurately.
Would the route duplicate an existing useful path?
More links are not inherently better.
Can someone implement the instruction without guessing?
If not, the row is not ready for handoff.
Limitations
Findings depend on what the agreed audit can see.
Blocked resources, JavaScript rendering, incorrect canonicals, generated URLs, authentication, incomplete exports, and site changes during the audit can alter the evidence.
Known limitations belong in the handoff. The method does not claim full technical coverage, indexing control, or guaranteed search outcomes.
Common questions
Answers before you start.
Does a high candidate score guarantee approval?
No. A score can order review work. It cannot verify that the exact source context supports a useful, natural route.
Are optional GSC exports required?
No. They may support prioritisation when agreed, but the standard method does not require account access and does not replace contextual review.
Does the service publish links?
No. The standard deliverable is an implementation pack. Your team retains CMS control and publication responsibility.
How are rejected candidates handled?
They remain visible with a rejection reason when that context helps explain the decision.
See the output
Explore a complete sample internal-link plan.
The Sample Pack shows how approved, edited, and rejected recommendations appear in the final handoff.