ChatGPT users differentiated information less (never by credibility) and rarely verified information compared to Google users¶

Claim · All claims
Evidence · 1 study (2 entries) · causal r1–r2 · q2
Currency · newest study 2026 · checked against its sources 2026-10-20 (automated, 2 of 2 entries) · not reviewed by a person

Subclaims¶

q2 i? ChatGPT participants showed fewer differentiation instances, with zero entries filtering by credibility versus 30.71% for Google, and verified information via another source in only 3.06% of entries versus 19.29% for Google (verification via prior knowledge: 4.08% versus 2.86%). → Shravika Mittal 2026 q2 i? ChatGPT participants checked accuracy in only seven instances (three via another source, four via prior knowledge), versus more frequent verification in the Google condition (31 instances). → Shravika Mittal 2026 (2)

Evidence¶

Shravika Mittal 2026¶

Shravika Mittal, Su Lin Blodgett, Q. Vera Liao. (2026). Learning by Chatting? Investigating the Impact of Generative AI on Information Seeking and Learning. https://arxiv.org/abs/2606.11669 doi:10.48550/arxiv.2606.11669

q2 · i? · causal · r1

Thematic analysis of diary entries using the Table 1 codebook: credibility-based differentiation appeared in 0% of ChatGPT entries versus 30.71% of Google entries; type-of-need differentiation 0% vs 10.00%. The article attributes this to offloading information selection to ChatGPT.

"Participants in the ChatGPT condition reported fewer instances of differentiating information, i.e., assessing (and filtering) information based on specific criteria (e.g., personal relevance, credibility, type of information need) and their indicators"

Shravika Mittal 2026 (2)¶

Shravika Mittal, Su Lin Blodgett, Q. Vera Liao. (2026). Learning by Chatting? Investigating the Impact of Generative AI on Information Seeking and Learning. https://arxiv.org/abs/2606.11669

q2 · i? · causal · r2

Diary-entry analysis of the verifying stage: three ChatGPT instances via another source, four via prior knowledge, plus three expressions of intent to verify without action; "participants in the Google condition verified information more frequently (31instances)."

"in the verifying stage (after information extraction), there were only seven instances where ChatGPT participants checked the accuracy of the information."

Studies bearing on this claim¶

Do learners check information from a generative AI the way they check sources found by search? This table gathers the studies behind this claim and the 5 narrower claims that bear on it: 6 distinct studies (2 qualitative, 1 causal, 1 design, 1 theoretical, 1 review), q1–q2, 0 of 6 with an effect size coded. One row per study, codes as each entry records them; no effect is averaged and no verdict is drawn. Built by scripts/synthesize_claim.py from eval/synthesis/; a row that looks wrong is fixed on the claim it comes from.

Discussion¶

Related Claims¶

• ChatGPT biased information access toward solution-oriented artifacts over principle-oriented information — related
• ChatGPT participants experienced meta-cognitive load and frictions, with slightly more unsatisfied endings than Google participants — related
• ChatGPT users began information seeking with narrow, specified queries far more often than Google users — related
• ChatGPT users were significantly less likely to correctly compare and evaluate animal and plant protein sources — related
• ChatGPT users showed a non-significant trend toward worse overall informal learning outcomes than Google Search users over 8 days — related
• ChatGPT users were significantly less likely to correctly answer a critical-learning question requiring interpretation of a nutrition label — related
• Students using ChatGPT 3.5 experience lower extraneous cognitive load than students using Google during a web research task — related
• Students using Google produce higher-quality justifications in their final recommendations than students using ChatGPT 3.5 — related
• Of 178 reference sources produced by AI dialogue systems, 69 lacked a DOI and 28 could not be found via Google search and had no existing DOI — related
• Google Trends data show ChatGPT search interest surged globally within weeks of its November 2022 public release, with plagiarism among the top related queries — related
• Recommendation homogeneity does not differ between students using ChatGPT 3.5 and students using Google — related
• Automation bias toward fluent LLM output surrenders learners' epistemic agency — related
• Conversational synthesis can hide source plurality behind one voice, weakening verification relative to search — related
• Novice AI users showed a Trust-but-Can't-Verify pattern: accepting unfamiliar constructs they could not assess, then symptom-level debugging — related