WordPress SEO automation, honestly

WordPress runs a large share of the web partly because it is endlessly automatable. That is also how sites get into trouble: the plugin ecosystem makes it trivial to automate exactly the decisions that should stay manual. This is the split that holds up in practice.

Worth automating

Publishing mechanics

Everything between "the draft is approved" and "the post is live" is mechanical and should never be done by hand: creating the post, setting the slug from the title, assigning category and tags, uploading and attaching the featured image, filling the SEO title and meta description fields, scheduling. The REST API does all of it. A publishing step that takes fifteen minutes of clicking is fifteen minutes taken from the part that decides whether the post ranks.

Structured data

Schema should be generated from the post's own fields, not typed. Article type, headline, publish and modified dates, author, image, publisher — all of it is derivable. Hand-maintained schema drifts out of sync with the page within a month and then actively misinforms.

Internal linking suggestions

Automatically surfacing candidate internal links is genuinely useful; automatically inserting them is not. The difference matters: a suggestion engine that says "these four older posts are about this subject" saves real time, while an auto-linker that replaces every occurrence of a keyword with a link produces pages with eleven links to the same URL and anchor text that reads like spam.

Monitoring and audits

• Broken internal links and 404s from removed content.
• Posts with no internal links pointing at them.
• Missing or duplicated meta titles and descriptions.
• Images without alt text, or oversized images.
• Pages that dropped out of the index, and pages that entered it that should not have.
• Core Web Vitals regressions after a theme or plugin update.

These checks are boring, high-frequency and objective — the definition of a job for a machine. Run them weekly and act on the diffs rather than staring at a dashboard.

Content refresh detection

A scheduled job that lists posts whose average position or clicks fell materially over the last quarter is more valuable than most content strategy exercises. Updating a post that used to rank is consistently cheaper and faster than writing a new one.

Not worth automating

What to publish

Topic selection is where the ranking outcome is mostly decided, and it depends on things a scheduler does not know: what you can credibly claim, what your customers actually ask, where the current results are weak. A pipeline that publishes three posts a week because Thursday arrived will fill the site with pages targeting queries you cannot win. Picking winnable queries is the part that has to stay human.

Publishing without review

The failure mode is not bad grammar; it is confident, specific, wrong facts — a price, a statistic, a legal claim, an integration that does not exist. One human read-through before publish catches nearly all of it and costs minutes — it is the same edit pass described in how to make AI content actually rank. Sites that skip it eventually publish something they have to take down, and by then it has been crawled.

Redirects and deletions

Automated bulk redirect rules are a fast way to build a redirect chain, a loop, or a rule that quietly sends a valuable page to the homepage. Redirects are cheap to add and expensive to unpick — decide them one at a time.

Anything touching money or law

Pricing pages, refund and shipping terms, medical or financial claims, anything with a regulatory surface. Automate the draft if you like; never automate the publish.

A workflow that holds up

1. Monthly: a human picks the queries, based on what the site can credibly answer and where the current results are weak.
2. Weekly: drafts generated against those queries, with the brief carrying the angle, the audience and the internal links to include.
3. Per post: one human editing pass — cut the padding, verify every specific claim, check the structure reads on its own.
4. On approve: automated publish — post created, slug, category, featured image, meta fields, schema, scheduled.
5. Day after publish: automated check that the URL is indexable, the canonical is self-referencing and the schema validates.
6. Weekly: automated audit diff — broken links, orphans, missing metadata, index changes.
7. Quarterly: automated decay report, and a human decides what to update.

The shape to notice: humans decide, machines execute and machines watch. Every failure I have seen in an automated content operation comes from inverting one of those.

Plugin hygiene, since it is the other half of the problem

• One SEO plugin. Two will both emit canonicals and schema, and the duplicates conflict.
• Check what the theme emits before adding a plugin to emit it again.
• Every plugin that adds front-end scripts is a performance cost on every page, including pages it does not touch.
• After any update, spot-check a post, a page and an archive for changed canonicals or missing schema. Updates change output more often than release notes admit.

Frequently asked questions