5 WordPress Tasks You Can Automate With AI (Using MCP)

Every WordPress site owner has a list of tasks they know they should do but keep putting off. Not because they're hard — because they're tedious. Adding internal links between 50 related articles. Writing meta descriptions for posts that don't have them. Cleaning up unused tags and categories. Setting up redirects after pruning content.

These are the tasks that separate well-optimized sites from mediocre ones. And they're exactly the kind of tasks that AI is perfect for — repetitive, pattern-based, requiring consistency across a large number of posts.

Using Connect My Site to AI (an MCP server for WordPress), I've automated five recurring tasks that used to eat hours of my week. Here's what they are, how they work, and what the results look like.

If you're new to MCP, start with my explainer on what MCP servers are and the full setup guide.

Task 1: Internal Linking

Manual time: 2-4 hours per category (reading articles, finding connections, editing each post, choosing anchor text, inserting links)

With AI: 15-20 minutes per category

Internal linking is the highest-leverage SEO activity that most site owners neglect. Every article on your site should link to 3-5 other relevant articles. These links help Google understand your site structure, pass authority between pages, and keep visitors browsing longer.

The problem is that it's mind-numbingly tedious. For every article, you need to read the content, think about which other articles are relevant, find the right paragraph to insert a link, choose appropriate anchor text, open the other article in the editor, and add the link. Do that across 100 articles and you've lost a full day.

Here's how I do it with AI now. I connect Claude to the site and say: "Map the internal links for all turtle diet articles. Find which articles should link to each other but don't. Suggest specific anchor text and paragraphs for each new link."

Claude reads all the articles in the category, maps the existing internal links, identifies gaps, and presents a list of suggested links. Each suggestion includes the source article, the target article, the proposed anchor text, and the paragraph where the link should go. After I review and approve, Claude applies the links with the Pro write tools.

On The Turtle Hub, this workflow added 47 internal links across 24 articles in one session. It would have taken me an entire afternoon to do manually, and I probably would have missed half the opportunities because I can't hold 24 articles in my head simultaneously. Claude can.

Task 2: Meta Description Generation

Manual time: 1-2 minutes per post (reading the article, writing a compelling 150-character summary)

With AI: About 3 minutes for 20 posts

Meta descriptions are the snippets that appear below your page title in Google search results. They don't directly affect rankings, but they significantly impact click-through rate. A compelling meta description can be the difference between someone clicking your result or the one below it.

The problem is that WordPress doesn't require meta descriptions, and most SEO plugins (Yoast, RankMath) leave the field blank by default. On my sites, I regularly find 10-20% of articles with empty or auto-generated meta descriptions. Auto-generated ones are usually terrible — they just pull the first 160 characters of your content, which is often an intro paragraph that doesn't describe the article at all.

With AI connected via MCP, I ask Claude to find all posts with missing or auto-generated meta descriptions. It uses the find_missing_seo_elements tool to identify the gaps, then reads each article and writes a tailored meta description. Each description is:

• Under 160 characters (Google's display limit)
• Specific to the article's actual content, not generic fluff
• Written with the target keyword naturally included
• Compelling enough to drive clicks ("Learn how..." or "The exact process for...")

Claude applies them in bulk with the bulk_update_meta tool. What used to take 30-40 minutes across 20 posts takes about 3 minutes now.

Task 3: Content Pruning Preparation

Manual time: 4-6 hours for a full site audit

With AI: 30-45 minutes including review and decisions

I've written extensively about content pruning — the process of deleting, merging, or rewriting underperforming articles to improve overall site quality. It's the single most impactful SEO strategy I've used across my portfolio of 12 sites.

The preparation phase is where AI saves the most time. Before pruning, I need to identify candidates — thin articles, stale content, duplicate topics, and orphan pages. Then for each candidate, I need to decide: delete, merge, or rewrite?

With AI connected to the site, the entire diagnostic phase becomes a conversation:

• "Find all posts under 800 words" — Claude returns the list with word counts and edit links
• "Find all posts not updated in the last 6 months" — stale content identified
• "Find posts with duplicate titles or keyword cannibalization" — cannibalization pairs flagged
• "For each thin post, check if there's a stronger article on the same topic I can merge it into" — merge candidates identified with specific recommendations

The AI can read both the thin article and the potential merge target, compare their content, and tell me exactly what unique information from the thin article should be preserved in the merge. That comparison step alone saves an hour or more, because manually reading and comparing 10+ pairs of articles is exhausting.

I still make the final delete/merge/rewrite decisions myself — that requires editorial judgment and traffic data that the AI doesn't have. But by the time I'm making those decisions, the AI has already done 80% of the work. I used ScanMyPosts alongside MCP for this — ScanMyPosts catches the content-level issues (broken shortcodes, dead images), while MCP handles the site-level analysis (orphans, cannibalization, content structure).

Task 4: Redirect Management

Manual time: 2-3 minutes per redirect (decide target URL, navigate to redirect plugin, add the redirect, test it)

With AI: About 5 minutes for 10-15 redirects

Every time you delete or merge a post, the old URL needs to redirect somewhere relevant. If you just delete a post without redirecting, visitors get a 404 error and Google removes it from the index — but any backlinks pointing to that URL lose their value.

Redirect management is one of those tasks that's simple in concept but tedious in practice. For each deleted or merged post, you need to identify the most relevant remaining page to redirect to, create the redirect, and verify it works. On a major pruning session where I'm removing 20 articles, that's 20 individual redirects to set up.

With the MCP write tools, Claude handles this as part of the pruning workflow. When I decide to merge Article A into Article B, Claude creates the redirect from A's URL to B's URL using the manage_redirects tool. When I decide to delete an article with no merge target, Claude identifies the most topically relevant remaining page and sets up the redirect there.

The best part is consistency. When I set up redirects manually, I'd sometimes forget to redirect one or two URLs in a batch. The AI doesn't forget — it tracks every deletion and ensures every URL has a corresponding redirect. I can also ask it to audit the existing redirects: "Are there any redirect chains (A → B → C instead of A → C)? Are there any redirects pointing to pages that have since been deleted?" These are problems that accumulate silently and are nearly impossible to catch manually.

Task 5: Taxonomy Cleanup

Manual time: 1-2 hours (reviewing all categories and tags, identifying unused or redundant ones, reassigning posts, deleting empty terms)

With AI: 15-20 minutes

WordPress taxonomies — categories and tags — get messy over time. You create a tag in the heat of writing an article, then never use it again. You have three tags that mean essentially the same thing ("turtle-food", "turtle-diet", "what-turtles-eat"). You have categories with one post in them. Every unused or redundant taxonomy term is a dead page on your site that Google crawls and finds empty.

The cleanup process with AI looks like this:

• "List all categories and tags with their post counts" — Claude uses list_categories_tags to get the full taxonomy inventory
• "Which tags have fewer than 3 posts?" — identifies candidates for deletion or merging
• "Which tags overlap in meaning?" — Claude reads the post titles under each tag and identifies semantic duplicates
• "Merge 'turtle-food' and 'turtle-diet' into 'turtle-diet', then delete 'turtle-food'" — Claude uses reassign_taxonomy to move posts and delete_category_tag to remove the empty term

On Acuario Pets, this workflow eliminated 23 unused tags and merged 8 redundant ones in a single session. The site went from 67 tags to 36 — each one now has at least 3 posts, and there are no semantic duplicates. That's 31 fewer empty taxonomy pages for Google to waste time crawling.

The Compound Effect

Individually, each of these tasks produces modest improvements. Meta descriptions improve CTR by a few percent. Internal links boost time on site. Pruning concentrates authority. Redirects preserve backlink value. Taxonomy cleanup reduces crawl waste.

But they compound. A site with strong internal links, optimized meta descriptions, pruned thin content, proper redirects, and clean taxonomies is categorically different from a site where these tasks are neglected. Google evaluates your site holistically — it's the total quality signal that determines your rankings, not any single factor.

The reason most site owners neglect these tasks isn't that they don't know they're important — it's that they're too tedious to do consistently. That's exactly why automating them with AI changes the game. When a task that used to take 4 hours takes 30 minutes, you actually do it. Monthly instead of never.

Getting Started

To automate these tasks on your own WordPress site:

• Install Connect My Site to AI — the free version handles all the read-only auditing (Tasks 3 and 5's diagnostic phase). The Pro version ($49/year) adds the write tools needed for Tasks 1, 2, 4, and the action phase of Tasks 3 and 5.
• Connect Claude Code or Claude Desktop to your site using the MCP setup guide.
• Start with Task 2 (meta descriptions) — it's the quickest win and the lowest risk. Then work up to internal linking and pruning as you get comfortable with the workflow.

For a real-world example of what an AI-powered audit looks like, check out what happened when I let AI audit The Turtle Hub — 23 issues found and fixed in about an hour.