How We Test Search Tools
A fair search comparison starts with a specific job. We do not run one random query, declare a winner and turn it into a ranking.
Start with the job
Academic search, reverse-image search, code search and everyday web search are different jobs. We choose a task first, then compare tools that are actually meant to do that task.
Use the same questions
When we compare tools, we use the same query set and the same rules for each one. We record the date because results and product features can change.
What we can compare
- Whether the tool found the expected item or source.
- Whether the result gave enough context to judge it.
- Whether sources or destination pages were visible.
- How much work it took to get a useful result.
- Important limits, paywalls or account requirements.
What we do not pretend to prove
A small comparison cannot prove that one search engine is universally “more accurate.” It only tells you what happened on the stated tasks, on the stated date, under the stated rules.
The query set, date, rules and limitations should all be visible. If they are not, treat the result as an editorial impression rather than a benchmark.