LearnGrokUse one query export and one page inventory to produce a list of pages to create, update, merge or retire. This is for content marketers and SEO leads who need to make an editorial decision, not present a large spreadsheet.
The habit is simple: review the same fields, group queries by search need, then assign one action to each page theme. Keep the source files and the final decisions together so you can revisit them after the next reporting period.
1. Prepare two small working files
Export your search query data for a consistent period. Keep these columns where available:
- Query
- Clicks
- Impressions
- Average position
- Landing page
- Country or market
- Device, if it changes the decision
Create a separate page inventory. One row should represent one indexable page or one clearly defined page theme. Include:
- URL
- Page title
- Content type, such as guide, category page, comparison page or product page
- Primary topic
- Intended audience
- Publication or last substantial update date
- Current owner
- Your initial verdict: keep, update, merge, redirect or retire
Remove branded navigation queries, internal site searches, obvious misspellings and queries outside your market before asking the model to review the data. Do not remove awkward queries merely because they do not fit your current plan. Those are often the useful ones.
Watch out
Keep query and page data separate first
A landing page is evidence of what currently ranks. It is not proof that it is the best page to serve the query.
If the export is too large to inspect properly, split it by topic, market or directory. Preserve the column headings in every part. The model you are using can help structure the review, but its file and input limits are version-dependent. Check the current guidance in the xAI documentation overview.
2. Define the decision rules before reviewing queries
Write the rules in the report before you look for opportunities. This stops a high-impression query from automatically becoming a new article.
Use these actions:
| Action | Use it when | Output you need |
|---|---|---|
| Create | A repeated query theme has no suitable page | Proposed page, audience, angle and primary query theme |
| Update | One existing page fits the need but is incomplete, stale or poorly aligned | Brief with sections to add, remove or rewrite |
| Merge | Several pages answer the same need and split relevance | Surviving URL, merge plan and redirect list |
| Retire | A page has no distinct purpose and no query theme worth preserving | Replacement destination or removal decision |
| Monitor | The evidence is not yet clear | Review date and the missing evidence |
A page can have clicks and still need replacing. A query can have many impressions and still be unsuitable if its intent is unrelated to what you offer. Record that distinction in the report.
Key point
Decide by search need, not by keyword count
Your priority list should say which page best serves a group of related queries and what must change on that page.
3. Group queries into page themes
Sort the query export by impressions, then review the queries in groups rather than one at a time. A page theme is a cluster of queries that a single useful page could answer without becoming a catch-all.
For each cluster, record:
- The representative query, usually the clearest wording rather than the largest number
- The search need: learn, compare, choose, solve, find or buy
- Modifiers that change the need, such as audience, industry, location, template, example, cost or alternative
- The current landing page or pages
- Whether the existing page actually answers the query
- The recommended action
Give the model a bounded task. Paste a topic-sized batch, your page inventory rows for that topic, and this prompt:
Group these search queries into distinct page themes. For each theme, identify the likely search need, relevant existing URLs, and one action: create, update, merge, retire or monitor.
Do not assume the current landing page is correct. Keep themes separate when the audience, task, format or decision stage differs. Flag uncertain groupings.
Return a table with: page theme, representative queries, search need, existing URL, evidence from the export, recommended action, and rationale.
Review the table yourself before accepting it. The model can spot repeated language and apparent overlap. You decide whether your business can credibly meet the need and whether a new page would duplicate an existing one.
4. Turn themes into a priority list
Do not make a long backlog sorted only by impressions. Put each approved theme into a decision table with a clear owner and next step.
| Priority | Theme | Action | Why now | Next deliverable |
|---|---|---|---|---|
| High | [theme] | Update | Relevant page already receives impressions but misses the query need | Update brief |
| Medium | [theme] | Create | Clear need, no suitable existing page | Page brief |
| Low | [theme] | Merge | Overlapping pages create a confused route to the answer | Consolidation plan |
| Review | [theme] | Monitor | Intent or page fit remains unclear | Check next export |
Set priority using evidence you can explain in one sentence. Useful evidence includes repeated query demand, a close existing page that fails to answer the task, multiple competing URLs, or a strategically important audience modifier. Avoid a combined score unless every input is defined and the score leads to a different decision.
For each high-priority item, create one brief with:
- The page action and target URL.
- The primary search need and intended reader.
- Queries the page must answer.
- Existing content to retain, remove or redirect.
- The page angle that makes it distinct from adjacent pages.
- The person responsible and the review date.
Note
Treat retiring as a content decision
Before retiring a page, check whether another page should absorb its useful material and whether visitors need a clear destination.
5. Check whether the report is wrong
A neat cluster is not necessarily a valid one. Check a sample of queries from every high-priority theme, especially the queries with the most impressions and the most unusual modifiers. Read the current landing page beside them.
The report is likely wrong when you see this:
| What you see | What it usually means | What to do |
|---|---|---|
| One theme contains queries for different audiences | The grouping is too broad | Split by audience and reassess the page action |
| A proposed new page has an existing near-match | The inventory is incomplete or titles hide overlap | Read both pages and choose update or merge before creating |
| Queries and landing page have different tasks | The current ranking is incidental | Do not optimise the page until you identify the right page type |
| A theme is based on one unusual query | The evidence is weak | Mark it monitor, then review later |
Check
The list is usable when every high-priority row names one page, one action and one reason another person can verify in the source files.
Also check that your actions add up. If ten themes all require new pages, ask which existing pages should have been updated instead. If nothing is marked for retirement or merging, check whether the inventory has been reviewed honestly.
When the process does not work
If the query export lacks landing-page data, make the report provisional. Group the search needs first, then manually map only the high-priority themes to the site. If your inventory is outdated, rebuild it for the relevant directory before recommending mergers or retirements.
If the model produces vague themes, reduce the batch size and provide two or three examples of page themes you consider distinct. If it recommends actions without evidence, require it to quote the query group and the relevant inventory row. Keep uncertain items in monitor rather than forcing a create or retire decision.
Repeat this review on the same cadence and keep prior reports. The useful output is not just this month's priority list. It is a record of why each page was created, changed, merged or left alone.
