We built a GSC MCP server for automated ecommerce SEO audits — here’s what it found

Any ecommerce SEO consultant will tell you the same thing: Google Search Console holds the most honest data you will ever get about a site's organic performance. It does not lie, guess, or average across channels. What it does do is sit behind a dashboard that makes large-scale pattern recognition genuinely painful.

Exporting 16 months of search analytics, splitting it by device, country, query and page, cross-referencing impressions against click-through rate, building a striking-distance keyword list, then mapping every ranked URL to its primary query for a redirect plan — that is a full day of work before a single recommendation is written. We built a faster path. This post documents the full loop end to end, with real output from a live audit.

01Why manual GSC analysis breaks down

The data is excellent; the access pattern is the bottleneck. The default export grabs a 28-day window, one dimension at a time, into a CSV that then has to be reshaped by hand. The signal that actually drives strategy — which pages hold ranking equity, which queries sit one position from page one, which impressions are pure noise — only emerges once you cross several exports against each other.

02What an MCP server changes

MCP stands for Model Context Protocol — an open standard that lets an AI assistant like Claude connect directly to external tools and APIs, rather than waiting for a human to paste in data. Instead of exporting a CSV, uploading it somewhere and prompting an AI to “analyse this”, an MCP server lets the model query the API directly, in real time, with full parameter control. For ecommerce SEO work that matters because:

• No loss in translation. You query the GSC Search Analytics API directly — same data, filters and date windows as the dashboard.
• The full 16-month window. GSC stores 16 months of history; most manual exports only reach back 28 days. The MCP query uses the full range.
• Parallel datasets. Queries by clicks, queries by impressions, pages by clicks, page×query maps and country splits are fetched simultaneously, not one after another.
• Structured output. The model receives JSON, not a screenshot — so it can sort, filter and group without inventing numbers.

The connector we use is plugin_seo_gsc, which wraps the Search Analytics API and exposes it as a callable tool inside Claude's session.

03The full audit loop, in four stages

Every ecommerce SEO audit we run now follows the same loop — from a direct API pull to a brief the design and content teams can act on.

04Fetching 16 months of search data

The first call is always list_sites — confirming the correct property identifier before any data pull. For a URL-prefix property like https://seventy2digital.com/ that matters, because the API distinguishes domain properties (sc-domain:example.com) from URL-prefix properties, and mixing them up returns empty results. Once the property is confirmed, four parallel queries fire:

Date window: 2025-02-14 → 2026-06-12 (16 months, respecting GSC's 2-day data lag). On a small-to-mid ecommerce site, Q1 typically returns 1,000–5,000 unique queries. On the site audited here it returned 1,341 unique queries across 100 indexed URLs — a manageable but revealing dataset.

05Six structured datasets, one export file

The raw API responses are processed with jq and written to a single Markdown file, gsc-export.md, with six named sections:

• Dataset 1 — Top 100 queries by clicks. The true performance ranking. For most ecommerce sites the top 100 clicked queries account for 90%+ of organic traffic — what the site actually earns, not just shows up for.
• Dataset 2 — Top 100 queries by impressions. The opportunity inventory. High impressions with low clicks means either deep rankings (page 4+) or a CTR problem solvable with better titles and meta descriptions.
• Dataset 3 — Top 60 pages by clicks. The URL equity map. Any page that earned a click in 16 months must survive a migration with a 301 if its URL changes.
• Dataset 4 — Page × query map. The most strategically important set — it answers “what does Google think each page is about?” The primary query per page becomes the anchor keyword for any rewrite or new slug.
• Dataset 5 — Germany-only top 60 by impressions. Germany is the primary market; this surfaces the German-language and local terms that drive revenue-relevant traffic.
• Dataset 6 — Non-Germany top 60 (USA + GBR + IND). The English-speaking audience. Since the API has no negation filter, we aggregate the three largest non-German markets. Here the USA contributes 45,238 impressions — roughly triple Germany — at a CTR of 0.02% against Germany's 0.55%.

06What the data revealed

With the structured export in hand, Claude runs the analysis — no manual sorting, no pivot tables. Five findings from this audit, each shaping the strategy directly. Start with the headline numbers across 16 months:

Finding 1 — 171,000+ impressions in 16 months: visibility is there, the opportunity is in the click

The audit’s first headline finding is a strong foundation: the site has built genuine search visibility across 1,341 unique queries and 100 indexed URLs — Google is finding, indexing and surfacing the content at scale. That makes the next phase tractable rather than starting from scratch. The opportunity the data reveals is positional: most high-impression pages rank at positions 35–95, deep enough that even excellent CTR wouldn’t yet produce meaningful traffic. This is a ranking-consolidation story — the assets exist, the visibility is proven, and the work is moving ranked pages up the page, not rebuilding from zero.

Finding 2 — Precision diagnosis: 30,000 competitor-brand impressions identified and isolated

This shows why impression volume alone is never enough — and why the workflow earns its keep. Dataset 2 surfaces a striking anomaly: the query online visibility services by garage2global drives 29,982 impressions at position 6.1, and the automated analysis identifies it immediately as a competitor’s branded term. The /category/resources/online-visibility/ page had been accidentally ranking for it — high apparent visibility with structurally zero click potential. Without the 16-month impression-to-query breakdown, this pattern stays invisible in any standard dashboard. The fix is clean and high-value: one structural change recovers that impression inventory and redirects it toward owned keyword themes — in seconds, not days.

Finding 3 — Already at position 3 for a commercial keyword: one diagnostic fix from earning traffic

The ecommerce SEO services page (/ecommerce-seo-services-reutlingen/) carries a page-level average position of 55.3 — but that aggregate hides the query-level picture:

These are not targets to work toward — they are rankings the site already holds. The audit identified them; the standard dashboard would not. Ranking #3 for a commercial keyword without clicks almost always points to a narrow, fixable cause: a missing or under-optimised meta title/description, a SERP feature pushing the result below the fold, or a rendering issue — any one a single-session fix. The value here is surfacing a high-value, already-earned ranking and pointing precisely at where to look next.

Finding 4 — Germany confirmed as the high-intent market: 0.55% CTR vs 0.07% internationally

Germany produces 56% of all clicks from just 13% of impressions — an 8× efficiency advantage over the international audience. The data confirms Germany as the primary high-intent market and gives the content strategy a concrete mandate: protect and extend the DE rankings. The English-speaking audience generates substantial impression volume but through a different content mix (DCX frameworks, CX strategy) that serves awareness rather than conversion. That’s not a failure — it’s a segmentation insight — the EN strategy needs its own brief targeting commercial ecommerce SEO terms.

Finding 5 — Seven queries at positions 5–20: ranked, proven, ready to optimise

Google has already validated the page relevance and the content is indexed — no new content needed, targeted on-page optimisation moves them to page one. A prioritised action list:

07Handover to design — from data to brief

The analysis feeds directly into a design and development brief. Concretely:

• Redirect map. Every URL that earned a click in 16 months gets a named entry in the 301 map. The export is the source of truth: if a URL is in Dataset 3, it survives the migration or is explicitly redirected.
• New slug recommendations. Dataset 4 drives slug naming. A page found primarily via one query keeps those exact words in its new slug; the phantom-query page is restructured or removed.
• Content-brief inputs. Dataset 5 defines the German keywords that anchor the /de/ pages; Dataset 6 defines the /en/ targets. Position data sets realistic expectations — queries at 5–20 need optimisation, queries at 50+ need new content.
• Design constraints. The CTR analysis flags meta-description problems versus ranking-position problems — different swimlanes. Meta fixes are a design and CMS task; position fixes are content and link-building.

The design team receives gsc-export.md plus the structured analysis report. Everything needed to build the information architecture, write the content briefs and configure the redirect rules lives in those two documents.

08Why this workflow matters

The 2025–2026 ecommerce search landscape rewards specificity. Generic “SEO audit” processes that treat every site the same miss the patterns that actually move performance. This loop matters because:

1. It uses your data, not industry benchmarks. The striking-distance keywords, the phantom queries eating impression budget, the redirect-critical URLs — all specific to your domain's 16-month history. No template produces them.
2. It closes the loop between audit and execution. Most reports go unread because the gap between “what we found” and “what we are building” is too wide. This outputs a redirect map, slug recommendations and country keyword lists in the same session as the analysis.
3. It separates German and English intent. For businesses in DACH plus international markets this is not optional — the two audiences find you through completely different queries.
4. It makes migrations survivable. Migrating without a GSC-derived redirect map is how you lose six months of ranking equity overnight. The page × query dataset shows exactly which URLs hold equity and what they rank for.