Make your best pages rank for more keywords with Claude

Episode 1104 of Edward Sturm's daily SEO podcast is an eight-minute walkthrough of what he calls the 1-hour SEO update: take a page that already ranks, find the queries it half ranks for, and add the sections it's missing. The idea is sound. A page with clicks has its own authority, so widening what it covers is cheaper than ranking a new page from nothing.

His version has sixteen steps and leans on the Search Console UI, a CSV export, and copy-pasting between tabs. Below are the same sixteen steps with Search Console wired into Claude over MCP instead. The work is identical. The plumbing mostly disappears.

Step 1: go to Google Search Console

Skip the browser tab. Open Claude Code, Cursor, or Claude Desktop with Month17 connected. It's the same GSC data, served from your archive, and the whole workflow happens in this one conversation.

Step 2: see your most visited pages

Sturm clicks into the full report and switches from queries to pages. Here you ask: "Show my top pages by clicks over the last three months." That's one get_top_pages call, sorted by clicks like the report he's describing.

Step 3: pick a page to update

Still your call, in both versions. If nothing jumps out, quick_wins gives the shortlist: it finds queries sitting on page two with real impressions, which is exactly the half-ranking profile this workflow feeds on.

Step 4: export the page's queries as a CSV

This step goes away. "Get every query /guides/example/ ranks for, sorted by impressions" runs get_page_queries and puts the full list straight into context. No export button, no download folder.

Step 5: go to Claude

Also gone. You've been in Claude since step 1.

Step 6: paste in your article and the CSV, then prompt

Half of this survives. You still provide the article, either pasted or, if your content lives in a repo, by pointing Claude Code at the file. The CSV upload is unnecessary because the queries are already in context. Sturm's prompt works verbatim from here: analyze the queries against the existing content, identify keyword clusters with meaningful impressions that aren't well covered, and suggest new H2 sections.

The archive lets you push the same question further back. "Which queries did this page rank for two years ago that it has since lost?" is content_decay, and a lost cluster is often a better section candidate than a new one, because the page already proved it can rank for it. "Are any of these clusters seasonal?" is seasonal_patterns, and it can save you from cutting a section that only looks dead because it's July. A CSV export from the live API sees neither; it's boxed into Google's 16-month window.

Step 7: ask for the first H2, placement only

Unchanged. "Give me the first H2 to add and tell me where to insert it. Don't write the rest of the content yet." His reasoning holds: the body copy comes out better once the placement is settled.

One check worth adding before you commit to a section: cannibalization_check tells you whether another page on your site already competes for that cluster. If one does, the right move is usually an internal link to it, not a duplicate section.

Step 8: have it write the section

Unchanged.

Step 9: rewrite the output

Unchanged, and still the longest step. Sturm's fixes match what we see: the strategic advice needs correcting because you're the expert and the model isn't, the prose runs long, and it hedges. "This can help your SEO" should be "this will help your SEO." Budget real time here.

Step 10: repeat until you've added two or three sections

Unchanged. Loop steps 6 through 9.

Step 11: figure out which pages to boost

Your freshly updated page ranks and gets clicks, so it has authority to pass on through internal links. Sturm picks the targets from memory. Here you can ask: "Which of my pages have declining clicks but still-decent impressions?" That's traffic_drops doing the target-picking, with data instead of a hunch.

Step 12: start a new chat and give it your updated article

Unchanged. A fresh conversation with the final article avoids dragging along the editing history. "This is my page. Just confirm you received it."

Step 13: paste the page you want to boost

Unchanged, prompt and all: "This is the page I want to boost. Find the most natural place in the first article to add an internal link to it. Give me only where to add it, the sentence or H2 section to insert, and the anchor text. Keep it concise and in my voice." His footnote applies too: if the two pages share no topic, add a short H2 to the strong page first, so the link has context around it.

Step 14: reread your introduction

Unchanged. After two or three new sections the intro may be describing an article that no longer exists.

Step 15: update the published date

Unchanged. You made a substantial update, so say so.

Step 16: request indexing

The one step that stays in the Search Console UI, because Google's API exposes URL inspection but not the request-indexing button. inspect_url will confirm the recrawl happened afterward; asking for it is still a click.

Sturm calls the whole thing an hour. Steps 1 through 5 collapse into two prompts, step 6 loses its uploads, and step 11 gets its own tool. What's left of the hour goes where it should: step 9, deciding what the page actually says.