Internal links are promises people and crawlers must be able to follow

An internal link makes three promises

Every internal link makes a discovery promise, a meaning promise, and a continuity promise. Discovery means the destination is exposed through a crawlable anchor and a usable interface. Meaning means the link text and surrounding sentence let a reader anticipate what the destination contains. Continuity means the destination advances the current task rather than forcing an unrelated detour. Google’s link guidance documents anchor elements with destinations as the normal crawlable pattern; W3C’s link-purpose guidance asks that purpose be determinable from text together with context. Those requirements overlap but are not interchangeable. A crawler can reach an anchor labelled ‘click here,’ while a human may not understand it. A descriptive link can still fail if redirects, status codes, or a mismatched canonical send the request elsewhere.

Evidence: Google Search Central; World Wide Web Consortium

Separate crawl access from human understanding

The useful mental model is a path contract: source decision, visible label, final destination, expected answer, and a reason this transition belongs here. Link quantity is not one of the contract’s promises.

Evidence: Google Search Central; Google Search Central

Links connect decisions, not merely documents

A site graph becomes understandable when edges represent reader transitions. A prerequisite edge supplies knowledge required for the current explanation. An evidence edge opens the source or method behind a claim. A continuation edge offers a logical next decision. A navigation edge establishes place within a stable catalogue. Labeling these purposes in an internal map reveals why a link exists and who should maintain it. It also exposes ornamental cross-linking: if an editor cannot name the transition, the edge may be serving an SEO superstition rather than a person. Google encourages descriptive anchor text and sensible organization but does not specify an ideal number of internal links. Dense connections can blur answer ownership, interrupt reading, and make maintenance harder. Sparse connections can strand valuable pages. The path contract lets evidence, not a quota, determine the balance.

Keep the destination’s reader job alongside the source context. When that job changes, the edge must be reviewed even if the URL still returns a successful status.

Evidence: Google Search Central; GOV.UK Service Manual

Crawlability is a rendered property

A link in a design file or source repository is not necessarily available in the rendered page. Inspect the final HTML or browser DOM for an anchor and a meaningful href, then follow redirects to the preferred destination. Check robots directives, response status, and canonical intent separately; a reachable URL may still be excluded or consolidated. Avoid using onclick-only controls for essential navigation when they do not expose a destination. Likewise, do not repeatedly link through a temporary redirect when templates can point directly to the final URL. Google’s documentation supports crawlable anchors, while its starter guide emphasizes that search engines discover pages through links and other signals. These principles help establish eligibility for discovery, not a guarantee of indexing or ranking. A path contract should record the tested rendered state and date because deployment behavior can differ from templates.

Technical reachability is the foundation of the edge, but the contract remains incomplete until a human can understand and use the transition.

Evidence: Google Search Central; World Wide Web Consortium

Meaning must survive scanning and assistive navigation

Readers often encounter links out of the author’s intended sequence: they scan headings, tab through controls, arrive from a search result, or hear a list of links through assistive technology. Repeated labels such as ‘read more’ force them to reconstruct context or guess. Descriptive text should identify destination purpose without stuffing a full title into every sentence. When identical text leads to different destinations, add enough local context to distinguish the choices. W3C allows link purpose to be determined from the link together with programmatically associated context, yet plain, concise labels are usually easier to reuse. GOV.UK’s task-based usability approach offers a practical check: give someone a realistic starting page and observe whether the chosen edge advances the task. Count a technically successful click as a failure when the destination violates the expectation set by the label.

The path contract records the prediction a reader makes before activation and the answer found afterward, connecting accessibility language to task evidence.

Evidence: Google Search Central; World Wide Web Consortium

Example path through an affiliate disclosure topic

Imagine a foundational disclosure article that defines when a material relationship should be revealed. A prerequisite link points to the jurisdiction-specific source used for the article’s scope. An evidence link opens the official guidance supporting placement. A continuation edge labelled ‘audit disclosure placement on a live page’ leads to a checklist, not to a generic marketing hub. A separate link to product-review standards appears only where readers decide how the relationship affects editorial treatment. The internal map records each source decision, label, preferred URL, expected answer, and maintenance owner. It deliberately omits an unrelated email tool page even if that page needs more inbound links. No traffic or ranking result is claimed. The example demonstrates that a smaller set of typed edges can make the intended learning route more explicit than a dense web of topical associations.

If the checklist later becomes a broad compliance guide, its inbound continuation labels must be reassessed because the destination contract has changed.

Evidence: Google Search Central; World Wide Web Consortium; GOV.UK Service Manual

Maintain contracts when content moves

Create path-contract rows for the ten most important reader transitions rather than inventorying every edge immediately. The next action is to validate one row end to end in the rendered site: source context, visible label, anchor destination, redirect chain, canonical target, answer match, and keyboard use. Assign an owner and trigger review after a destination rewrite, consolidation, navigation change, or repeated wrong-turn report. Limits remain: crawl tools may not reproduce every rendered state, click data cannot show whether a choice was understood, and fewer clicks are not always better when prerequisites matter. Search systems decide how to crawl and index pages, so a valid contract cannot promise visibility. Stop when the essential transitions are explainable and testable; expand the map only when another decision path materially affects the audience. The artifact should prevent accidental breakage without turning the site into a frozen graph.