Map Google and Bing Queries to Pages and a GEO Question Bank

A register row keeps two page roles visible

Use the query exactly as exported. The reported page records where the platform attributes impressions; the intended answer page records where the editorial plan places the answer. A difference between them is something to review, rather than a reason to overwrite either field.

The row below is illustrative. The query, pages, question identifier, dates and reviewer are placeholders, with no measured results.

Keep the export or request reference alongside the row so a reviewer can recover the reported pairing. Record aggregation settings with that reference. The article on recording AI visibility test environments describes how to record the surrounding observation conditions.

Google attribution governs the URL in the row

Search Console's Search Analytics API says that page aggregation uses the canonical URI. Its byPage option aggregates by URI; byProperty aggregates at property level. When grouping or filtering by page, the documentation instructs callers to choose auto and rules out property aggregation. Preserve the returned page and the actual aggregation setting; do not substitute the preferred answer URL.

If you group or filter by page, you cannot aggregate by property.

Google Search Analytics API

The scope of available query rows also needs a note. Google's October 2022 explanation retains a specific exception:

While the actual anonymized queries are always omitted from the tables, they are included in chart totals, unless you filter by query.

Google performance data deep dive

The register therefore describes the identifiable exported queries. A chart total does not disclose additional query wording that can be mapped to a bank question. Record whether a query filter was used rather than treating the visible rows as a complete list of every query.

Bing pairs need a page or query context

GetQueryStats is documented as top-query statistics; its displayed response contains a query field without a page field. A top-query list and a separate top-page list cannot establish which query belongs to which page.

Use statistics requested for a specific page or query. GetPageQueryStats takes a page parameter and returns query statistics. Retain that input page with its returned query rows. GetQueryPageStats takes a query parameter; register a pair only when the retrieved material explicitly identifies the associated page. The example on that documentation page does not itself expose a page field, so it cannot supply a page identity.

All three Microsoft Learn pages carry this qualification:

Some information relates to prerelease product that may be substantially modified before it’s released.

Microsoft Learn: GetPageQueryStats

Retain the page or query request context as evidence of the association. Do not manufacture a pairing from the order of two independent lists.

Four labels connect queries to existing questions

Choose the relation by comparing the query's apparent need with the existing question's scope. Here, “narrower” and “broader” describe the query relative to that question. These examples are illustrative; every query and page is fictional.

One question can have several query rows. One page can receive queries associated with different questions. If a row relates to multiple questions, keep a separate question-ID/label link for each under that row; a shared page does not make those questions interchangeable. Add a brief rationale when the wording leaves room for more than one interpretation.

For the distinction between search results and AI answers, see how Google visibility and ChatGPT mentions differ. The step for keeping question wording fixed and the source-gap matrix are listed in Choosing a GEO Agency for Chinese Brands. Keep each bank question's existing wording unchanged during a query review.

Grounding phrases and candidates need separate handling

Bing's 10 February 2026 public-preview announcement describes grounding queries as phrases the AI used to retrieve content referenced in AI-generated answers; the displayed data represents a sample of overall citation activity. This describes the announcement at that time. Record these phrases in the dedicated platform-report field, with the report reference. The field holds a separate Bing report; it does not establish a query–page pairing with the register row. Do not insert these AI retrieval phrases into the verbatim user-search-query column or treat them as a complete inventory. Bing AI Performance announcement

For “no matching question,” send the query, reported page, proposed need and reviewer rationale to a candidate list for the bank owner. Keep the mapping label until that owner decides whether to add a question. A later reassignment should preserve the old link, replacement link, reason and review date. The article on evidence pages for technical questions describes how to place supporting documentation beside technical claims.

At the next review, start with rows whose reported and intended pages differ. Open the reported page, compare its answer with the existing bank question, and record whether the mapping should remain or change. Keep that decision attached to the row.

Related reading

• From Google Rankings to ChatGPT Answers
• Choosing a GEO Agency for Chinese Brands
• Build Evidence Pages for B2B Technical Questions
• Record AI Visibility Environments and Repeat Tests
• How AI Engines Decide Which Brands to Recommend
• E-Commerce GEO in China: Question Bank to 90-Day Plan