Google Search Console
Keyword Research Skill.
Give your agent GSC data. Get defensible new-page opportunities.
Reusable instructions and Python helpers for your AI workspace. Your agent matches search queries to existing content through JEV, separating genuine gaps from topics your website already covers.
This is a package enquiry, not an analysis commission. Price, release contents and delivery terms are confirmed before payment. Model access, GSC connectivity and paid APIs are separate.
The package you buy
SKILL.md instructions, agent configuration, Python helpers, validation rules and references in the agreed release.
The output you build
A query → current page → JEV check → decision table, a new-page candidate queue and evidence behind each recommendation.
Inside Search Console SEO Skill
The source module is gsc-new-page-opportunities. This is an agent toolkit, not a standalone SaaS or access to Google or JEV. You run the method with your authorized data sources.
- export_gsc.py
- Collect available GSC data with dates and query-to-page relationships.
- crawl_site.py
- Inventory website URLs and content to check existing answers.
- discover_tasks.py · assemble_tasks.py
- Discover user tasks and prepare reviewable query groups.
- new_pages.py · jev_batch.py
- Prepare JEV comparisons, preserve responses and support page decisions.
- spend_budget.py
- Control the approved monetary budget for paid requests.
- SKILL.md · references/ · agents/
- Workflow instructions, evidence rules, configuration and supporting documentation.
How your agent uses Google Search Console for SEO
You provide the website, market and authorized access or exports. The agent performs the research; this is not a long manual checklist that leaves every action to you.
- 1
Export 16 months of Google Search Console data
The agent collects available finalized queries, pages, clicks, impressions, CTR and average position. It records exact dates, country, device and search type. Shorter exports stay labelled partial.
- 2
Match queries to the website
The agent groups search queries by user task, reads observed landing pages and checks other current URLs. The sitemap, internal discovery and GSC URLs contribute to the inventory.
- 3
Check the match through JEV
The agent sends the task and page content to TypeSafe JEV. JEV checks whether a page answers fully, partly or not at all. Query membership and duplicate tasks receive separate checks.
- 4
Return a decision ledger
A suitable existing URL stays. A distinct uncovered task becomes a new-page candidate. Unclear judgments remain in review; the agent does not create a page for every keyword.
Example: engine repair
Illustrative example: matching engine-repair queries to existing and proposed service pages.
| Search query | Observed URL | JEV content check | Decision |
|---|---|---|---|
| engine repair | /engine-repair/ | The page answers the task | Keep the existing page |
| engine diagnostics | /engine-repair/ | Same service in this example | Keep in the same cluster |
| diesel engine repair | /engine-repair/ | Distinct service; check every other current page | Candidate /diesel-engine-repair/ if no suitable URL exists |
In a real project, diagnostics may deserve its own service page. Business scope, current content and comparable search results decide. JEV provides model judgments, not a Google verdict. A mismatch with one landing page does not prove that the website lacks an answer elsewhere.
Download teaching table · CSVWhat the agent needs from you
Do not share a password. Direct collection requires an authorized connection to the selected Search Console property; suitable retained exports provide an alternative input.
You provide
Website, real services, target country and language, selected GSC property or export with dates and filters. Approve specific data sources and spending limits.
Your agent runs
Data collection, website reading, search intent grouping, approved JEV checks, other-page matching and a traceable decision log. It preserves failed extraction and unanswered questions.
You approve
Ambiguous tasks, business priorities and next actions. Content writing, existing-page updates and publication are outside this skill. No permissions or website content change silently.
Not a list of ideas. An inspectable result.
The purchase gives you the package. Running the agent with it produces these files; completeness depends on the supplied data and completed checks.
- new-pages.csv
- A queue of candidates and qualified new-page proposals.
- query-decisions.csv · task-decisions.csv
- Queries, user tasks, decisions and reasons, including exclusions.
- page-relations.csv · membership-checks.csv
- Task-to-page coverage and query-cluster membership checks.
- duplicate-pairs.json · qa.json
- Duplicate tasks, unresolved evidence and QA status.
Each recommendation retains query members, observed URL, other matching pages, proposed destination, state and missing evidence. EXISTING means a current answer; REVIEW_SECTION means a possible section review; CANDIDATE is a hypothesis; CREATE means qualification gates passed, not a published page. A workbook is prepared separately in an agreed format.
Google Search Console keyword research: the evidence layer
The package gives your agent rules for interpreting search performance, not just exporting a keyword list. Every content opportunity must connect a real search query to page evidence and a useful business task.
Search queries and ranking pages
The agent uses Google Search Console Performance report data: query, clicks, impressions, CTR and average position. A separate query-and-page export identifies which URL appears for each search term. Property, dates and filters stay with the dataset so you can review the keyword research rather than trust an unexplained AI summary.
Clicks, impressions and average position
An impression records exposure of your website in Google Search, not total keyword search volume. Click-through rate is clicks divided by impressions. Average position depends on the date range, country, device and search results. A keyword ranking near page one or a low CTR is a reason to investigate, not automatic evidence that you need new content.
A consistent keyword research dataset
To use Google Search Console for keyword research, the agent retains available finalized data over the selected 16 calendar months. Query/byProperty and query,page/byPage exports remain separate: page impressions cannot be summed into property-level keyword totals. Pagination can retrieve more returned rows, but the API still does not guarantee every query. Short exports, hidden queries and failed requests remain explicit limitations.
From a keyword opportunity to a page proposal
A new page needs a distinct search intent, business fit, no suitable answer elsewhere on the website and a concrete useful-content plan. The agent checks comparable Google search results and duplicate proposals. A related keyword may belong on an existing page; a broader keyword cluster is not automatically a new URL. Until all required evidence is reviewed, the opportunity stays a candidate rather than CREATE.
This Google Search Console SEO workflow helps prioritize defensible new content opportunities. It does not replace Search Console, measure the entire keyword market or make a model judgment a ranking guarantee. Existing-content optimization stays with the separate page-update skill.
Ranking, clicks and low CTR: what the agent checks
The agent uses real search data to open an evidence check. Each keyword opportunity keeps its exact search query, current page and reason for the next action. Good keywords are business-relevant tasks, not simply the biggest number in an export.
Keywords with high impressions
A high number of impressions puts a query into review, not straight into a writing queue. If the existing page answers the task, the agent retains that URL. Keywords with low CTR may reflect ranking position, result features or search intent. Any snippet improvement is referred to a separate workflow rather than answered with a duplicate page.
Long-tail keywords with a distinct service
A specific keyword can reveal a separate service, such as diesel engine repair. The agent checks the literal user task behind the keyword, then compares it with the whole website. A modifier alone does not justify another page. The output records whether the task really needs distinct useful content or fits an existing service.
Ranking keywords on a weak landing page
Keywords you are already ranking for can still land on a broad page. JEV reads the content instead of inferring relevance from ranking metrics. If another current URL answers fully, the agent records that destination. Otherwise the keyword remains a candidate with its missing qualification checks, not a promise of more search traffic.
One keyword cluster, several current URLs
The agent preserves query-to-page relationships and checks duplicate tasks. Multiple ranking pages do not automatically prove cannibalization: they can satisfy different needs. The skill records page relations without silently merging content or creating redirects. You review the keyword strategy before implementing changes to the website.
An SEO keyword decision you can inspect
Every output row preserves the original search term, task cluster, observed URL, other matching pages, JEV judgment and unresolved questions, proposed destination, state, reason and dated search metrics. You can review the recommendation without repeating the keyword search. Candidate counts do not become forecasts of organic search traffic to your site.
Your Google Search Console account and source checks
The agent uses the selected Google Search Console account through an agreed access route. It retains the property, filters and data dates. Domain verification through DNS or an HTML file, submitting a sitemap to Google Search Console and changing permissions are not silent package actions. If you need to check whether a page is indexed by Google, the URL Inspection Tool is a separate diagnostic, not an inference from a missing keyword row.
Recent 7, 28 or 90-day views can show how the site performs in Google Search alongside the main research window. They stay separate from the 16-month dataset. Google Trends or other search engines may offer additional context, but they are not relabelled as Google Search Console data. You review what is happening in organic results before approving a page proposal.
Where the skill stops
Not the entire keyword market
Google Search Console shows queries your site appeared for. Impressions are not market-wide search volume. Topics without observed visibility need separate keyword research.
Not a new URL for every phrase
Current content, duplicate tasks and comparable search engine results pages must support a distinct useful destination. A partial answer may need a section instead of another page.
No hidden evidence gaps
Anonymized queries and API limits mean not every row is available. Failed page extraction and unclear model responses stay visible rather than becoming successful checks.
No growth or publishing guarantee
Candidate count is not required page count or an organic traffic forecast. Implementation needs separate approval and outcome measurement.
How the skill fits your SEO tools
Keep each source in its proper role. Buying the package does not automatically include an account, integration or subscription to another SEO tool.
Search Console versus a keyword tool
Google Ads and Google Keyword Planner, Semrush or other approved sources can extend keyword discovery beyond your existing search visibility. Their market estimates are not interchangeable with GSC impressions. Tools like these need a configured connection or suitable export; naming a provider does not establish a shipped universal adapter. The skill helps you find gaps in current coverage, while broader keyword opportunities belong to a separate discovery run.
Analysis versus content optimization
Google Analytics can add landing-page conversion context when separately authorized, but it does not replace Search Console keyword data. An index check, title or meta description rewrite, structured data change and internal linking implementation belong to separate SEO strategies and workflows. A page can appear in search results without fully answering the task. Ranking metrics alone cannot tell the agent whether its content is useful; it must read the page before recommending another destination.
Start with your environment
Tell us your AI workspace and whether you will use an authorized Search Console connection or an export. Before purchase, we confirm the setup, Python requirements, GSC/JEV access and launch instructions for your environment.
Request Search Console SEO Skill
Confirm the price, included files, delivery method, usage rights for own or client projects, updates, support and applicable refund terms before payment. Checkout and automatic delivery are not connected here.
I would like the Search Console SEO Skill package. My AI environment: [environment]. Website and market: [details]. GSC: [access/export]. Please confirm price, contents, dependencies, delivery, usage rights and support terms.
Before you buy: FAQ
Am I buying a finished website analysis?
No. You receive the agreed reusable agent package. Analysis is the output of applying it in your own environment. Research carried out by Dima is a separate engagement.
Can I use Google Search Console for keyword research without direct access?
A suitable export can support the workflow when it preserves property, dates, filters, queries and page relationships. A query-only keyword list cannot establish missing pages. Import limitations remain visible in the report.
Are JEV and paid data APIs included?
No. Use your authorized connections and a separate budget. Request counts and costs are approved before dispatch. An access failure cannot become a successful evidence check.
Will the agent optimize existing content?
Not with this skill. It records existing coverage and refers possible improvements to a separate page-update workflow. Writing, redirects, content optimization and publication are different tasks.
Can I use the skill on client projects?
Usage rights, support and update terms are confirmed before payment. They do not automatically transfer from another package.
What does Google Search Console give you, and what does the skill add?
Google Search Console is a free Google tool for observing how your site appears in Google. The skill does not sell access to that service. It gives your agent a repeatable method to analyze Search Console data, read pages, check coverage through JEV and retain the reasons for recommendations. Artificial intelligence assists the analysis; it does not gain access to hidden Google ranking signals. Search Console helps you understand observed search performance, while the package adds content matching and reviewable page decisions. It does not create a Google account, verify a domain or change permissions. Nor does it turn your property data into a competitor keyword database. Instead of a generic suggestion to add content, the output connects a proposed page to an observed keyword and a specific uncovered task.