Start with six questions from visitors’ everyday work
Choose two queries with a known answer page, two requiring comparison and two with no available answer. Use terms from real enquiries where that evidence exists. If it does not, identify the terms as hypotheses rather than measured popular searches.
A fictional training provider might test Introduction A, the topic bookkeeping and the query Sunday course. The last offer may not exist. The test should establish whether the visitor receives an understandable message instead of an unrelated list.
Define what a useful result would mean
Write the expected result before each test. It might be a particular page, two reasonable alternatives or an honest empty result. An old news item mentioning the term is not necessarily a useful first choice.
Review the title, description and any category together. Can someone distinguish the course description from last year’s course announcement? Ambiguity may originate in headings or the content model rather than search ranking alone.
| Test | Expectation | Evidence |
|---|---|---|
| Known page | The right destination can be identified | Chosen result and working link |
| Similar resources | The difference is understandable | Title and description support the choice |
| No answer | An understandable way forward | Query and filters can be changed |
Try alternative words and languages for a reason
Add a common abbreviation or alternative name where there is reason for the audience to use it. Decide whether results should belong to the active language and how other versions are discoverable. An English query leading to an unlabelled page in another language may stop the task.
Typo tolerance, synonyms and searching inside PDFs are separate capabilities to verify. Do not assume they exist because a search field is visible. Record actual support and try an example before describing a capability as a benefit.
Make an empty result understandable
Show which query returned no results and offer a concrete next step: change words, remove a filter or open a relevant topic index. Do not imply an offer exists merely to avoid an empty list.
Show active filters and let them be changed. Retain the query so it does not need unnecessary retyping. Check with a keyboard that search, results and resetting can be reached in a logical order.
Follow the result to the actual page
Open the chosen result in a published preview. Check the link, language and whether the answer is present. A search index may need updating after publication; verify the actual process in the solution in use.
Change a marked test page in a safe environment and check when the update appears in results. Document required steps, ownership and limitations. Do not use private documents or real personal information for the exercise.
Prioritise what prevents the task
Record reproducible problems with the query, expected result, actual match and next step. Separate content gaps from missing search capabilities. More filters cannot create an absent answer page.
Ask another person to try the list without explanation. Fix blocking links and misleading matches first, then rerun the affected queries. Keep the list as a small regression check when content or the search solution changes.