AI NewsGo-to-marketAnnouncement
Semrush added a Claude connector that reads Site Audit data
Semrush added a Model Context Protocol connector that pulls completed Site Audit data into Claude on any plan, so a technical audit's 140-plus issues can be summarised, ranked and assigned by asking the model, with Site Health score, severity and affected pages all reachable in a chat.
Image: Semrush
Why it mattersA search team can take a Site Audit's long issue list and have Claude list affected pages, trace trends between audits, and assign owners, using the real audit data instead of a dashboard.
A technical SEO audit returns hundreds of issues at once, and the step that follows, picking which ones to fix first, still happens in a spreadsheet. Semrush published a workflow that moves the picking into Claude, using a new Semrush connector that pulls audit findings into a chat.
The connector is a Model Context Protocol server, which gives Claude read access to Semrush's Site Audit API on any Claude plan. Data flows one way: from Semrush into the model, so Claude can read completed audits but cannot start new ones or change their settings. The audits themselves come from Semrush's existing Site Audit, which the author Zach Paruch describes as scanning a website for more than 140 technical issues. Once the connector is installed from Claude's connector gallery, Claude can list every completed audit, summarise one by Site Health score and issue counts, pull the issue data for a single page, or trace an issue's history across audits over time.
What the connector costs to run
Each request through the Semrush connector uses between 100 and 10,000 API units. Most requests sit at 100 to 1,000 units; pulling a single past audit uses 10,000 of them. The plans that include connector access, Semrush One Starter, Pro+, and SEO Classic Pro and Guru, come with 50,000 units per month, which reset on renewal. Semrush One Advanced and SEO Classic Business need an added API units package. Only read endpoints on Site Audit are supported.
What the model does with the data
Paruch's prompts cover the usual next steps: list every page affected by one issue, compare how many pages were affected in this audit against the last, propose causes and fixes, rank issues by impact against effort, and assign them to the SEO, content or development team. Across multiple Site Audit campaigns Claude can also surface issues that appear on two domains at once, by name.
The post does not measure how often the model proposes a fix that is wrong, how often it misreads the severity Semrush assigned, or how long the whole flow takes from one end to the other. The worked examples are single-site. Teams running a hundred campaigns will have to decide for themselves whether the monthly budget of API units is enough.
Where the ranking step moves
On a backlog of 140 flagged issues, drafting a prioritised list with named owners is where the real work tends to stop, because a dashboard view supports clicking through issues one at a time, and ranking by business impact against effort is a different kind of task. The connector lets the model do that ranking from the real audit data, in plain language, with the business context the team brings into the chat. The audit itself continues to run inside Semrush, and the connector takes over at that point, drafting the owner-by-issue list that a team would otherwise build by hand.
Source
Semrush: How to do an SEO audit with Claude & the Semrush MCP.
This item was written by an AI system from the linked source. Reveneau is responsible for what it publishes.
Get AI News in your inbox
New developer tools, model and agent releases, and how teams are actually using them to release software. Short, and only when there is something worth reading.
