Colquhoun.AI
← All guides

Research Guides

Separate Suggested Text From the Research Query Submitted

· Colquhoun.shop editorial

A research search box can show a useful suggestion and still send the wrong query. The important distinction is whether the suggestion was displayed, highlighted, accepted or actually submitted. Test those moments separately instead of assuming every visible highlight means the reader made a choice.

This worksheet is for editable search inputs that suggest queries. It does not certify accessibility or describe a feature of Colquhoun.shop.

1. Identify the intended interaction

W3C's ARIA Authoring Practices combobox guide distinguishes manual and automatic list autocomplete. With manual selection, the typed value remains unless the reader chooses a suggestion. In the automatic variant described there, the first suggestion is highlighted and can become the value when the field loses focus. Neither variant is a reason to leave the interaction undocumented.

Also confirm whether the field accepts arbitrary queries or only predefined values. A site search that accepts arbitrary text is not the same as a report picker that requires an existing record. Use a different expectation for each.

2. Use a fixture where the values visibly differ

Imagine an editorial library search configured for manual selection. Typing crawl budget offers crawl budget checklist. A test that types the full suggestion cannot detect whether the field silently substituted it. Starting with the shorter phrase makes an accidental replacement visible.

Run three separate journeys from a clean field. First, type the shorter phrase and move focus away without selecting anything. Second, type it and dismiss the open list using the intended keyboard behavior. Third, deliberately select the longer suggestion. Inspect the field after each journey before doing anything that submits the search.

For this fictional manual-selection design, the first two journeys retain the shorter phrase and the third adopts the longer one. If a product intentionally follows the documented automatic variant, leaving the field may instead adopt the highlighted option. Judge what actually happens against the selected contract, not against a screenshot of the menu.

3. Record what the search actually received

The accepted text and the submitted query can disagree. For example, the visible field can still show the shorter phrase while the results heading or request uses the longer one. Inspect the result context or the relevant request in a test environment to distinguish a display bug from a query substitution.

4. Sample the confusing boundaries

Include no suggestions, one suggestion and several options. Try a case where suggestions arrive after the tester has moved on or edited the phrase again. For each scenario, verify whether the popup remains appropriate to the current input and whether stale options affect the accepted or submitted value.

Check keyboard operation independently from pointer selection. Verify that the interaction does not unexpectedly discard the draft query. Record assistive technology observations only for combinations you actually tested; an accessible name in markup is not a full interaction result.

5. Close with a reproducible result

A clear acceptance note reads: “In the tested manual-selection field, typing crawl budget and leaving the field retained that phrase. Explicitly selecting crawl budget checklist changed the value, and the results request used that longer query.” If dismissal, late suggestions or mobile interaction were not tested, say so.

This evidence explains why the search behaved as expected in the sampled journeys. It does not demonstrate search ranking, indexing or the quality of the returned sources. Those checks begin after confirming that the site received the query the reader meant to send.

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.