Search Intent Mismatch Audit: Search Console Notebook for Site Architecture

A page can rank well and still attract the wrong searchers. A pricing page that mostly shows up for “what is” queries, or a guide that pulls in people ready to buy, is a sign your site structure and your search intent have drifted apart. Checking that by eye works for ten pages. This free Colab notebook checks the whole site from a Search Console export.

Who it’s for

In-house SEOs and consultants auditing site architecture, or anyone trying to work out why a page ranks for queries it was never written for.

How to use it

1. Load your Search Console data. Upload a page + query export with clicks, impressions and position, or run the built-in example (a fictional SaaS site) first.
2. Map pages to page types. Edit the URL rules that sort pages into resource, company and product pages, and the intents each type should serve.
3. Classify the queries. Rule-based filters label each query informational, commercial, transactional or navigational. Brand + modifier queries take the modifier’s intent and are flagged as branded.
4. Flag mismatches. Each page gets a click-weighted (or impression-weighted) intent mix, a dominant intent and a verdict: match, mismatch or mixed.
5. Check the KPI trade-offs. Clicks, impressions, CTR and average position per page type, plus sessions, engagement rate and key events if you add a GA4 landing-page export.
6. Find intent bridges. For each resource page, the product page with the most shared query vocabulary: your first candidate for a contextual link or CTA.
7. Export queries_with_intent.csv, page_intent_audit.csv and intent_bridges.csv.

What you need

• A Search Console export with page and query in the same table. The web interface can’t combine them, so use the Search Console API, the Search Analytics for Sheets add-on or the BigQuery bulk export.
• Optionally, a GA4 landing-page export with sessions, engagement rate and key events.
• No API keys and no cost. It runs on pandas in the free Colab runtime.

What you get

A list of pages whose search demand doesn’t match their job on the site, the numbers to show why it matters per page type, and a shortlist of links that move readers from learning pages to buying pages.

Limits

• The intent filters are deliberately simple and English-only. For other languages, swap in the multilingual intent filters from the same repo.
• The intent bridge score is word overlap (Jaccard similarity). Treat it as a shortlist for a human, not an automatic internal linking rule.
• A mismatch isn’t always a problem. Sometimes the fix is a new page; sometimes the page is doing exactly what you need.

The thinking behind the audit is in How to Build a Search Intent-First Website Architecture, and intent work is covered in depth in the Semantic ML-enabled Keyword Research course.