Semrush gave Claude eight SEO tasks on a live site and wrote up what happened

Semrush ran eight typical SEO tasks through Claude on a live site and wrote down what came out. The idea is simple: instead of a long prompt every time, you describe the procedure once, connect the agent to Search Console, analytics and Semrush, and then give it a short task. The write-up is on the Semrush Blog.

The four layers under the agent

The first layer is a project: a document with business context, competitors and priorities. The second is skills, procedures you write once and keep improving. The third is MCP connections, live access to Semrush, Google Search Console, Google Analytics, SerpAPI and Firecrawl. The fourth is the prompt itself, which can be one line because the system already knows the rest. A full set of connections needs a paid Claude plan; the free one handles a single custom connector.

Anyone running an agency or an in-house team knows the pain: the analysis done one way on Monday gets repeated differently on Thursday because someone skipped a step. A saved skill removes exactly that. If you are still choosing a stack, see our overview of AI SEO tools for small business.

What the agent found on the live site

The examples in the piece are down to earth, which is what makes them interesting.

• Content decline diagnosis flagged three pages with traffic drops: one was an intentional consolidation, another simply lost its sitemap entry. Nothing needed rewriting.
• The technical sprint grouped issues by root cause: ten "temporary redirect" errors traced back to internal links without a trailing slash.
• The internal linking workflow found that the site's second-largest revenue driver had zero internal links pointing to it, and suggested exact sentences for placing them.
• The release regression check caught two critical problems: temporary redirects across domains and a page removal that never happened during a migration.

Look at the character of these findings. They are not a quarterly strategy but blind spots: things a person misses after looking at the same report for the tenth time. An agent fits that niche well because it does not get bored.

Where the limits are

Semrush itself says that the content portfolio workflow, which sorts every URL into keep, update, consolidate, redirect or investigate, needs properly configured conversion tracking. Without it the agent recommends selectively: it sees traffic but not which pages make money. Citation mapping has the same catch: 32 prompts is a sample, and you still need to check whether it resembles what your customers actually ask.

A few things the original does not say outright. All the examples come from one site, and this is a post by a company that sells a tool built into the setup. There is no comparison against the human effort for the same tasks, so "time saved" stays an impression rather than a number. There is also a practical point for smaller markets: MCP connectors to Search Console take about fifteen minutes to set up, but access to client accounts at an agency usually takes weeks to agree, and that bottleneck does not go away. Teams already reporting in GA4 will feel the difference faster; the basics are covered in our guide to GA4 reports for SEO.

The example that stuck with me is the page with no internal links that brings in the second-largest revenue. I wonder how many of those sit on sites where internal linking was last touched two years ago.

Related reading: SGE and AI Overview: Optimising for AI Search · SEO Website Promotion