Ali Fakhar

How to turn Search Console queries into a content plan using AI

  • Level: Beginner
  • A few hours to set up

The short answer

Export a year of queries with their landing pages, ask a model to group them by what the searcher wants rather than by shared words, compare each group with the page that should answer it, and let an editor decide which gaps need a new page, an update or nothing.

On this page

At a glance

The marketing problem
Search Console holds thousands of queries, but a query list is not a content plan. Similar searches need grouping by the need behind them, with a decision about which page should serve it.
Who this is for
SEO and content teams deciding what to improve or write next.
Tools involved
  • Google Search Console
  • Claude API
  • Screaming Frog

Where I have run it

1,000+ ranking keywords, average position 40 to 11

View case study ↗
Google Search Console, rolling twelve months against the twelve months before the work started. The same programme grew impressions 447% to 4.55 million.

What you need

Before you start

  • Search Console access for the site
  • A list of your existing pages and what each is for
  • An understanding of the buyer’s journey, so clusters can be mapped to stages

What goes in

  • Queries with their landing pages, impressions, clicks and average position, over a comparable period
  • Queries with impressions but few clicks: where you appear but do not satisfy
  • Questions sales and support hear, to check the clusters against

How it works

  1. 01 · ToolExport queries with their pages
  2. 02 · AI stepGroup queries by intent
  3. 03 · AI stepCompare clusters with existing pages
  4. 04 · You decideEditor decides the priorities
  5. 05 · OutputContent plan with a reason attached
  • Tool
  • AI step
  • You decide
  • Output

Step by step

  1. 01 Tool

    Export queries with their pages

    Export queries and pages for the last twelve months, keeping the query-to-page relationship. The interface export stops at 1,000 rows, so use the Search Console API, Looker Studio or a Sheets add-on for more. Keep queries with impressions and no clicks: that is where the gaps are.

  2. 02 AI step

    Group queries by intent

    Ask the model only for a cluster name per query, then total clicks and impressions yourself with a pivot table. Models are unreliable at arithmetic over long lists.

    Prompt to copy

    Below are search queries. Group them by what the searcher is trying to achieve, not by shared words. For example, “how much does X cost” and “X pricing” belong together. Return a table with two columns: query, cluster name. Then list each cluster with a one-line description of the need and its buying stage (learning, comparing, ready to buy). Do not drop queries; put any that fit nowhere in a cluster called “unclustered”. Queries: [paste]

  3. 03 AI step

    Compare clusters with existing pages

    For each cluster, find the page that should answer it and judge whether it does.

    Prompt to copy

    Here are query clusters, the page that currently ranks for each query, and a list of our pages with their purpose. For each cluster say: which page best serves it, whether that page answers the need (yes, partly, no) and what is missing. Suggest “update page”, “new page” or “no action”, with one sentence of reasoning. Clusters: [paste] Our pages: [paste]

  4. 04 You decide

    Editor decides the priorities

    Check overlap, business relevance and whether a new page is really justified. Most clusters do not deserve a page; saying so is the job.

  5. 05 Output

    Content plan with a reason attached

    Each planned piece lists the queries it exists for, the page it creates or updates and the outcome it should affect.

Where a person decides

An editor decides what gets written or updated; a cluster is an argument for a page, not a commission.

What to watch for

  • Search Console hides some low-volume queries, so the export is not all demand.
  • A cluster is a suggestion, not proof of a new page opportunity; check existing coverage first.
  • A weak title and description is a different problem from missing content.

How to tell it is working

  • A sample of clusters is checked by hand for sensible grouping.
  • Proposed pages are checked against existing ones for overlap.
  • Affected pages are tracked over time, separately from seasonal or site-wide changes.

An illustrative example

Input

Illustrative: several queries about the cost of software development land on a broad services page.

Output

A “pricing questions” cluster with its queries, a note that the services page never mentions cost, and a recommendation to add a pricing section before considering a separate guide.

Want this running in your team?

AI workflow implementation. Turn a repetitive marketing task into a workflow your team can use with confidence.

View service
All AI in marketing workflows ↗
Analytics preferences

Current preference: off