Use query and page evidence to choose a specific article improvement

Start from a page and its questions

Choose one URL and inspect its queries over a consistent period. Group product-name queries by the question asked: pricing, free allowances and cancellation are different jobs. Look at mismatched impressions as well as the phrases already generating clicks. A site-wide query list can obscure which answer needs work. Starting with one page makes the relationship between reader questions and existing scope easier to inspect. The useful deliverable is a concrete editorial decision, rather than a long spreadsheet that nobody can map back to a page.

Locate the answer in the body

For each question, identify the paragraph that answers it. A keyword in a heading is not sufficient. A free-allowance question, for example, may require the distinction between a trial and ongoing free use plus what happens beyond the allowance. Research missing evidence, or link to an existing complete answer when appropriate. Not every phrase needs its own URL. Sometimes a clearer prerequisite or limitation inside the current guide answers the reader better and creates less maintenance than another narrowly differentiated page.

Fix the comparison boundary

Keep search type, country, device and date comparisons consistent. Move from a whole-property change to relevant page groups and individual URLs. Where a business has strong weekly or seasonal patterns, avoid selecting a convenient short window that masks those differences.

Do not treat the query table as complete demand

Some queries are omitted for privacy and the query table is not a complete reconstruction of the chart total. Keep that limitation when reporting a filtered segment. Combine visible queries with the page’s purpose and other audience evidence instead of treating the list as an exhaustive demand map.

Example: two pages attract a similar question

When two pages attract a similar question, do not immediately delete one. Inspect whether their purposes differ, such as overview versus plan selection. Consider consolidation when they repeat the same answer; clarify headings and navigation when they have distinct jobs. Shared query wording alone is not a diagnosis. The practical question is whether readers reach the explanation they need.

Choose evidence-led additions

Specify a missing decision input rather than asking for more detail generally: plan eligibility, billing period, output restriction or an export step during migration. Do not invent a feature or price simply because a query suggests readers want it. If evidence is unavailable, narrow the claim. When a question lies outside the guide’s purpose, clarify that boundary and point to an appropriate next page rather than continuously broadening the article. This keeps a focused answer usable and prevents search exposure from dictating unsupported content.

Prioritize with impact and effort

A frequently displayed question may need no edit when the page already answers it well, while a low-exposure contractual error still deserves correction. Prioritize reader impact, obtainable evidence and effort together. Shortly after publication or migration, keep technical indexing checks separate from judgments about the body and avoid forcing conclusions from sparse data. Missing metrics should remain unavailable, not become zero. This makes the work queue explainable and prevents a simple sort by impressions from outranking necessary factual maintenance.

Review the same question group later

Record release date, edited sections, checked sources and the question group to revisit. Later, inspect impressions and query composition alongside clicks. New exposure can change total CTR or average position, so an aggregate movement is not automatically the effect of the rewrite. Consider demand and wider site changes as well. Even an inconclusive result is useful when the delivered change is traceable. The record reduces repeated investigation and helps distinguish a genuinely new problem from the same unresolved hypothesis returning at the next review.

• Google Search Console: choose a page-level investigation from search evidence