Colquhoun.AI
← All guides

Research Guides

Verify Zero Search Matches Before Recording Them

· Colquhoun.shop editorial

Record “zero matches” only when the research tool has evidence that a search completed and found none. A blank list after the tool received no response body, failed to parse the result, or never ran the request should not enter the audit as a zero. Otherwise, a broken search and a clean result can look identical in your report.

Start with what the response actually says

RFC 9110 defines HTTP 204 as a successful response with no content to send. It does not assign an application-level result count. An endpoint may use 204 for an empty result if its contract explicitly says so. An empty array in a valid 200 response can be evidence of zero if the contract identifies that array as the complete set for the query. An empty response body and an empty JSON array are different cases.

Four fixtures for an empty-state audit

On an isolated copy or with recorded fixtures, hold the query, filters, scope, and page number fixed. Then compare four results and the words the interface uses to describe them:

  1. One known match. Return a valid response with one synthetic result. Verify the interface shows it and the exported count is one.
  2. Documented empty result. Return the service's defined no-match response. Verify the interface clearly says no results for this query, and check whether the count applies to the whole query or only the current page.
  3. No content without a defined count. Return 204 in the fixture. Unless the service contract explicitly maps that response to an empty result, record that the body did not establish a count.
  4. Unavailable or malformed result. Simulate a failed request or a 200 with invalid JSON. Verify that the interface reports the failure or inability to load results instead of silently substituting an empty array.

Worked example: a directory search

Imagine an audit tests a directory's “meeting notes” filter. Page one returns {"items":[],"total":0} under a documented schema, so the worksheet can record zero matches within that directory snapshot. A second test simulates a missing body and the same interface displays “0 results.” That second row is not another zero: it is evidence of misleading fallback copy. If you later test page three of a query with two results, an empty page still does not erase those two earlier matches.

Save one row per case: query and scope | fixture or timestamp | HTTP status | payload state | documented result count | displayed text | export behavior. Use “unknown” for counts that were never established. The worksheet tests the evidence trail, not the directory's completeness or its search engine visibility.

For a related distinction outside API payloads, see checking whether a timing value was missing before recording zero.

This guide describes the built-in Web Reader and Keyword Density behavior verified for this publication. Available data depends on the page response; these tools do not measure visits or search rankings.