Gsc Analyst

GSC Analyst

A Claude Code skill that turns a Google Search Console export into a prioritized SEO action plan.

Dropping Search Console CSVs into an AI agent already works. The problem is that the analysis it produces is fluent and often wrong, because Search Console data invites specific mistakes: average position read as a rank, CTR anomalies that are sampling noise, traffic "declines" that are last November, and AI Overview impact asserted as fact when Search Console does not report it.

This skill encodes what an experienced SEO checks before drawing a conclusion.

Install

git clone https://github.com/aurelien-dio/gsc-analyst.git
cd gsc-analyst
./install.sh

Then open Claude Code and say: *"analyze this Search Console export"* with the ZIP from Search Console's Export button.

Requires Python 3.9+ and Claude Code. Dependencies install into an isolated venv at `~/.claude/skills/gsc-analyst/.venv/`; your system Python is untouched, and `./uninstall.sh` removes everything.

What it analyzes

Analysis	What it answers
Brand / non-brand split	Is SEO growing, or is it harvesting existing demand?
Year-over-year trend	Is traffic actually falling, or is this seasonal?
SERP absorption cohort	Which queries lost clicks with ranking intact?
CTR anomalies	Which titles underperform their position, significantly?
Striking distance	Where does a rank gain convert into real clicks?
Cannibalization	Which queries have several of your URLs competing?
Concentration	How much traffic rests on how few pages?
Device and country	Where does the traffic mix explain poor aggregates?

The parts that are not obvious

**Average position is never treated as a rank.** It is an impression-weighted mean, a query alternating between 3 and 15 reports 9, a position it never held. Every position-derived claim is scoped accordingly.

**The expected CTR curve is fitted from your own data**, not taken from a published ta