What the report found
In the public Duolingo sample report, participants could not verify whether outcome evidence applied to their language, level, speaking goals or paid tier. The report points to duolingo.com/efficacy/studies and recommends an auditable index of research.
This is a finding from the recorded test, not a claim about the current live Duolingo website. The complete sample report and its structured export are linked below.
Before: a concern without a clear implementation
‘Make speaking and writing outcomes easy to find by language and skill’ identifies a direction, but leaves the editor to decide what page to change, what to add and when the work is done.
An implementation-ready task connects that concern to a specific information structure. It should help a learner answer: does this study support the outcome I want, for the course I use?
After: a proposed studies index
On the efficacy studies page, place a ‘Find evidence for your course’ section before the research listings. Organize the studies by course language and skill, with a clear way to distinguish speaking, writing, reading and listening.
For every study, show
Course language and level, learner population, measured skill, sample size, attrition, study duration, findings and limitations. Disclose funding and whether the study is independent. Link the original publication.
Separate evidence from inference
Do not use a reading or listening result as proof of speaking proficiency. Where no applicable study is listed, say so instead of implying all courses and skills are covered.
Connect the course page
Beside a course’s outcome claim, link directly to the relevant study entry. State which skill the evidence measures and what remains untested.
What counts as implemented
A learner can identify the evidence for their course and skill without opening every paper. Each entry links to its source, exposes limitations and distinguishes measured outcomes from broader marketing claims.
No numerical result should be added without the underlying study. These are proposed content and navigation changes; this example does not represent a Duolingo implementation or customer endorsement.
What the retest should establish
Rerun the original buying requests with compatible profiles. Inspect whether the agent finds the relevant evidence, represents its scope correctly and still raises the original concern. Compare the cited sources as well as the recommendation.
Retest status: not measured for this proposed change. We have not published a matched post-edit result for this example. A completed case study would include the published edit, both dated reports, configuration differences and repeated observations before making any claim of lift.