Strandline Research No. 05: long conversational queries in Google Search Console Method notes: data pull, classification rules and aggregation. Data pulled on 23 September 2026. SOURCE - Google Search Console API v1, searchanalytics.query, read with one Search Console account that holds the properties of one portfolio of websites. - Search type: web. Data state: final (fresh, unfinished data left out). - Dates: 2026-06-24 to 2026-09-21 inclusive, 90 days. On 2026-09-23, 2026-09-21 was the last date with final data. PROPERTIES AND SITES - The account listed 46 properties and had data access to 43 of them. The other 3 were listed without data access. - One property with data access belongs to another company. It is left out before any analysis, and no figure, table or file of this report uses its data. - Included: every other property with at least 1,000 impressions in the 90 days, summed from its daily rows. 27 properties. - Excluded: the 15 properties with data access and fewer than 1,000 impressions. - One property is one site: 21 domain properties (all subdomains) and 6 URL-prefix properties. - 6 sites first appeared in search results after 2026-06-24, the last of them on 2026-07-18. They count from that date. - Site types (number of sites): Free online tools and templates (7) sites built around one free tool, such as a calculator, a diagram maker or a document template Single-topic guide sites (11) sites built to answer one search question in depth Directories and databases (4) searchable directories, databases and a job board B2B services (2) company sites for services sold to businesses Content sites and blogs (3) article sites and a personal site PULLS (one set per included site) 1. dimensions [date] daily totals: total impressions and clicks 2. dimensions [query] 90-day query rows: classes, site types, sites, word counts, positions 3. dimensions [date, query] daily query rows, summed into Monday-to-Sunday weeks 4. dimensions [query, device] query rows by device (DESKTOP, MOBILE, TABLET) - rowLimit 25,000, startRow paging until a page came back with fewer than 25,000 rows. The largest pull, [date, query] for one site, returned 25,038 rows in two pages. No other pull needed a second page. - Rows grouped by query, date or device are aggregated by property (two results from the same site in one search count as one impression), so the query rows and the daily totals count impressions the same way. - Checks: the [date, query] rows and the [query, device] rows sum to the same 358,745 impressions as the [query] rows, class by class. VISIBLE IMPRESSIONS - Search Console leaves rare queries out of query rows ("we might not track some queries that are made a very small number of times or those that contain personal or sensitive information", support.google.com/webmasters/answer/96568). - Visible impressions: the sum of impressions over the [query] rows. Share visible: visible / total impressions. - Every class figure uses visible rows only. Totals: 641,159 impressions, 358,745 visible (56.0%). CLASSIFICATION Applied to the query text as Search Console returns it (lower case), after trimming spaces at both ends. Words: the query split on white space. 1. Long conversational, if any of: a. 9 or more words b. the query ends in "?" c. the query contains one of these 16 prompt phrases (plain substring match): "answer the following" "the following question" "user question" "question:" "naturally and concisely" "if relevant" "include specific" "you are a " "you are an " "act as a " "act as an " "respond with" "respond in " "in your answer" "in your response" "based on the following" 2. Question, if not long conversational and the first word is exactly one of: who what where when why how which can does is are should 3. Keyword: everything else. In this data, every query with a prompt phrase also had 9 or more words or a closing "?" (14 queries), so rule (c) added no query of its own. METRICS - Queries: distinct query texts per site, summed across sites (a query seen on two sites counts twice). - Impressions, clicks: sums of row values. - CTR: clicks / impressions. - Average position: impression-weighted mean of the rows' average position. Search Console defines a row's position as the average position of the topmost result from the site. Lower is closer to the top. - Impressions at position 10 or better: impressions from query rows whose average position is 10.0 or lower. - Weeks: Monday to Sunday, labeled by the Monday. weekly.csv and the weekly chart hold 13 weeks: the partial week of 2026-06-22 (only 2026-06-24 to 2026-06-28 are in the window, 5 days) and 12 full weeks, 2026-06-29 to 2026-09-20. 2026-09-21 counts in the 90-day figures only. - Periods: first four full weeks (2026-06-29 to 2026-07-26) and last four (2026-08-24 to 2026-09-20), for all sites, for the 24 sites left after removing the 3 with the most long conversational impressions, and for the 21 sites with visible impressions in each of the 12 full weeks. The partial week is not used here. - Within-site comparisons: sites with at least 30, 50 or 100 visible impressions in both classes being compared. Position gap: keyword average position minus the other class's. Desktop gap: long conversational desktop share minus keyword desktop share, in percentage points. The report text uses the 50-impression cut-off. - Word counts: 1 to 8 words one by one, then 9 to 10, 11 to 14 and 15 or more. PUBLISHED TABLES AND PRIVACY - Property names, domains and query text are not published. Tables are aggregates by class, site type, week or word count, plus sites.csv, which labels sites S01 to S27 (ordered by type, then by share visible), gives total impressions only as a band (1,000 to 9,999; 10,000 to 49,999; 50,000 or more), and shows positions and desktop shares only where both classes have at least 50 impressions. - Percentages and positions are rounded to one decimal from unrounded values.