CoChat
Sign Up
Cited by 341Artificial Intelligence in Healthcare and EducationHealth, Technology, Consumer BehaviorTechnology Adoption and User Behaviour

User Intentions to Use ChatGPT for Self-Diagnosis and Health-Related Purposes: Cross-sectional Survey Study

Open access
Shahsavar, Yeganeh, Choudhury, Avishek·2023·JMIR Human Factors
DOI 10.2196/47564PMC PMC10233444

Published version of a preprint

Open this paper in CoChat — free accountOpen PDF

Abstract

Background: With the rapid advancement of artificial intelligence (AI) technologies, AI-powered chatbots, such as Chat Generative Pretrained Transformer (ChatGPT), have emerged as potential tools for various applications, including health care. However, ChatGPT is not specifically designed for health care purposes, and its use for self-diagnosis raises concerns regarding its adoption’s potential risks and benefits. Users are increasingly inclined to use ChatGPT for self-diagnosis, necessitating a deeper understanding of the factors driving this trend. Objective: This study aims to investigate the factors influencing users’ perception of decision-making processes and intentions to use ChatGPT for self-diagnosis and to explore the implications of these findings for the safe and effective integration of AI chatbots in health care. Methods: A cross-sectional survey design was used, and data were collected from 607 participants. The relationships between performance expectancy, risk-reward appraisal, decision-making, and intention to use ChatGPT for self-diagnosis were analyzed using partial least squares structural equation modeling (PLS-SEM). Results: Most respondents were willing to use ChatGPT for self-diagnosis (n=476, 78.4%). The model demonstrated satisfactory explanatory power, accounting for 52.4% of the variance in decision-making and 38.1% in the intent to use ChatGPT for self-diagnosis. The results supported all 3 hypotheses: The higher performance expectancy of ChatGPT (β=.547, 95% CI 0.474-0.620) and positive risk-reward appraisals (β=.245, 95% CI 0.161-0.325) were positively associated with the improved perception of decision-making outcomes among users, and enhanced perception of decision-making processes involving ChatGPT positively impacted users’ intentions to use the technology for self-diagnosis (β=.565, 95% CI 0.498-0.628). Conclusions: Our research investigated factors influencing users’ intentions to use ChatGPT for self-diagnosis and health-related purposes. Even though the technology is not specifically designed for health care, people are inclined to use ChatGPT in health care contexts. Instead of solely focusing on discouraging its use for health care purposes, we advocate for improving the technology and adapting it for suitable health care applications. Our study highlights the importance of collaboration among AI developers, health care providers, and policy makers in ensuring AI chatbots’ safe and responsible use in health care. By understanding users’ expectations and decision-making processes, we can develop AI chatbots, such as ChatGPT, that are tailored to human needs, providing reliable and verified health information sources. This approach not only enhances health care accessibility but also improves health literacy and awareness. As the field of AI chatbots in health care continues to evolve, future research should explore the long-term effects of using AI chatbots for self-diagnosis and investigate their potential integration with other digital health interventions to optimize patient care and outcomes. In doing so, we can ensure that AI chatbots, including ChatGPT, are designed and implemented to safeguard users’ well-being and support positive health outcomes in health care settings.

AI snapshot

Written from the paper’s full text.

TL;DR

In a cross-sectional survey of 607 ChatGPT users, 78.4% were willing to use ChatGPT for self-diagnosis. PLS-SEM supported a model in which performance expectancy and risk-reward appraisal were positively associated with decision-making, which in turn predicted intention to use ChatGPT for self-diagnosis.

Key findings

  • Most respondents were willing to use ChatGPT for self-diagnosis (n=476, 78.4%).
  • The model explained 52.4% of the variance in decision-making and 38.1% of the variance in intent to use ChatGPT for self-diagnosis (adjusted R² 52.2% and 37.9%, respectively).
  • Performance expectancy (β=.547, 95% CI 0.474-0.620) and risk-reward appraisal (β=.245, 95% CI 0.161-0.325) were positively associated with decision-making; decision-making was positively associated with intent to use (β=.565, 95% CI 0.498-0.628).
  • Indirect effects were significant for performance expectancy → decision-making → intent to use (β=.309, 95% CI 0.255-0.366) and risk-reward appraisal → decision-making → intent to use (β=.138, 95% CI 0.087-0.191); all 3 hypotheses were supported.

Methods

  • Cross-sectional online survey conducted in February 2023 via Centiment; 607 participants who used ChatGPT at least once per month were recruited, with IRB approval and informed consent.
  • Measures included performance expectancy (4 items adapted from UTAUT), decision-making (2 items), risk-reward appraisal (1 item), and intent to use ChatGPT for self-diagnosis (1 item), all assessed on a forced 4-point Likert scale.
  • Analysis used descriptive statistics and partial least squares structural equation modeling (PLS-SEM), assessing the measurement model for reliability/validity and then the structural model for path coefficients, significance, and R².

Limitations

  • The paper states that it did not control for potential confounders such as age, medical condition, health literacy, previous experience with comparable technologies, or demographic characteristics.
  • The paper states that the cross-sectional design limits causal inference and examination of evolving user interactions, and that self-reported measures may introduce social desirability, recall, or imprecise-reporting biases.
  • The paper states that focusing on ChatGPT, which is not specifically intended for health care, may limit applicability to other AI chatbots designed for health care; evident from the text, the sample was recruited through Centiment and restricted to monthly ChatGPT users, and risk-reward appraisal and intent to use were measured with single items.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.2196/47564

Cites

42 of 42 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Claims the paper makes

0 of 16 cited sources checked · 0 supported · 0 flagged

  • Findingown data

    Higher performance expectancy of ChatGPT was positively associated with improved perception of decision-making outcomes among users (β=.547, 95% CI 0.474-0.620).

    “The higher performance expectancy of ChatGPT (β=.547, 95% CI 0.474-0.620) and positive risk-reward appraisals (β=.245, 95% CI 0.161-0.325) were positively associated with the improved perception of decision-making outcomes among users”
    Quote in the paperSubstantial claim
  • Findingown data

    Positive risk-reward appraisals were positively associated with improved perception of decision-making outcomes among users (β=.245, 95% CI 0.161-0.325).

    “The higher performance expectancy of ChatGPT (β=.547, 95% CI 0.474-0.620) and positive risk-reward appraisals (β=.245, 95% CI 0.161-0.325) were positively associated with the improved perception of decision-making outcomes among users”
    Quote in the paperSubstantial claim
  • Findingown data

    Enhanced perception of decision-making processes involving ChatGPT positively impacted users' intentions to use the technology for self-diagnosis (β=.565, 95% CI 0.498-0.628).

    “and enhanced perception of decision-making processes involving ChatGPT positively impacted users’ intentions to use the technology for self-diagnosis (β=.565, 95% CI 0.498-0.628).”
    Quote in the paperSubstantial claim
  • Findingown data

    Most respondents were willing to use ChatGPT for self-diagnosis (n=476, 78.4%).

    “Most respondents were willing to use ChatGPT for self-diagnosis (n=476, 78.4%).”
    • Also stated in: § Participant Details · plain
      “Most respondents were willing to use ChatGPT for self-diagnosis (n=476, 78.4%).”
    Quote in the paperNot a substantial claim
  • Findingown data

    Risk-reward appraisal has a positive impact on intent to use via decision-making (β=.138, t=5.191).

    “Likewise, risk-reward appraisal demonstrated a meaningful positive impact on the intent to use via decision-making (β=.138, t=5.191).”
    Quote in the paperSubstantial claim
  • May combine two claims. 15
  • Classified as probably not a substantial claim. 25

  • May combine two claims.

    Findingown data

    The model accounted for 52.4% of the variance in decision-making and 38.1% of the variance in intent to use ChatGPT for self-diagnosis.

    “The model demonstrated satisfactory explanatory power, accounting for 52.4% of the variance in decision-making and 38.1% in the intent to use ChatGPT for self-diagnosis.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    Performance expectancy significantly affected decision-making (β=.547) and intent to use (β=.309).

    “Performance expectancy significantly affected decision-making (β=.547) and the intent to use (β=.309).”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingown data

    Most respondents used ChatGPT for information search and entertainment purposes.

    “Most of the respondents used ChatGPT for information search and entertainment purposes.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data

    The decision-making construct had a Cronbach α of .668, rhoC of 0.858, and AVE of 0.751.

    “The decision-making construct exhibited a Cronbach α coefficient of .668, a rhoC of 0.858, and an AVE of 0.751.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data

    Most respondents were familiar to a certain degree with ChatGPT and perceived it to be persuasive.

    “Most of the respondents were also familiar (to a certain degree) with the technology of ChatGPT and perceived the technology to be persuasive.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingown data

    Most respondents had a bachelor's degree, a high school diploma, or a master's degree.

    “Most respondents had a bachelor’s degree, a high school diploma, or a master’s degree.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data

    The model explains 52.4% of the variance in decision-making and 38.1% in the intent to use ChatGPT for self-diagnosis.

    “The R2 values indicated that our model can explain 52.4% of the variance in decision-making and 38.1% in the intent to use ChatGPT for self-diagnosis.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    Adjusted R2 values were 52.2% for decision-making and 37.9% for intent to use.

    “When adjusting for the number of predictors in the model, the R2 values were 52.2% for decision-making and 37.9% for the intent to use ChatGPT for self-diagnosis, demonstrating satisfactory explanatory power.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    The performance expectancy construct had a Cronbach α of .783, rhoC of 0.860, and AVE of 0.606.

    “Regarding reliability, the performance expectancy construct had a Cronbach α coefficient of .783, a rhoC of 0.860, and an AVE of 0.606.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data

    Risk-reward appraisal substantially impacted decision-making (β=.245) and indirectly intent to use (β=.138).

    “In contrast, risk-reward appraisal substantially impacted decision-making (β=.245) and, indirectly, the intent to use (β=.138).”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingown data

    A total of 607 individuals participated in the study.

    “A total of 607 individuals participated in the study, providing comprehensive responses to the questionnaire.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Interpretationown data

    People are inclined to use ChatGPT in health care contexts even though the technology is not specifically designed for health care.

    “Even though the technology is not specifically designed for health care, people are inclined to use ChatGPT in health care contexts.”
    Quote in the paperSubstantial claim
    • Rests on an assumption: ChatGPT is not specifically designed for health care purposes.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    TAM asserts that users' actual system usage is impacted by their behavioral intention.

    “TAM also asserts that users’ actual system usage is impacted by their behavioral intention”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [16] source

      Not checked in the free run.

  • Classified as probably not a substantial claim.

    Premisecitation

    Cognitive fit theory suggests that the alignment between an individual's cognitive processes and the representation of information by technology influences the effectiveness of problem solving and decision-making.

    “Cognitive fit theory complements this relationship by suggesting that the alignment between an individual’s cognitive processes and the representation of information by technology influences the effectiveness of problem solving and decision-making [18].”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [18] source

      Not checked in the free run.

  • Classified as probably not a substantial claim.

    Premisecitation

    TAM posits that users' intention to adopt a technology is influenced by their perceptions of its usefulness and ease of use.

    “TAM posits that users’ intention to adopt a technology is influenced by their perceptions of its usefulness and ease of use [16].”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [16] source

      Not checked in the free run.

  • Classified as probably not a substantial claim. May combine two claims.

    Premisecitation

    Prospect theory posits that individuals evaluate potential gains and losses during decision-making processes and that their choices are influenced by the perceived risks and rewards associated with each alternative.

    “Prospect theory posits that individuals evaluate potential gains and losses during decision-making processes and that their choices are influenced by the perceived risks and rewards associated with each alternative [23].”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [23] Kahneman, D, 2013, Handbook of the Fundamentals of Financial Decision Making: Part I source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    Trust in a technology is positively associated with performance expectancy.

    “Trust in a technology, which is positively associated with performance expectancy [17]”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [17] source

      Not checked in the free run.

  • Classified as probably not a substantial claim.

    Premisecitation

    UTAUT posits four factors influencing an individual's behavioral intention to use a given technology: performance expectancy, effort expectancy, social influence, and facilitating conditions.

    “The UTAUT framework posits 4 factors influencing an individual’s behavioral intention to use a given technology: performance expectancy, effort expectancy, social influence, and facilitating conditions.”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [15] source

      Not checked in the free run.

  • Classified as probably not a substantial claim.

    Premisecitation

    PMT suggests that individuals' intentions to engage in protective behaviors are influenced by their perceived severity of a threat, perceived vulnerability, response efficacy, and self-efficacy.

    “PMT suggests that individuals’ intentions to engage in protective behaviors are influenced by their perceived severity of a threat, perceived vulnerability, response efficacy, and self-efficacy [24].”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [24] source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductnone given

    The survey incorporated a checking question to verify that respondents thoroughly read all questions before providing their answers.

    “Additionally, the survey incorporated a checking question to verify that respondents thoroughly read all questions before providing their answers, further ensuring data quality.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Premisecitation

    People often turn to online sources for self-diagnosis and health information, despite potential negative impact on mental health.

    “Similarly, as evidenced by our prior study on internet use and mental health, people often turn to online sources for self-diagnosis and health information, despite the potential negative impact on mental health [41].”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [41] source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductnone given

    The cross-sectional survey design constrained our capacity to examine the evolving nature of users' interactions with AI chatbots.

    “Second, the cross-sectional survey design constrained our capacity to examine the evolving nature of users’ interactions with AI chatbots.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Premisecitation

    PLS-SEM is able to handle small- to medium-size samples and is suitable for exploratory research.

    “This method was chosen for its ability to handle small- to medium-size samples and suitability for exploratory research [33].”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [33] source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Interpretationnone given

    Focusing on ChatGPT, which is not specifically intended for health care applications, may limit the applicability of the findings to other AI chatbots explicitly designed for health care purposes.

    “Lastly, focusing on ChatGPT, which is not specifically intended for health care applications, may limit the applicability of the findings to other AI chatbots explicitly designed for health care purposes.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductnone given

    Informed consent was obtained from participants.

    “Informed consent was obtained from participants.”
    Quote in the paperNot a substantial claim

  • Classified as probably not a substantial claim.

    Study design and conductown data

    The structural model was evaluated to test hypotheses, including path coefficients, significance levels, and R2.

    “After confirming the measurement model’s adequacy, we evaluated the structural model to test our hypotheses. This analysis included assessing the path coefficients, significance levels, and determination coefficients (R2) for each endogenous latent construct.”
    Quote in the paperSubstantial claim

  • Classified as probably not a substantial claim.

    Study design and conductown data

    The measurement model was assessed for reliability and validity focusing on indicator reliability, internal consistency reliability, convergent validity, and discriminant validity.

    “We assessed the measurement model for reliability and validity by focusing on 4 aspects: indicator reliability, internal consistency reliability, convergent validity, and discriminant validity.”
    Quote in the paperSubstantial claim

  • Classified as probably not a substantial claim.

    Premisecitation

    PLS-SEM is a powerful multivariate analysis technique that allows researchers to estimate complex cause-effect relationships between latent constructs and their indicators.

    “PLS-SEM is a powerful multivariate analysis technique that allows researchers to estimate complex cause-effect relationships between latent constructs and their indicators [32].”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [32] Hair, JJ, 2017, Advanced Issues in Partial Least Squares Structural Equation Modeling

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    Misuse (using ChatGPT for tasks it is not designed for) can affect user trust, resulting in underuse of the technology.

    “This is important because misuse (using it for tasks it is not designed for) can affect user trust, resulting in the underuse of the technology [11].”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [11] source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Study design and conductnone given

    Centiment's platform implemented various quality control measures and prevented duplicate responses.

    “Centiment’s robust platform facilitated the research team in designing and distributing the survey, while implementing various quality control measures and preventing duplicate responses.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Study design and conductnone given

    The data gathered through the online survey were securely stored on Centiment's platform and remained accessible exclusively to the research team.

    “The data gathered through the online survey were securely stored on Centiment’s platform and remained accessible exclusively to the research team.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductnone given

    The cross-sectional design of the study limited our ability to draw causal inferences over time.

    “Furthermore, the cross-sectional design of the study limited our ability to draw causal inferences over time.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductnone given

    Relying on self-reported measurements may introduce various biases, including social desirability, recall, or imprecise reporting.

    “Moreover, relying on self-reported measurements may introduce various biases, including social desirability, recall, or imprecise reporting.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductown data

    The study received IRB approval from West Virginia University (protocol 2302725983).

    “The study, classified as a flex protocol type, received approval from the Institutional Review Board (IRB) of West Virginia University (IRB protocol number 2302725983).”
    Quote in the paperNot a substantial claim

  • Classified as probably not a substantial claim.

    Study design and conductcitation

    Eliminating a neutral option encourages participants to critically contemplate their responses, yielding more accurate data.

    “Eliminating a neutral option encourages participants to critically contemplate their responses, yielding more accurate data [28].”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [28] Garland, R, 1991, Mark Bull

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based). May combine two claims.

    Study design and conductnone given

    The online survey format allowed participants to complete it at their discretion, contributing to an increased response rate and enhanced sample diversity.

    “Furthermore, the online survey format allowed participants to complete it at their discretion, contributing to an increased response rate and enhanced sample diversity.”
    Quote in the paperNot a substantial claim
36 of 36

  • The paper relies on this without support.

    ChatGPT is not specifically designed for health care purposes.

    “ChatGPT is not specifically designed for health care purposes”
    “ChatGPT is not specifically trained in medical literature.”

  • The paper relies on this without support.

    Users are increasingly inclined to use ChatGPT for self-diagnosis.

    “Users are increasingly inclined to use ChatGPT for self-diagnosis, necessitating a deeper understanding of the factors driving this trend.”

  • The paper relies on this without support.

    Inaccurate or incomplete information provided by ChatGPT may result in misguided self-diagnosis or exacerbation of existing conditions.

    “Inaccurate or incomplete information provided by ChatGPT may result in misguided self-diagnosis or exacerbation of existing conditions.”

  • The paper relies on this without support.

    The misalignment between AI-generated information and users’ mental models can lead to erroneous decision-making and unfavorable health outcomes.

    “From a cognitive human factor standpoint, the misalignment between AI-generated information and users’ mental models can lead to erroneous decision-making and unfavorable health outcomes.”

  • The paper relies on this without support.

    The study is the first to investigate the effects of perceived ChatGPT effectiveness and risk-reward appraisal on decision-making.

    “Although our study is the first to investigate the effects of perceived ChatGPT effectiveness and risk-reward appraisal on decision-making”

  • The paper relies on this without support.

    Our study is unique in specifically examining the factors influencing users’ intentions to use ChatGPT for self-diagnosis.

    “However, our study is unique in that it specifically examines the factors influencing users’ intentions to use ChatGPT, an AI chatbot not designed for health care purposes, for self-diagnosis.”

  • The paper relies on this without support.

    Human behavior has consistently demonstrated a tendency to repurpose technology for purposes beyond its original design, even when aware of the potential risks or drawbacks.

    “Human behavior has consistently demonstrated a tendency to repurpose technology for purposes beyond its original design, even when aware of the potential risks or drawbacks.”
7 of 7

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

Each contradiction between two passages listed here was confirmed by an independent check.

  • Decision-making's total effect conflicts with its direct effect

    Minor

    Table 3 gives the decision-making-to-intent coefficient as .565 under direct effects but .138 under total effects. With no indirect path from decision-making to intent shown, its total effect should equal its direct effect.

1 of 1

Cites

Flagged first, then by citations

Research works printed but not linked: 2

  1. 10.5555/3495724.3495883 · DOI
  2. Garland, R, 1991, Mark Bull, 2, 66

Cited by

Sorted by citations

Related

Authors
Shahsavar, Y., & Choudhury, A.
Affiliations
West Virginia University (US)
Corresponding author
Choudhury, Avishek
ORCID
Venue
JMIR Human Factors
Publisher
JMIR Publications Inc.
Published
2023-05-17
Volume
10
Pages
e47564
ISSN
2292-9495 (ISSN-L)
DOI
10.2196/47564
PMC
PMC10233444
OpenAlex
W4376872703
Semantic Scholar
e416da90b398b6c598662bc4b4310b46fa9ca96f
Cited by
341
References
42
Citations by year
2026: 92 · 2025: 124 · 2024: 98 · 2023: 26
FWCI
13.58
Type
journal article
Language
en
Open access
Licence
cc-by
Locations
  • JMIR Human Factors (publishedVersion, cc-by)
  • PubMed Central (publishedVersion, cc-by)
  • PubMed (publishedVersion)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
Topics
  • Health Sciences › Medicine › Health Informatics › Artificial Intelligence in Healthcare and Education
  • Health Sciences › Health Professions › Health Information Management › Health, Technology, Consumer Behavior
  • Social Sciences › Decision Sciences › Information Systems and Management › Technology Adoption and User Behaviour
Keywords
ChatGPT, self-diagnosis, technology acceptance, performance expectancy, decision-making, intention to use, responsible AI use, user perceptions of AI

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Claims read by deepseek-v4.1-flash under definition version 3 (extract-v3) on 2026-10-07.

Classified by jev-1.13-20260917 (gold-guideline-v0.2) on 2026-10-07.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review