The short answer
Do not sort Search Console queries by impressions and call the first 20 rows your content calendar.
A defensible opportunity combines five signals:
Observed demand: relevant queries are already generating impressions.
Existing visibility: your site has a real foothold, not a hypothetical keyword score.
Intent fit: the searcher’s job matches something your business can help with.
Page fit: an existing page either serves the intent, partly serves it, or clearly does not.
Confidence: the pattern survives a sensible date comparison and is large enough to act on.
AI is useful for grouping messy query rows, comparing periods, joining queries to landing pages, and applying a consistent scoring rubric. A practitioner still decides whether the right action is to update a page, create a distinct page, consolidate overlap, or leave the result alone.
What counts as an SEO content opportunity?
An opportunity is not merely a keyword with impressions. It is a supported editorial decision.
Evidence pattern | Likely decision | Human review gate |
|---|---|---|
A relevant page appears for a coherent query cluster, but its visibility or clicks lag its own recent baseline | Update the existing page | Confirm the page genuinely serves the intent and inspect the current search results before rewriting titles or content |
One page earns impressions for related questions that its content only partially answers | Expand the existing page | Add material that helps the reader complete the job; do not bolt on unrelated headings for coverage |
A distinct, valuable intent appears across multiple related queries and no current page serves it well | Create a new page | Verify the intent is meaningfully different from existing coverage and that the team can add original value |
Several pages split visibility for the same intent or the preferred page changes repeatedly | Consolidate or clarify ownership | Check canonicals, internal links, page purpose, and search results before merging or redirecting anything |
The query is irrelevant, navigational, branded, seasonal, low-sample, or already well served | Monitor or ignore | Record why it was excluded so it does not reappear as a “new” idea next month |
The same row can support different decisions in different businesses. A position of 12 may be attractive for a high-value, well-matched topic and meaningless for an irrelevant query. Likewise, a low click-through rate is not automatically a bad title: result features, intent, brand familiarity, device, country, and ranking position all shape CTR.
A practical Search Console opportunity workflow
1. Define the market and decision first
Start with one Search Console property, one search type, and a clear audience or market. For most editorial research, use Web search and a country filter only when the content is intended for that country.
Then define the decision you are trying to make. Examples:
Which existing pages should we refresh this month?
Which query clusters deserve a net-new page?
Where are two pages competing for the same intent?
Which declining topics need investigation before we change content?
This keeps the analysis from turning into a generic list of “interesting keywords.”
2. Use complete comparison periods
For a routine review, compare two equal 28-day periods and exclude the newest few days. Google says Search Console data is typically available after two to three days, and daily data outside the 24-hour view is reported in Pacific Time.
Use finalized data for planning. Fresh data is useful for launches or directional checks, but it should not drive a content rewrite before it settles.
Seasonal sites should also compare year over year. A month-over-month decline in a seasonal topic can be normal; a year-over-year loss in the same weeks is a different signal.
3. Pull a query-level discovery report
Retrieve query, clicks, impressions, CTR, and average position for the current and comparison periods. Keep the property, dates, search type, country, device, and data state in the output.
If the brand has a distinctive name, create branded and non-branded views. Treat the split as approximate: Google notes that anonymized queries are omitted when query filters are applied.
Do not ask the AI to diagnose pages yet. First use query rows to find coherent themes:
relevant clusters gaining impressions but not clicks
clusters with improving visibility and enough demand to justify attention
important clusters losing both clicks and position
repeated questions adjacent to a page’s current purpose
commercial or problem-aware language that maps to a product or service
Average position is a diagnostic clue, not a fixed rank. Google calculates it from the topmost position of a property’s result and notes that it does not always correspond to a single stable on-screen placement.
4. Pair shortlisted queries with pages
Now rerun the report using both query and page dimensions for the shortlisted themes. This answers the question the query-only report cannot: which URL is Google already associating with this intent?
Use this paired view selectively. Adding dimensions fragments the data into more rows, and the Search Console API returns top rows rather than a guaranteed complete dataset. HireOtto can export up to 25,000 rows from a performance request, but that is still a top-row sample – not a complete crawl of every query.
Look for three page relationships:
Clear owner: one relevant page receives most of the cluster’s visible demand.
Weak owner: a page appears, but its purpose or coverage only partially matches the cluster.
Split ownership: multiple pages appear for the same intent, or the leading page changes between periods.
Split ownership is a prompt to investigate, not automatic proof of keyword cannibalization. Different URLs may appropriately serve different sub-intents.
Ask the AI to group queries by the job the searcher is trying to complete. Queries that share a noun can still require different pages.
For example, these belong near one another but do not necessarily share one page:
“google search console mcp” – category or product investigation
“analyze search console queries with ai” – workflow intent
“check if url is indexed” – narrow troubleshooting intent
“search console api row limit” – technical reference intent
Require the AI to name the intent, show the supporting queries, identify the current page owner, and explain any ambiguous rows. A cluster without representative query evidence should not enter the queue.
6. Inspect the existing coverage
Search Console describes performance; it does not read your live page and decide whether the content is good.
For every promising cluster, review the current page, its title and main sections, the page’s intended audience, internal links, and the current search results. Also check the URL’s indexed status when indexing or canonical selection might explain the data.
This is the point where many “new content opportunities” disappear. If a relevant page already answers the job well, the right action may be better distribution, stronger internal linking, a clearer snippet, or simply more time – not another article.
Google’s guidance is a useful guardrail: content should primarily help people, demonstrate original value, and leave the reader able to achieve their goal. Search data can reveal demand; it does not excuse thin or duplicative production.
7. Score evidence, fit, and actionability
Use a small rubric rather than a single magic metric. Score each factor from 0 to 3 and keep the underlying evidence beside the score.
Factor | 0 | 1 | 2 | 3 |
|---|---|---|---|---|
Demand evidence | Too little data to interpret | Small or volatile cluster | Consistent impressions | Strong, sustained or growing cluster |
Visibility foothold | No meaningful visibility | Early or inconsistent visibility | Clear page-two or lower page-one foothold | Strong visibility with a specific click or coverage gap |
Intent and business fit | Irrelevant | Peripheral | Relevant | Directly useful to the audience and business |
Coverage gap | Existing page already satisfies the job | Minor edit only | Material expansion needed | Distinct intent with no suitable page |
Confidence | One weak signal | Mixed or immature data | Pattern holds across cuts | Pattern holds across periods, pages, and market context |
Do not let the total score choose the action by itself. A high score can still recommend an update rather than a new page. Add effort, strategic priority, and potential conflict with existing pages before scheduling work.
8. Build four editorial queues
The final output should separate decisions:
Update: improve an existing page’s usefulness, structure, title, snippet, or internal links.
Create: build a page for a distinct intent that current coverage does not serve.
Consolidate: resolve overlapping purpose, duplication, or unclear page ownership.
Monitor: wait for more data, seasonality, indexing stability, or a clearer business case.
For each item, include the query cluster, current page owner, comparison evidence, proposed action, reason, confidence, human check, and next review date. That turns an AI analysis into an editorial backlog someone can actually review.
The exact prompt
Copyable prompt
Connect to my Google Search Console data. Do not make changes.
First list the properties I can access and wait for me to select the correct property. Then build an SEO content-opportunity queue for Web search in my target country.
Compare the most recent complete 28-day period ending three days ago with the preceding 28 days. Use finalized data and state the exact inclusive dates.
Pull query-level clicks, impressions, CTR, and average position for both periods. Keep the property, search type, country, device, data state, returned-row count, and truncation status visible.
Ask me for brand terms and exclude them only for the non-branded view. Warn that query filtering removes anonymized queries, so the branded split is approximate.
Group relevant non-branded queries by search intent, not merely shared words. For each cluster, show representative queries and current-versus-previous metrics. Flag small denominators and volatile percentages.
Shortlist clusters with useful demand, an existing visibility foothold, a meaningful decline, or a clear business fit. Do not use a universal CTR or position cutoff.
For shortlisted clusters, run a query-and-page report to identify the current page owner. Classify ownership as clear, weak, split, or absent. Do not label split ownership as cannibalization without reviewing the pages and search results.
If I provide a current content inventory or public page URLs, compare each cluster with the page’s purpose and coverage. Otherwise mark content-fit conclusions as unverified.
Score demand evidence, visibility foothold, intent and business fit, coverage gap, and confidence from 0 to 3. Show the evidence behind every score.
Produce four queues: Update, Create, Consolidate, and Monitor. For every item include the query cluster, current page, evidence, proposed action, reason, confidence, human review gate, and next review date.
End with the five highest-priority actions and the exact evidence that would change each recommendation. Do not draft or publish content unless I ask separately.How HireOtto supports the workflow
HireOtto can run the evidence-gathering portion of this workflow inside Claude, ChatGPT, Make, Grok, Perplexity, or another supported MCP client.
It can list the Search Console properties you can access, run filtered performance reports across query, page, country, device, date, hour, and search-appearance dimensions, and return clicks, impressions, CTR, and average position. For larger analyses, it can provide a CSV export. It can also list submitted sitemaps and inspect Google’s indexed version of a URL when indexing or canonical selection needs a closer look.
The Search Console connection is read-only. HireOtto cannot edit a page, publish content, submit a sitemap, request indexing, or run a live URL test. It also cannot prove that an opportunity deserves a new page without reviewing the live content, the search results, and the business context. Keep those as human review gates, and treat a large export as top-row Search Console data rather than a complete keyword census.
For setup and exact parameters, use the Search Console workflow guide and Search Console tools reference. To start in a new AI client, follow Connect a HireOtto server.
What a useful opportunity queue should contain
The deliverable should make the editorial decision auditable:
exact property, search type, market, device scope, and inclusive dates
current and comparison metrics, with small-sample warnings
representative queries for each intent cluster
the current page owner and whether ownership is clear, weak, split, or absent
an Update, Create, Consolidate, or Monitor decision
evidence and confidence – not just an AI score
a human review gate and next review date
If the output contains only keywords and suggested titles, the workflow stopped too early.
Common mistakes to avoid
Treating high impressions as demand you should chase
An impression only means a result was shown under Google’s measurement rules. The query can be irrelevant, too broad, or poorly matched to your business. Validate intent before assigning work.
Using one universal CTR benchmark
CTR changes with position, result format, brand, device, country, and intent. Compare a page or cluster with its own history and similar queries before calling it weak.
Creating a page for every query variation
Cluster by task. Several phrasings may belong on one useful page; one phrase can also hide several distinct intents. A keyword list is not an information architecture.
Ignoring missing and truncated data
Search Console omits anonymized queries, and its API does not guarantee every row. Query filters can remove additional anonymized data. Keep row counts, filters, and truncation status visible.
Asking AI to infer page quality from performance rows
Clicks and position cannot tell you whether a page is accurate, original, current, or satisfying. Read the page and inspect the live search landscape before changing it.
Frequently asked questions
What Search Console positions are best for content opportunities?
There is no universal cutoff. Many teams begin with queries that already have a visible foothold – often lower page one or page two – but intent, demand, page fit, and business value matter more than an arbitrary position band.
Is high impression and low CTR always an optimization opportunity?
No. It can reflect ranking position, result features, an informational query with no click need, an irrelevant match, device or country mix, or weak title and snippet alignment. Inspect the search results and the page before choosing the fix.
Should I update an existing page or create a new one?
Update when the existing page already owns the intent and can satisfy the job without losing focus. Create when the query cluster represents a distinct task or audience that the current page cannot serve cleanly. Consolidate when multiple pages overlap without a defensible purpose.
Why do query rows not add up to the chart total?
Search Console protects user privacy by omitting anonymized queries from tables. Filters and grouping can also change the visible totals, and the API returns top rows rather than guaranteeing every row.
Can AI automatically build the content calendar?
AI can build a review-ready queue and explain the evidence behind it. A practitioner should still validate intent, existing content, search results, originality, business priority, and production capacity before committing the calendar.
About Me
I’m Suyash – badminton junkie, ex‑GroupM ad‑ops grunt, first marketer at a B2B SaaS startup, and creator of Hiretto: MCP servers for performance marketers.
My mission: less clicking, more thinking.
Let’s build leverage together.