Search Console query insights: low CTR, striking distance and cannibalization

What it does

Analyzes a Google Search Console Performance export locally (nothing is uploaded).

• Finds queries and pages with many impressions but low CTR, compared with your own site's CTR at the same position tier, not a generic CTR curve
• Finds striking-distance queries (average position 5-20) and estimates extra clicks if they reached the CTR of the next position tier
• Detects keyword cannibalization: one query served by two or more of your pages (needs data with both query and page columns)
• Writes a prioritized table (Markdown, CSV, JSON); the agent then proposes concrete title, content and consolidation changes

English and Japanese column headers are recognized, and the export ZIP can be read directly.

How to use

python scripts/gsc_insights.py export.zip --exclude-regex "brand" --out ./gsc-report

Ask your agent to find improvement opportunities in your Search Console data and give it the export. Use --exclude-regex to keep brand queries out of the benchmarks.

Requirements

• Python 3 (standard library only)
• A Search Console export (ZIP or CSV)

Caveats

• Extra-click numbers are rough estimates derived from your own data, not forecasts.
• The standard UI export keeps queries and pages in separate files, so cannibalization needs a query+page export (API, Looker Studio or BigQuery).
• UI exports are capped at 1,000 rows per table and omit anonymized queries.