Study & Research Skills
16 gains13 Traces14 problems
Every measured gain and problem on the record for this topic, newest first. Updated .
One square per record, darker for stronger evidence:
0 records
Showing the records in one thread of the summary: Accuracy and checking AI answers. Show every record
No records of this kind in this topic yet.
AI and how students study
Most of what is known about students and AI comes from surveys and small interview studies, with only a few controlled comparisons. TRV-2026-2205TRV-2026-1464TRV-2026-0548 Students in many countries say AI makes hard ideas easier to grasp and speeds up their work, and many also say they worry about leaning on it too much. TRV-2026-1674TRV-2026-2100TRV-2026-1832
A randomized experiment with 117 university students found that those who used ChatGPT improved their essay scores more than students given a human expert, writing tools or no extra support. TRV-2026-0548 A systematic review of 27 studies found ChatGPT supported personalized learning, feedback and coding; it also linked heavy reliance to cognitive offloading and weaker memory retention, without measuring how large either effect is. TRV-2026-1900 In a controlled comparison of 35 foundation level students, weekly Socratic questioning led by AI brought no statistically significant change in critical thinking test scores. TRV-2026-1836 Among 44 English-as-a-foreign-language students in Oman, writing with AI help was associated with 42% fewer grammatical errors than traditional writing. TRV-2026-1842
A check of 500 references generated by chatbots for restorative dentistry found frequent errors in author names, journal titles and DOIs; Copilot and DeepSeek-v3.2 scored worst. TRV-2026-1415 In a survey of 381 Lahore students, 70.9% said ChatGPT's results contained inaccuracies, yet 50.1% agreed they would use generated text without further validation. TRV-2026-1674 Among 454 final-year dental students surveyed, only 20.0% said they always verify outputs and 25.6% reported using AI during exams. TRV-2026-0675 An observational study of 666 people found frequent AI tool use linked to weaker critical thinking, with younger participants more dependent on AI and scoring lower. TRV-2026-0545
Studies without comparison groups, one of 25 foundation students and one of 31 high school students, reported higher scores after AI use or prompt training, so the gains cannot be pinned on the tools. TRV-2026-1835TRV-2026-2007 On K-12 learning, little is known: Stanford reported little evidence either way, while parents and child development experts say chatbots may let children offload thinking. TRV-2026-0087 Many findings rest on what students said rather than what they learned; 300 preservice teachers reported high acceptance of ChatGPT, with no measured learning results. TRV-2026-0443 Formal training looks scarce: 82.6% of 310 health science students had received no formal AI instruction, and only 30.3% of 175 Iraqi nursing students had formal AI training. TRV-2026-2376TRV-2026-2257
Critical thinking is where the evidence points both ways: in a survey of 64 Economics Education students, heavier use of AI learning apps went with higher self-rated critical thinking. TRV-2026-1414 An observational study of 666 people linked frequent use to weaker critical thinking, and a comparison of 35 students found no change in test scores. TRV-2026-0545TRV-2026-1836 In a survey of 40 US students, ChatGPT users reported improved critical thinking, but those who used it passively reported neutral views at the resolution stage. TRV-2026-1312
Students say AI helps with brainstorming and grammar, and a randomized experiment found higher essay scores, though some interviewed students said full AI drafts reduce their independence. TRV-2026-0548TRV-2026-1464TRV-2026-2358TRV-2026-1842
See these 6 recordsThe evidence runs both ways: one survey links AI use to stronger thinking, while an observational study and a review tie heavy reliance to weaker thinking and memory. TRV-2026-1414TRV-2026-0545TRV-2026-1900
See these 11 recordsChatbots invented citation details in a reference check, and surveyed students said they often skip verification. TRV-2026-1415TRV-2026-1674TRV-2026-0675
A Trace is one record with both a gain and a problem, so it counts on both sides: 29 records show a gain and 27 show a problem. 20 context records are in the record list but not counted, because they report no measured gain or problem.
| Evidence | Gain side | Problem side |
|---|---|---|
| HighStrong study design, backed by at least two independent sources | 0 | 0 |
| ModerateA measured study or an official source | 18 | 16 |
| EmergingEarly, small, unreviewed or reported second-hand | 11 | 11 |
By type of evidence: 23 surveys; 10 interview studies; 4 controlled comparisons; 2 observational studies; 1 systematic review or meta-analysis; 1 benchmark or lab test; 1 randomized trial; 1 news report. How evidence is scored
In a survey of 401 Chinese university students, respondents perceived AI and social media use as positively linked to their academic performance and mental well-being, with smart learning amplifying the effect.
300 preservice teachers at one university reported high acceptance of ChatGPT as a tool for self-regulated learning, based on questionnaires, reflections and interviews rather than measured learning results.
In a randomized experiment with 117 university students, those who used ChatGPT improved their essay scores more than students given a human expert, writing tools or no extra support.
The same review linked heavy reliance on ChatGPT to cognitive offloading, weaker memory retention, superficial learning and diminished creativity in students, without measuring how large these effects are.
In the same Omani study, 63% of students did not fully follow task instructions, which the researchers read as limited deep understanding of content.
Many surveyed students had little formal AI training, and low confidence in campus rules and privacy worries were tied to less use. TRV-2026-2376TRV-2026-0428TRV-2026-0362
See these 9 recordsSurveys link perceived usefulness, ease of use and university support to students' intention to use AI. TRV-2026-1324TRV-2026-2198TRV-2026-0408
See these 4 recordsEach record sits in one thread at most. A record with both a gain and a problem counts on both sides. 8 records fit none of these threads.
In a study of 35 foundation-level students, weekly AI-led Socratic questioning brought no statistically significant change in critical thinking test scores, though the students viewed AI positively.
ace