Keyword research with Google Search Console and AI (no third-party tools)

Google Search Console already contains your most valuable keyword research data — every query Google has ever shown your site for, with real clicks, impressions, CTR, and average position. Most SEOs export this to a spreadsheet and manually sort through rows.

With Claude connected to Search Console via MCP, you skip the spreadsheet entirely. Ask Claude to analyse, cluster, filter, and prioritise your keyword opportunities in one conversation.

Why GSC data beats third-party estimates for your own site

Search Console shows real data:

• Actual queries users typed into Google
• Actual impressions (Google showed your page)
• Actual clicks
• Actual position for each query

Third-party tools like Ahrefs and Semrush show estimated volumes based on panel data and algorithmic modelling. For your own domain, GSC is more accurate by definition.

What GSC cannot show: competitor keyword data, keywords you have zero impressions for. For gap analysis (finding keywords you don't rank for at all), you still need a third-party tool or to look at competitor-facing queries manually.

Setting up the workflow

Connect Search Console to Claude via GenieSeo MCP — free, one OAuth, 60 seconds. Then open Claude and follow the phases below.

Phase 1: Map your keyword landscape

Understand the full picture first.

Using get_search_analytics, pull the last 90 days of query data for my site.
Give me:
1. Total unique queries I get impressions for
2. Queries split into four buckets: positions 1–3, 4–10, 11–20, 21–50
3. Queries with impressions > 100 in each bucket
Output as a summary table.

This tells you how many keywords you're playing for and where the bulk of opportunity sits.

Phase 2: Find quick wins (the highest ROI action)

Quick wins are queries where you're already visible but not ranking highly enough to get clicks. Positions 5–15 with decent impressions are the most valuable.

From the last 90 days of query data:
- Find all queries with average position between 5 and 15
- Filter for impressions > 100
- Sort by impressions descending
- For each query, show: query text, position, impressions, clicks, CTR

These are my quick-win keywords. I want to know:
1. Which pages are ranking for these queries?
2. What's the gap between my current position and top 3?
3. Which quick wins are easiest to move (lowest competition, highest impressions)?

Phase 3: CTR analysis — bad titles and meta descriptions

High impressions + low CTR = your title/meta description is losing to competitors who have better copy.

From the last 90 days:
- Find queries where average position is 1–5 AND CTR is below 20%
- Find queries where average position is 6–10 AND CTR is below 5%
Sort by impressions. For each, tell me:
- The query
- My position
- My CTR
- What a better title might look like based on the query intent

Phase 4: Group keywords by intent

Manually grouping keywords into informational / commercial / navigational / transactional takes hours. Claude does it in seconds.

Take my top 100 queries by impressions from the last 90 days.
Group them into four intent buckets:
- Informational: user wants to learn something
- Navigational: user wants to find a specific page/site
- Commercial: user is researching before buying
- Transactional: user is ready to buy/sign up

For each group, list the top 5 queries.
Which bucket has the most volume? Which bucket has my lowest CTR?

Phase 5: Identify content gaps from impression data

If you have high impressions but no clear content targeting those queries, you need new pages.

From my top 200 queries by impressions:
- Group queries that appear to be about the same topic (even if worded differently)
- For each topic cluster, tell me:
  a. The main query (most impressions)
  b. All variants in the cluster
  c. Total impressions across the cluster
  d. Which page of mine is currently ranking

Then highlight clusters where: total impressions > 500 BUT there is no clear dedicated page on my site for that topic.
These are content gaps I should create pages for.

Phase 6: Find question queries (featured snippet opportunities)

Questions rank well for featured snippets, People Also Ask, and voice search.

From my last 90 days of query data:
- Find all queries that start with: how, what, why, when, where, which, can, does, is, are
- Sort by impressions
- Show position and CTR for each
- Which question queries am I getting impressions for but ranking below position 3?
These are my featured-snippet opportunities.

Phase 7: Branded vs non-branded split

From the last 90 days:
Split my queries into branded (contains my brand/domain name) and non-branded.
For each group show: total clicks, total impressions, average CTR, average position.

What percentage of my clicks are branded?
What does this tell me about my organic visibility for non-branded queries?

Phase 8: Track keyword velocity (what's growing)

Compare the last 30 days vs the 30 days before that.
Which queries gained the most impressions?
Which queries lost the most impressions?
These are my momentum signals — what topics are growing for my site, what's shrinking.

Complete keyword research session (single prompt)

Run a complete keyword research session for my site using the last 90 days of Search Console data.

Step 1: Quick wins — positions 5–15, impressions > 100, sorted by impressions.
Step 2: CTR failures — position 1–5 with CTR below 20%. List top 10.
Step 3: Intent grouping — categorise top 100 queries by informational/commercial/transactional.
Step 4: Content gaps — topic clusters with > 500 combined impressions but no dedicated page.
Step 5: Question queries — queries starting with how/what/why/when for featured snippets.
Step 6: Momentum — top 10 queries gaining impressions vs prior 30-day period.

Output each step as a Markdown table. End with 5 prioritised action items.

What you can do without paid tools

With only GSC data and Claude:

• ✓ Find quick-win keywords to push from page 2 to page 1
• ✓ Fix CTR problems with better title/meta rewrites
• ✓ Discover content gaps from impression clusters
• ✓ Prioritise featured snippet opportunities
• ✓ Track keyword velocity

What you need a paid tool for:

• ✗ Keywords competitors rank for (no impression data on your end)
• ✗ Volume estimates for keywords you have zero impressions for
• ✗ Backlink analysis

For most SEO teams, the quick wins from GSC data alone generate more return than any paid tool research.

Next steps

• Get GenieSeo MCP URL (free) — connect GSC + GA4 to Claude in 60 seconds
• 50 AI SEO prompts for Google Search Console
• Copy-paste keyword research recipes