Internal Links Are the Agent Task WordPress Keeps Missing

Google’s link guidance turns a familiar WordPress chore into an agent test. Internal linking only works when the writer knows the site before it writes.

That matters because publishing agents are usually judged on draft quality: outline, citations, tone, metadata. The harder task is co nective. An agent has to understand inventory, choose crawlable markup, write anchors that set expectations, and know when a link is editorially relevant. Otherwise, it creates isolated pages at scale.

Google’s AI features guide says normal SEO practices still apply to AI Overviews and AI Mode, which is the same operational point behind this blog’s earlier workflow warning. AI search does not make internal links optional. It makes bad site memory more visible.

The Link Is the Assignment

Internal linking is often treated as cleanup after publication. In an agent workflow, that is backward. The link decision belongs upstream because it changes what the draft should say, what supporting pages it should acknowledge, and which destination deserves a reader’s next click.

Google’s link guidance says links help Google find pages and understand relevance, but it also pushes the work toward ordinary editorial craft. The linked page must be reachable, the anchor must make sense, and the sentence around it must provide context.

That is why a link agent ca not merely scan for keyword overlap. It needs to know the current article’s job, the destination page’s job, and the reason the reader would move between them.

Inventory Comes Before Anchor Text

On WordPress, the inventory is there if the agent is allowed to read it. The WordPress REST API posts reference exposes fields and filters around links, dates, slugs, status, titles, content, categories, and tags. That is not prose data. It is the map.

A useful agent should know which pages are live, which posts were recently modified, which URLs already rank or convert, which categories define topical clusters, and which old pieces still answer the same intent. Without that, anchor text becomes decoration.

The sitemap matters too, but it is not the same as editorial linking. Google’s sitemap guidance frames sitemaps as a way to tell search engines about important pages and their relationships. A contextual internal link tells a reader why one page belongs after another.

Crawlable Is Not Cosmetic

The technical floor is plain. Google generally recommends crawlable links as anchor elements with href attributes. A button, script event, or visual card may feel like a link to a user and still be a weak discovery path if the rendered HTML does not expose the destination clearly.

That distinction matters in WordPress because links can come from blocks, menus, related-post plugins, shortcodes, reusable patterns, and theme templates. An agent that only edits paragraph text may miss half the link graph. An agent that only changes UI modules may miss the strongest contextual opportunities.

The agent should inspect rendered output, not just stored post content. It should recognize nofollow, redirects, canonical conflicts, draft targets, broken URLs, and links buried in components that search engines or users may not treat as meaningful.

Relevance Is the Human Test

Yoast’s internal linking guide gets the practical sequence right: link newly published content from similar-topic pages, not random pages. That sounds simple until a site has years of archives, overlapping service pages, multiple authors, seasonal posts, and product-led pages fighting for the same intent.

This is where agents either become useful or noisy. A bad agent suggests every page that shares a keyword. A better one distinguishes a definition from a comparison, a feature page from a support page, a current guide from a stale a nouncement, and a revenue page from an informational bridge.

Anchor choice is part of that judgment. The anchor should be specific enough to stand alone but not stuffed with every related phrase. The surrounding sentence should make the relationship visible: background, next step, deeper guide, evidence, pricing, migration path, or objection handler.

For SEO editors, the review question changes from “did we add links?” to “did we add the right next step?” That is a smaller checklist and a harder editorial standard.

The Workflow Has to Loop Back

Internal linking also has a lifecycle. A new article needs incoming links after publication. An old article may need links to newer answers. A refreshed article may need anchors rewritten because the old destination changed. A deleted article may leave behind links that now leak users into redirects or 404s.

Google’s SEO Starter Guide says Google primarily finds pages through links from pages it has already crawled. That is the reason the task ca not end when the agent imports a post. Publishing creates a new URL, but linking makes it part of the site.

The Search Console Links report can support that loop by showing top internally linked pages and which pages link to a specific URL. It is not a strategy by itself, but it gives editors a reality check against the agent’s proposed map.