Before choosing the next article, check which hostname your Search Console rows describe

Historical evidence and scope

Read the URL before treating a Search Console page row as evidence about your public website. Our saved DoneHat domain-property export had 36 page rows: 16 for donehat.com and 20 for staging.donehat.com.

The window was 23 September–20 October 2026, retrieved 22 October. The staging ChatGPT visibility guide recorded 140 impressions in its returned page row. That is a clue about a staging URL, not 140 impressions for the production guide or proof of a later traffic recovery.

Keep the number attached to its page

Each row below lists the page URL, impressions, then clicks.

donehat.com/resources/board-ready-ai-visibility-reporting — 9 — 0

staging.donehat.com/resources/board-ready-ai-visibility-reporting — 25 — 0

staging.donehat.com/resources/chatgpt-brand-visibility-guide — 140 — 0

donehat.com/blog/ai-summary-clicks-pew-search-evidence — 5 — 0

Those are selected page rows for the same 28-day window. Do not combine production and staging URLs into one production-page result. Zero clicks in a small returned row also does not explain reader intent or content quality on its own.

A query is a clue, not a page join

The export separately returned 42 query rows. “How to see brand visibility in chatgpt” had 25 impressions and no clicks. A punctuation variant ending in a question mark had 21 impressions and no clicks.

The rows suggest a useful question to answer. They do not count unique buyers. Because query and page responses were requested separately, this export does not establish which page appeared for either query. Topic similarity is not enough to join them.

A 23 October read-only extraction retrieved the aggregate directly, without grouping dimensions, for that same historical window: 14 clicks and 288 impressions for the domain property. The deployed integration and a separate direct provider read agreed on those counts.

That total covers the domain property, including its subdomains. It is not a production-host total. It also is not a sum of the older page rows: property and page groupings have different aggregation rules, and available detailed rows can omit data.

Join query and page only when both were requested

The 23 October extraction also requested query and page together, restricted to production hostnames. It returned 13 available pairs. Four pairs pointed to our existing board-reporting guide:

Each row below lists the query, production page path, impressions, then clicks.

do any platforms produce a board-ready ai search visibility summary? — /resources/board-ready-ai-visibility-reporting — 3 — 0

how can ai visibility reports be made client- or board-ready? — /resources/board-ready-ai-visibility-reporting — 1 — 0

how to report on ai search visibility to the board — /resources/board-ready-ai-visibility-reporting — 2 — 0

reporting ai visibility to the board — /resources/board-ready-ai-visibility-reporting — 1 — 0

Those are real query/page relationships for that window. Their seven returned impressions are sparse evidence, not seven buyers or a property-wide total. The useful next step is to improve the existing guide with a report worksheet, rather than add a near duplicate.

Turn the rows into a publishing decision

First, choose the production page that should answer the observed question. Check whether improving an existing guide is more useful than adding a near duplicate. Our existing ChatGPT visibility and board-reporting guides are relevant starting points.

Second, separate a content decision from an indexing decision. Staging rows warrant a review of the staging host's indexing controls. They do not show that a different headline or a longer article would fix the hostname problem.

Third, retain the baseline: property, hostname, grouping, dates, timezone, retrieval date, requested row limit and known omissions. Compare the same scopes when another export is available. Keep any URL or measurement changes visible before claiming a result.

Choose the article because it answers a real question. This export supports testing a practical measurement guide and a reporting worksheet. It does not forecast traffic growth or prove that either article will receive AI citations.

Evidence and method

Source: DoneHat's saved sc-domain:donehat.com export. Dates 23 September–20 October 2026 inclusive, 28 days; Search Console dates use America/Los_Angeles. Retrieved 22 October 2026 at 19:49 UTC. Query and page responses were independent; row limit 1000 per request. Returned 42 query rows and 36 page rows, including 16 production and 20 staging page rows.

The older selected page rows and independent query rows remain separate. The later paired rows are reported only where query and page were requested together. The 23 October extraction used the same 23 September–20 October window, final web-search data and a 1,000-row detail limit without pagination. Its dimensionless domain aggregate was 14 clicks and 288 impressions, with direct and integration reads matching.

Production detail output was restricted to reviewed public query keys and public paths; all 13 provider-returned pairs passed that output filter, with no privacy omissions. This does not establish exhaustive query coverage: anonymized queries and provider limits can still omit data. Google impressions are not AI prompt volume, people or conversions. No post-fix ranking or traffic gain is claimed.

Original saved Search Console receipt SHA-256: d9fc8cd9ff960cc4e519953d31b83a703e68e12bd958603f5d4e1ccbd992ad15. This checksum identifies the saved file reviewed, not a complete property-total export.

Aggregate and paired-data retrieval: 23 October 2026 at 21:37 UTC. Read-only extraction SHA-256: 92e0461f86e12944473c4acc232de8042dd906b510bb0ceb3c776a3e150a1002. This is fresh retrieval of historical data, not a new performance window.

Google’s Search Analytics documentation distinguishes property aggregation from page aggregation and warns that query/page detail can omit data. See Google’s API reference and data aggregation guidance in the Sources below.

Sources

• API reference — Google for Developers (accessed 2026-10-23)
• Data aggregation guidance — Google for Developers (accessed 2026-10-23)