<?xml version="1.0" encoding="UTF-8"?>
<article article-type="research-article" xmlns:xlink="http://www.w3.org/1999/xlink"><front><journal-meta><journal-title-group><journal-title>Annals of Open Science</journal-title></journal-title-group></journal-meta><article-meta><article-id pub-id-type="doi">10.0000/aos.2026.0001</article-id><title-group><article-title>Generative-AI use and study habits of college students in fully remote, hybrid and fully onsite courses: a simulated comparative study</article-title></title-group><contrib-group><contrib contrib-type="author" corresp="yes"><name><string-name>Maya Okafor</string-name></name><aff>Department of Education, Example University</aff></contrib><contrib contrib-type="author"><name><string-name>Daniel Reyes</string-name></name><aff>Center for Learning Analytics, Example University</aff></contrib></contrib-group><pub-date><day>16</day><month>11</month><year>2026</year></pub-date><permissions><license license-type="open-access" xlink:href="https://creativecommons.org/licenses/by/4.0/"><license-p>CC BY 4.0</license-p></license></permissions><abstract><p>Simulated cross-sectional survey of 612 undergraduates comparing weekly generative-AI use and study habits across fully remote, hybrid and fully onsite course delivery, and their relation to final grade.</p></abstract></article-meta></front><body><sec id="introduction">
  <title>Introduction</title>
  <p>Universities moved large parts of teaching online during the pandemic and have kept many hybrid formats since [<xref alt="1" rid="ref-1" ref-type="bibr">1</xref>]. At the same time, generative AI assistants became freely available to students [<xref alt="2" rid="ref-2" ref-type="bibr">2</xref>]. Remote learners work with less external structure, so they may lean more on AI tools for explanation and drafting [<xref alt="3" rid="ref-3" ref-type="bibr">3</xref>]. Whether that reliance helps or harms learning is an open question [<xref alt="4" rid="ref-4" ref-type="bibr">4</xref>,<xref alt="5" rid="ref-5" ref-type="bibr">5</xref>]. We ask three questions: how does AI use differ across remote, hybrid and onsite delivery; how do study habits differ; and how are both related to course outcomes?</p>
</sec>
<sec id="methods">
  <title>Methods</title>
  <sec id="design-and-participants">
    <title>Design and participants</title>
    <p>We simulated a cross-sectional survey in one semester. Participants were 612 undergraduates taking the same three introductory courses delivered in three formats. Group sizes are given in Table <xref alt="1" rid="tbl-1">1</xref>. Students self-reported demographics, weekly hours of AI tool use, planning habit (a mean of five items on a 1–5 scale), study sessions per week and procrastination (1–5). Final course grade was taken from the registry.</p>
  </sec>
  <sec id="statistical-analysis">
    <title>Statistical analysis</title>
    <p>Group differences were tested with one-way analysis of variance and described with eta-squared. Pairwise differences use Welch intervals. Grade was modelled by ordinary least squares with delivery mode, AI use, AI use squared and planning habit; onsite delivery was the reference. All data were generated with a fixed random seed, and the analysis script is available from the corresponding author.</p>
  </sec>
</sec>
<sec id="results">
  <title>Results</title>
  <sec id="participants">
    <title>Participants</title>
    <p>Table <xref alt="1" rid="tbl-1">1</xref> summarises the three groups.</p>
    <boxed-text id="tbl-1">
      <table-wrap>
        <table>
          <thead>
            <tr>
              <th>Delivery mode</th>
              <th>n</th>
              <th>AI use, h/week</th>
              <th>Planning (1–5)</th>
              <th>Sessions/week</th>
              <th>Final grade, %</th>
            </tr>
          </thead>
          <tbody>
            <tr>
              <td>Fully remote</td>
              <td>204</td>
              <td>4.2 (2.2)</td>
              <td>3.37 (0.66)</td>
              <td>5.2 (1.4)</td>
              <td>71.1 (7.6)</td>
            </tr>
            <tr>
              <td>Hybrid</td>
              <td>211</td>
              <td>2.9 (1.7)</td>
              <td>3.56 (0.65)</td>
              <td>6.3 (1.4)</td>
              <td>75.2 (7.2)</td>
            </tr>
            <tr>
              <td>Fully onsite</td>
              <td>197</td>
              <td>2.1 (1.3)</td>
              <td>3.57 (0.64)</td>
              <td>6.0 (1.5)</td>
              <td>74.0 (7.6)</td>
            </tr>
          </tbody>
        </table>
      </table-wrap>
    </boxed-text>
    <p>Table <xref alt="1" rid="tbl-1">1</xref>. Participants and outcomes by delivery mode. Values are mean (standard deviation).</p>
  </sec>
  <sec id="ai-use-by-delivery-mode">
    <title>AI use by delivery mode</title>
    <p>Weekly AI use differed by delivery mode (F = 67.0, eta-squared = 0.18; Figure <xref alt="1" rid="fig-1">1</xref>). Remote students used AI 2.0 h/week more than onsite students (95% CI 1.7 to 2.4).</p>
    <graphic mimetype="image" mime-subtype="" xlink:href="fig1">
      <alt-text>Figure 1</alt-text>
    </graphic>
  </sec>
  <sec id="study-habits">
    <title>Study habits</title>
    <p>Planning habit differed by delivery mode (F = 6.1, eta-squared = 0.02), with remote students lowest and hybrid and onsite students similar. Hybrid students reported the most study sessions per week, and remote students reported the most procrastination (Figure <xref alt="2" rid="fig-2">2</xref>).</p>
    <graphic mimetype="image" mime-subtype="" xlink:href="fig2">
      <alt-text>Figure 2</alt-text>
    </graphic>
  </sec>
  <sec id="ai-use-and-course-grade">
    <title>AI use and course grade</title>
    <p>Grade rose slightly with AI use in the raw data, and the fitted curves for the three modes were similar in shape but offset in height (Figure <xref alt="3" rid="fig-3">3</xref>). After adjustment the association was not distinguishable from zero.</p>
    <graphic mimetype="image" mime-subtype="" xlink:href="fig3">
      <alt-text>Figure 3</alt-text>
    </graphic>
    <p>Table <xref alt="2" rid="tbl-2">2</xref> and Figure <xref alt="4" rid="fig-4">4</xref> give the adjusted model. Relative to onsite delivery, remote students scored about three points lower, and the hybrid difference was small. The coefficients for AI use and AI use squared both had confidence intervals that included zero.</p>
    <boxed-text id="tbl-2">
      <table-wrap>
        <table>
          <thead>
            <tr>
              <th>Term</th>
              <th>Coefficient</th>
              <th>95% CI</th>
            </tr>
          </thead>
          <tbody>
            <tr>
              <td>Intercept (onsite, reference)</td>
              <td>62.71</td>
              <td>59.21 to 66.21</td>
            </tr>
            <tr>
              <td>Fully remote</td>
              <td>-3.02</td>
              <td>-4.61 to -1.44</td>
            </tr>
            <tr>
              <td>Hybrid</td>
              <td>0.90</td>
              <td>-0.53 to 2.33</td>
            </tr>
            <tr>
              <td>AI use (h/week)</td>
              <td>0.42</td>
              <td>-0.41 to 1.25</td>
            </tr>
            <tr>
              <td>AI use squared</td>
              <td>-0.01</td>
              <td>-0.10 to 0.07</td>
            </tr>
            <tr>
              <td>Planning habit (1–5)</td>
              <td>2.94</td>
              <td>2.06 to 3.83</td>
            </tr>
          </tbody>
        </table>
      </table-wrap>
    </boxed-text>
    <p>Table <xref alt="2" rid="tbl-2">2</xref>. Linear model of final grade (R-squared = 0.12).</p>
    <graphic mimetype="image" mime-subtype="" xlink:href="fig4">
      <alt-text>Figure 4</alt-text>
    </graphic>
  </sec>
</sec>
<sec id="discussion">
  <title>Discussion</title>
  <p>In this simulated sample, remote delivery went with heavier AI use, weaker planning and more procrastination, while hybrid students studied in the most sessions. The lack of a clear adjusted association between AI use and grade is consistent with the view that the benefit of AI depends on how it is used alongside regular study rather than on how much it is used [<xref alt="4" rid="ref-4" ref-type="bibr">4</xref>,<xref alt="5" rid="ref-5" ref-type="bibr">5</xref>]. Because the study is cross-sectional and the data are simulated, no causal claim is made. A real study would need randomisation or longitudinal follow-up, validated habit scales and pre-registered analyses [<xref alt="6" rid="ref-6" ref-type="bibr">6</xref>].</p>
  <sec id="limitations">
    <title>Limitations</title>
    <p>The data are simulated, so the effects reflect the assumptions used to generate them. Self-reported AI use is subject to recall error, and students were not randomised to delivery mode.</p>
  </sec>
</sec>
<sec id="data-availability">
  <title>Data availability</title>
  <p>All data are simulated. The generating script and a CSV of the simulated dataset are available from the corresponding author on request.</p>
</sec>
<sec id="ethics-statement">
  <title>Ethics statement</title>
  <p>No human participants were involved. All records are synthetic.</p>
</sec>
</body><back><ref-list><ref id="r1"><label>1</label><mixed-citation>Bozkurt A, Sharma RC. Emergency remote teaching in a time of global crisis due to coronavirus pandemic. Asian Journal of Distance Education. 2020;15(1):i-vi.</mixed-citation></ref><ref id="r2"><label>2</label><mixed-citation>Kasneci E, Sessler K, Kuchemann S, et al. ChatGPT for good? On opportunities and challenges of large language models for education. Learning and Individual Differences. 2023;103:102274. doi:10.1016/j.lindif.2023.102274</mixed-citation></ref><ref id="r3"><label>3</label><mixed-citation>Broadbent J, Poon WL. Self-regulated learning strategies and academic achievement in online higher education learning environments: a systematic review. The Internet and Higher Education. 2015;27:1-13. doi:10.1016/j.iheduc.2015.04.007</mixed-citation></ref><ref id="r4"><label>4</label><mixed-citation>Zawacki-Richter O, Marin VI, Bond M, Gouverneur F. Systematic review of research on artificial intelligence applications in higher education. International Journal of Educational Technology in Higher Education. 2019;16:39. doi:10.1186/s41239-019-0171-0</mixed-citation></ref><ref id="r5"><label>5</label><mixed-citation>Means B, Toyama Y, Murphy R, Bakia M, Jones K. Evaluation of evidence-based practices in online learning: a meta-analysis and review of online learning studies. US Department of Education; 2010.</mixed-citation></ref><ref id="r6"><label>6</label><mixed-citation>Nosek BA, Ebersole CR, DeHaven AC, Mellor DT. The preregistration revolution. Proceedings of the National Academy of Sciences. 2018;115(11):2600-2606. doi:10.1073/pnas.1708274114</mixed-citation></ref></ref-list></back></article>