Students' Perceptions of Artificial Intelligence Tools for Study Productivity and Learning: An Exploratory Survey Study

Synthesis: This exploratory survey reports how 26 respondents — 23 of them current students — describe their use of generative AI for studying, and the paper's most useful contribution is a distinction its headline numbers hide. A large majority perceived the tools as effective: 92.3% agreed or strongly agreed that AI improved their understanding of subjects and 80.8% that it helped them complete assignments faster. Yet only 38.5% said AI reduced their overall study time while 50.0% said it did not, so faster completion of a single task did not translate into a shorter learning process — a gap the author reads partly as students spending saved time on more questions, revision, and verification. Alongside the positive perceptions, half of respondents reported sometimes or often relying on AI instead of trying to learn independently, and open-ended answers named dependence, reduced critical thinking and independent effort, inaccurate information, and academic dishonesty as the costs. The design is cross-sectional, descriptive, and non-probability, and the paper is explicit that perceived improvement is not measured learning.

Design, Sample, and Instrument

The study used a cross-sectional exploratory survey administered anonymously through Google Forms, recruited by non-probability convenience sampling. The dataset contained 26 completed responses, of which 23 (88.5%) identified as current students and three (11.5%) did not; because the research questions concern students, the student subgroup is the principal population of interest. The age distribution of the full sample was 12 respondents aged 19–25 (46.2%), 12 aged 26 or over (46.2%), and two aged 15–18 (7.7%).

The questionnaire had five sections — basic information, AI usage, productivity, understanding, and perceived benefits and disadvantages — combining categorical and Likert-style closed items with two open-ended questions asking for the main benefit and main disadvantage of AI-assisted studying. Closed items are reported as frequencies and percentages; open-ended responses were grouped into recurring themes. No inferential tests were run, because of the small sample and the absence of probability sampling, and percentages are calculated on the full 26 unless otherwise stated.

Two ethical features are reported with unusual candour and are worth noting as a model of the limitations disclosure this literature often lacks. The study states that it is not represented as having received institutional ethical approval, that two respondents were minors (raising consent, assent, and data-protection requirements that should be verified before further dissemination), and that individual-level data are withheld to reduce Privacy risk. The manuscript also declares its own use of AI: Claude assisted with LaTeX formatting, Overleaf with conversion and compilation, and ChatGPT with drafting and formatting portions of the text, while the research design, survey data, statistical results, and substantive conclusions are the author's own and all AI-assisted text was reviewed and verified. That declaration is itself a small case of disclosure practice.

Reported Use and Frequency

Twenty-three respondents (88.5%) reported using AI tools for studying and three (11.5%) did not. Reported frequency was daily for nine (34.6%), weekly for ten (38.5%), rarely for six (23.1%), and never for one (3.8%) — a distribution in which reported use is routine for most of the sample but not universal. The instrument listed ChatGPT, Google Gemini, Bing AI/Copilot, and "other" alongside a "do not use AI tools" option, though the analysis centres on aggregate frequency rather than tool-by-tool breakdowns.

The paper positions these numbers against systematic-review evidence that is deliberately paired rather than one-sided: reviews of ChatGPT in education document on-demand explanations, feedback, personalized support, and learning opportunities (Albadarin et al., 2024; Lo et al., 2024a) while also flagging inaccurate information, privacy, dishonesty, and overreliance, with Lo et al. noting that evidence on cognitive engagement and critical thinking remains comparatively weak and needs objective measures. The author's framing follows from this: students' perceptions are worth documenting, but perceived improvement should not be equated with objectively measured performance.

Perceived Productivity: Faster Tasks, Unchanged Study Time

The productivity results contain the page's most interesting internal tension. On assignment completion, 14 respondents (53.8%) strongly agreed and seven (26.9%) agreed that AI tools helped them complete assignments faster, four (15.4%) were neutral, and one (3.8%) disagreed — 21 of 26 (80.8%) in agreement.

On overall study time the picture inverts. Asked directly whether AI reduced their study time, ten (38.5%) said yes, 13 (50.0%) said no, and three (11.5%) were unsure. Most respondents experienced AI as making an individual task faster while a majority did not experience the total study process as shorter. The author's interpretation is that completing a task more quickly does not necessarily shorten learning: students may use the time saved to ask further questions, explore additional material, revise outputs, or verify information.

That reading matters beyond this sample because it names a measurement problem in productivity claims more broadly. Efficiency measured at the level of one assignment is not the same quantity as time-on-task across a course, and a saving in the first can coexist with no change, or an increase, in the second. In the vocabulary of the wider literature this is the difference between an efficiency gain on a task and a genuine reduction in work — and it is consistent with behavioural evidence that time on AI-susceptible problems falls while performance on proctored retention items does not necessarily follow (Faster Completion, Less Learning: Generative AI Reduced Study Time on Math Problems and the Knowledge They Build).

Perceived Understanding and Reported Reliance

The strongest positive perception concerned understanding: 17 respondents (65.4%) strongly agreed and seven (26.9%) agreed that AI improved their understanding of subjects, with one neutral and one strong disagreement, for 24 of 26 (92.3%) in agreement. The author notes that this is exactly the outcome self-report can capture and cannot verify, and that the parallel meta-analysis of 69 experimental studies (Deng et al., 2025) found positive effects of ChatGPT interventions on performance and some higher-order-thinking outcomes while recommending objective assessment and adequate sample-size planning.

Reported reliance cuts the other way. Asked whether they rely on AI instead of trying to learn independently, two (7.7%) said often, 11 (42.3%) sometimes, six (23.1%) rarely, and seven (26.9%) never — so 13 respondents (50.0%) reported sometimes or often substituting AI for their own learning. The coexistence of high perceived understanding with half the sample reporting substitution is the ambivalence the paper highlights, and it is the same tension the knowledge base documents at larger scale as over-reliance and as the performance–learning gap: perceived comprehension and durable comprehension are different outcomes, and only one of them is available from a questionnaire.

What Respondents Named as Benefits and Disadvantages

The open-ended answers map closely onto the review literature. Benefits clustered on understanding and explanation (simplifying complex material, explanations pitched at an understandable level), speed and convenience, and information access, including summarising, vocabulary and language support, and personalized or interactive assistance. One respondent reported no benefit. The paper notes the correspondence with previously documented uses — on-demand explanation, immediate feedback, information access, and personalized support.

Disadvantages centred on dependence or overreliance, reduced independent effort and Problem Solving, possible reductions in critical thinking, reading, focus, or memory retention, the need to verify outputs, and academic dishonesty or cheating, with several respondents reporting no disadvantage or being unsure. The author is careful about what this can establish: the survey cannot determine whether the perceived risks actually occurred or affected academic outcomes. The paper's recommendations follow the same division — students should treat AI as a learning aid rather than a substitute for their own reasoning and verify important AI-generated information against reliable academic sources; educators should teach students how to evaluate AI outputs, identify errors, and use AI ethically (an AI literacy and critical thinking agenda rather than a detection one); and future research should combine perceptions with pre/post tests, grades, or controlled learning tasks and with larger, more diverse samples.

Key Findings

1. AI use for studying was widespread in the sample. 23 of 26 respondents (88.5%) reported using AI tools for studying; nine (34.6%) used them daily and ten (38.5%) weekly.
2. Faster assignments were near-consensus. 21 of 26 (80.8%) agreed or strongly agreed that AI helped them complete assignments faster, including 14 (53.8%) who strongly agreed.
3. Faster tasks did not mean shorter study. Only 10 respondents (38.5%) said AI reduced their study time, while 13 (50.0%) said it did not and three (11.5%) were unsure — a dissociation the author attributes to saved time being reinvested in more questions, revision, or verification.
4. Perceived understanding gains were the strongest result. 24 of 26 (92.3%) agreed or strongly agreed that AI improved their understanding of subjects (17, or 65.4%, strongly).
5. Half the sample reported substituting AI for independent learning. Two respondents (7.7%) said they often rely on AI instead of trying to learn themselves and 11 (42.3%) said sometimes, against six (23.1%) rarely and seven (26.9%) never.
6. Benefits and risks were reported side by side. Open-ended answers praised explanations, speed, and information access while naming dependence, reduced independent effort and critical thinking, verification burden, and academic dishonesty.
7. The study makes no causal or generalizable claim. The design is cross-sectional and descriptive with a non-probability sample of 26, a non-validated instrument, and self-reported outcomes only; perceived improvement is explicitly not treated as evidence of measured learning.

Limitations and What the Findings Can Support

The paper states its limitations without softening them: the sample was small (N = 26), limiting precision and generalizability; recruitment was not random, so selection bias is possible; outcomes are self-reported rather than objective measures such as examination scores, assignment grades, study-time logs, or standardized assessments; the cross-sectional design cannot establish that AI use caused any change in learning or productivity; two respondents were aged 15–18, so consent, assent, and data-protection requirements should be verified before public dissemination; and the questionnaire was a small exploratory instrument, not a validated psychometric scale.

Read within those bounds, the contribution is a documented perception profile with one genuinely informative internal contrast — task speed rising while total study time does not — reported by an author who distinguishes throughout between perceived and measured outcomes. The knowledge base's larger-scale studies supply what this one cannot: whether assisted performance transfers to unassisted performance (From Enhancement to Over-Reliance: A Mixed-Method Study of Generative AI and Sustainable Learning Performance), how reliance calibrates against objective competence (Beyond checking: verification quality, reliance calibration, and learning in generative AI-assisted higher education), and how students' perceptions compare with faculty and institutional accounts of the same practices (Beyond the Hype: How Higher Education Stakeholders View the Benefits and Concerns of Generative AI for Teaching, Research, and Administration, Exploring AI perceptions in education: unveiling the role of student and teacher motivation and self-efficacy).

Connected Concepts

• Student Experience — self-reported study practices, perceived productivity, and perceived understanding
• Self-Report Measures — the study's core measurement constraint: perceptions cannot establish behaviour or learning
• Generative AI — how a small convenience sample uses and evaluates generative tools for studying
• Higher Education — the setting of the survey and of the systematic reviews it draws on
• Cognitive Offloading — half of respondents reported relying on AI instead of learning independently
• Critical Thinking — reduced critical thinking named among the reported disadvantages
• Academic Integrity — dishonesty and cheating appear in respondents' own account of the costs
• Learning Gains — the paper's explicit refusal to equate perceived improvement with measured learning
• AI Literacy — evaluating outputs, identifying errors, and verifying information as the recommended response
• Privacy — individual-level data withheld, and minors in the sample raising consent questions

Connected Articles

• Exploring AI perceptions in education: unveiling the role of student and teacher motivation and self-efficacy — Student and teacher perceptions of AI and their relation to motivation
• Beyond the Hype: How Higher Education Stakeholders View the Benefits and Concerns of Generative AI for Teaching, Research, and Administration — Stakeholder perceptions of GenAI in higher ed (Humble & Mozelius 2026)
• How Does Students' Perception of ChatGPT Shape Online Learning Engagement and Performance? — ChatGPT perceptions, online learning, and engagement
• Understanding student perspectives on generative AI chatbots: a human-centred mixed-methods study in higher education — Student perspectives on GenAI chatbots in higher education
• "It is a temptation to get it to do the work…" Student Experiences of Navigating the Generative AI Landscape in UK Higher Education: A Cross-Institutional Survey with International Comparison — GenAI Student Experiences: UK HE Survey
• The impact of risk perception on perceived benefits of generative artificial intelligence users—an empirical study based on university students — Risk perception of GenAI and perceived benefits
• From Enhancement to Over-Reliance: A Mixed-Method Study of Generative AI and Sustainable Learning Performance — From Enhancement to Over-Reliance: A Mixed-Method Study of Generative AI and Sustainable Learning Performance
• Is using artificial intelligence tools for academic work cheating? Student perceptions, ethics, and the impact — Student cheating behaviour with AI tools in academic work

Citation

Oladosu, A.-A. (2026). Students' Perceptions of Artificial Intelligence Tools for Study Productivity and Learning: An Exploratory Survey Study. OSF preprint.