CoChat
Sign Up
Cited by 708Misinformation and Its ImpactsArtificial Intelligence in Healthcare and EducationEthics and Social Impacts of AI

ChatGPT and the rise of large language models: the new AI-driven infodemic threat in public health

Open access
De Angelis, Luigi, Baglivo, Francesco, Arzilli, Guglielmo, Privitera, Gaetano Pierpaolo, Ferragina, Paolo, Tozzi, Alberto Eugenio, Rizzo, Caterina·2023·Frontiers in Public Health
DOI 10.3389/fpubh.2023.1166120PMC PMC10166793
Open this paper in CoChat — free account

Abstract

Large Language Models (LLMs) have recently gathered attention with the release of ChatGPT, a user-centered chatbot released by OpenAI. In this perspective article, we retrace the evolution of LLMs to understand the revolution brought by ChatGPT in the artificial intelligence (AI) field. The opportunities offered by LLMs in supporting scientific research are multiple and various models have already been tested in Natural Language Processing (NLP) tasks in this domain. The impact of ChatGPT has been huge for the general public and the research community, with many authors using the chatbot to write part of their articles and some papers even listing ChatGPT as an author. Alarming ethical and practical challenges emerge from the use of LLMs, particularly in the medical field for the potential impact on public health. Infodemic is a trending topic in public health and the ability of LLMs to rapidly produce vast amounts of text could leverage misinformation spread at an unprecedented scale, this could create an “AI-driven infodemic,” a novel public health threat. Policies to contrast this phenomenon need to be rapidly elaborated, the inability to accurately detect artificial-intelligence-produced text is an unresolved issue.

AI snapshot

Written from the paper’s full text.

TL;DR

This perspective article traces the evolution of large language models to ChatGPT and argues that accessible, user-centered LLMs create a new public health threat—an “AI-driven infodemic”—by enabling rapid production of human-like misinformation and fake scientific content. It calls for rapid multidisciplinary policy, detection tools, and a “detectable-by-design” approach.

Key findings

  • ChatGPT was released as a free research preview on November 30, 2022, and gained over one million users in 5 days; GPT-3 has 175 billion parameters, is 100 times bigger than GPT-2, and is approximately two times the number of neurons in the human brain.
  • ChatGPT achieved around 50–60% accuracy across USMLE examinations, near the passing threshold but inferior to Med-PaLM; GPT-4 exceeds the passing score on USMLE by over 20 points.
  • The authors introduce “AI-driven infodemic” as a novel public health threat: LLMs can rapidly generate vast amounts of human-like text, fake news, fake scientific abstracts, papers, and fabricated bibliographic references, and can be used to create targeted misinformation.
  • Detection is unresolved: commonly used GPT-2 detectors were flawed at recognizing ChatGPT-generated text, newer detectors do not perform well on GPT-4-generated text, and ICML’s 2023 prohibition on LLM use in submitted drafts had no tool to verify compliance.

Methods

  • Perspective article; no empirical study, dataset, or systematic analysis is reported.
  • The authors narratively retrace LLM evolution and review existing evidence on LLM use in medical research, scientific publishing, and public health.
  • Main analysis is conceptual and argumentative, proposing the “AI-driven infodemic” concept and policy responses such as detectable-by-design and expert monitoring groups.

Limitations

  • Not stated in the paper: no formal limitations section is provided.
  • Evident from the text: as a perspective/commentary, it presents no primary data, systematic search, or quantitative evaluation, and its claims about future AI-driven infodemic impacts are anticipatory.
  • Evident from the text: proposed detection and policy solutions are constrained by the paper’s own observation that accurate AI-text detection remains unresolved and detectors lag behind rapid LLM advances.

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

Integrity signals

Registry

Crossref · 10.3389/fpubh.2023.1166120

Cites

Statistics

Nothing to recompute.

Claims the paper makes

0 of 39 cited sources checked · 0 supported · 0 flagged

  • Premisecitation

    Junk and manipulative research, through underperforming studies or study designs unfit to carry out the intended research objective, has had an impact on the behavior of the general population and health professionals.

    “In this context, it has been more evident than ever that junk and manipulative research, through underperforming studies or with study designs unfit to carry out the intended research objective, has had an impact on the behavior of the general population and, more worryingly, on health professionals (56).”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [56] Casigliani, 2020, Too much information, too little evidence: is waste in research fuelling the covid-19 infodemic?, BMJ [Internet] source

      Not checked in the free run.

  • Premisecitation

    The use of ChatGPT in scientific literature production has already become a reality, with many authors stating they used ChatGPT to write at least part of their papers.

    “The use of ChatGPT in scientific literature production has already become a reality during the writing of this draft, many authors stated to have used ChatGPT to write at least part of their papers (39).”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [39] Chen, 2022

      Not checked in the free run.

  • Premisecitation

    New detectors developed to address the gap do not perform well in identifying GPT-4 generated text.

    “new detectors were rapidly developed and released to address this gap, but these tools do not perform well in identifying GPT-4 generated text.”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [50] Detectors save us from ChatGPT? CanAI I tried 3 Online Tools to Find Out | ZDNET [internet].

      Not checked in the free run.

  • Premisecitation

    Some people used hydroxychloroquine as a treatment for COVID-19 based on false or unproven information endorsed by popular and influential people.

    “for example some people used hydroxychloroquine as a treatment for COVID-19 based on false or unproven information, endorsed by popular and influential people (61).”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [61] Schwartz, 2022, Hydroxychloroquine for COVID19: the curtains close on a comedy of errors, Lancet Reg Health -Americas source

      Not checked in the free run.

  • Premisecitation

    Despite limitations imposed by programmers, LLMs can be easily tricked into producing text on controversial topics, including misinforming content.

    “Despite the limitations imposed by programmers, LLMs can be easily tricked into producing text on controversial topics, including misinforming content (44).”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [44]

      Not checked in the free run.

  • Classified as probably not a substantial claim. 18
  • May combine two claims. 15

  • Probably not a substantial claim: may not pass S4 (based).

    Findingnone given

    In earlier versions of ChatGPT, up to the 15th December version, when asked to cite references to support its statements, the output was a list of fake bibliographic references.

    “In the earlier versions of ChatGPT (up to the 15th December version), when asked to cite references to support its statements, the output was a list of fake bibliographic references.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Premisecitation

    GPT-3 can generate text appropriate for a wide range of contexts but often expresses unintended behaviors such as making up facts, generating biased text, or not following user instructions.

    “GPT-3 can generate text that is appropriate for a wide range of contexts, but unfortunately, it often expresses unintended behaviors such as making up facts, generating biased text, or simply not following user instructions (14).”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [14] Ouyang, 2022, Training Language Models to Follow Instructions with Human Feedback, Advances in Neural Information Processing Systems source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing).

    Premisenone given

    Ensuring that an LLM behaves in a way aligned with human values and ethical principles is a major problem, known as the alignment problem.

    “Here comes the major problem of alignment: the difficulty of ensuring that a LLM is behaving in a way that is aligned with human values and ethical principles.”
    Quote in the paperNot a substantial claim

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

    Premisecitation

    ChatGPT spread to over one million users within 5 days of its release.

    “with over one million users in 5 days from its release, (2)”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [2] Everyone’s having a field day with ChatGPT –But nobody knows how it actually works [internet]

      Not checked in the free run.

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

    Premisecitation

    ChatGPT is a large language model trained by OpenAI, released in a free research preview on November 30th 2022.

    ““ChatGPT” is a large language model (LLM) trained by OpenAI, an Artificial intelligence (AI) research and deployment company, released in a free research preview on November 30th 2022, to get users’ feedback and learn about its strengths and weaknesses (1)”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [1] ChatGPT: Optimizing language models for dialogue [internet]

      Not checked in the free run.

  • May combine two claims.

    Interpretationcitation

    LLMs learn to replicate biases and stereotypes present in their training data because their objective is to predict the next element in a text based on a large corpus of internet text data.

    “This can be explained since the objective of many LLMs, including GPT-3, is to predict the next element in a text, based on a large corpus of text data from the internet, (15) thus LLMs learn to replicate biases and stereotypes present in that data (16).”
    Quote in the paperSubstantial claim0 of 2 sources checked
    • cites [15] Korngiebel, 2021, Considering the possibilities and pitfalls of generative pre-trained transformer 3 (GPT-3) in healthcare delivery

      Not checked in the free run.

    • cites [16] Kasirzadeh, 2022, In Conversation with Artificial Intelligence: Aligning Language Models With Human Values source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S2 (checkable), S4 (based).

    Interpretationnone given

    Intercepting a product generated by ChatGPT would be much more complicated than adopting systems at the editorial level to recognize potential plagiarism in scientific articles.

    “It took a while to adopt systems at the editorial level to recognize potential plagiarism in scientific articles, but intercepting a product generated by ChatGPT would be much more complicated.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Premisecitation

    GPT-4 exceeds the passing score on USMLE by over 20 points and outperforms earlier LLMs.

    “GPT-4 exceeds the passing score on USMLE by over 20 points and outperforms earlier LLMs.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [27] Nori, 2023, Capabilities of GPT-4 on Medical Challenge Problems [internet] source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S2 (checkable), S4 (based). May combine two claims.

    Interpretationreasoning

    The ability of LLMs to rapidly produce vast amounts of text could leverage misinformation spread at an unprecedented scale, creating an 'AI-driven infodemic,' a novel public health threat.

    “Infodemic is a trending topic in public health and the ability of LLMs to rapidly produce vast amounts of text could leverage misinformation spread at an unprecedented scale, this could create an “AI-driven infodemic,” a novel public health threat.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S2 (checkable), S4 (based).

    Interpretationreasoning

    The 'AI-driven infodemic' is a consequence of the use of LLMs' ability to write large amounts of human-like texts in a short period of time, not only with malicious intent, but in general without any scientific ground and support.

    “The “AI-driven infodemic” is a consequence of the use of LLM’s ability to write large amounts of human-like texts in a short period of time, not only with malicious intent, but in general without any scientific ground and support.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Premisecitation

    Elsevier created guidelines on the use of AI-assisted writing for scientific production, requiring authors to specify the AI tools employed and giving details on their use.

    “Also Elsevier created guidelines on the use of AI-assisted writing for scientific production, confirming the rules imposed by Springer and requiring the authors to specify the AI tools employed, giving details on their use.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [52] Elsevier declaration of generative AI in scientific writing [internet]

      Not checked in the free run.

  • May combine two claims.

    Premisecitation

    Springer Nature journals have added rules that LLMs cannot be listed as authors and their use should be documented in the methods or acknowledgments sections.

    “For example, Springer Nature journals have been the first to add rules in the guide to authors: to avoid accountability issues, LLMs cannot be listed as authors and their use should be documented in the methods or acknowledgments sections (3).”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [3] 2023, Tools such as ChatGPT threaten transparent science; here are our ground rules for their use, Nature [Internet] source

      Not checked in the free run.

  • May combine two claims.

    Premisecitation

    The DPA has imposed a temporary ban on ChatGPT in Italy due to the company's failure to provide adequate privacy information to its users and lack of a suitable legal basis for data collection.

    “The DPA has imposed a temporary ban on ChatGPT in Italy due to the company’s failure to provide adequate privacy information to its users its and lack of a suitable legal basis for data collection.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [69] ChatGPT banned in Italy over privacy concerns

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S4 (based). May combine two claims.

    Interpretationreasoning

    ChatGPT's ability to follow user instructions exposes it to misuse, for example by generating convincing human-like misinformation.

    “on the other hand being submissive ab origine exposes it to misuse, for example by generating convincing human-like misinformation.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S2 (checkable).

    Interpretationreasoning

    With the global diffusion of LLMs, the infodemic concept must evolve into the concept of 'AI-driven Infodemic.'

    “With the global diffusion of LLMs the infodemic concept must evolve again into the one of “AI-driven Infodemic.””
    Quote in the paperNot a substantial claim

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

    Premisecitation

    The President of the Italian Data Protection Authority (DPA) has taken action against Open AI for serious breaches of the European legislation on personal data processing and protection.

    “Recently, the President of the Italian Data Protection Authority (DPA) has taken action against Open AI for serious breaches of the European legislation on personal data processing and protection (69).”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [69] ChatGPT banned in Italy over privacy concerns

      Not checked in the free run.

  • May combine two claims.

    Premisecitation

    In the WHO agenda, AI is considered a possible ally to fight infodemics, allowing automatic monitoring for misinformation detection.

    “In the WHO agenda, AI is considered a possible ally to fight infodemics, allowing automatic monitoring for misinformation detection”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [68] Public Health Research Agenda for Managing Infodemics [internet]

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S2 (checkable), S4 (based).

    Interpretationreasoning

    Alarming ethical and practical challenges emerge from the use of LLMs, particularly in the medical field for the potential impact on public health.

    “Alarming ethical and practical challenges emerge from the use of LLMs, particularly in the medical field for the potential impact on public health.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S2 (checkable), S4 (based).

    Interpretationreasoning

    The impact that generating false and potentially mystifying texts can have on people's health is huge.

    “The impact that generating false and potentially mystifying texts can have on people’s health is huge.”
    Quote in the paperNot a substantial claim

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

    Premisecitation

    Political Deepfakes already contributed toward generalized indeterminacy and disinformation.

    “Political Deepfakes already contributed toward generalized indeterminacy and disinformation (67).”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [67] Vaccari, 2020, Deepfakes and disinformation: exploring the impact of synthetic political video on deception, uncertainty, and trust in news, Social Media+Society source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S2 (checkable), S4 (based).

    Interpretationreasoning

    Given the nature of social media content diffusion, LLMs could be used to create content specifically designed for target population groups.

    “Given the nature of social media content diffusion, LLMs could be used to create content specifically designed for target population groups”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Premisecitation

    Many scientific journals are questioning the policy matter, publishing editorials on the topic, and updating author guidelines.

    “Many scientific journals are questioning the policy matter, publishing editorials on the topic, and updating author’s guidelines (51).”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [51] Gordijn, 2023, Ten. ChatGPT: evolution or revolution?, Med health care Philos [internet] source

      Not checked in the free run.

  • May combine two claims.

    Interpretationreasoning

    It is possible to rapidly disseminate misinformation via social media platforms and other outlets and to produce exponentially growing amounts of health-related information using LLMs, regardless of one's knowledge, skills, and intentions.

    “Not only is it possible to rapidly disseminate misinformation via social media platforms and other outlets, but also to produce exponentially growing amounts of health-related information, regardless of one’s knowledge, skills, and intentions.”
    Quote in the paperNot a substantial claim

  • Classified as probably not a substantial claim.

    Premisecitation

    The diffusion of misinformation through rapidly disseminated channels such as mass media and social networks can generate the phenomenon known as infodemic.

    “The diffusion of misinformation conveyed through rapidly disseminated channels such as mass media and social networks, can generate the phenomenon known as infodemic (57).”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [57] 2020, The COVID-19 infodemic, Lancet Infect Dis source

      Not checked in the free run.

  • May combine two claims.

    Premisecitation

    ChatGPT achieved around 50-60% accuracy across all USMLE examinations, near the passing threshold, but inferior to Med-PaLM.

    “The performance of ChatGPT on USMLE was recently evaluated and it achieved around 50–60% accuracy across all examinations, near the passing threshold, but still inferior to Med-PaLM (26).”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [26] Kung, 2022

      Not checked in the free run.

  • May combine two claims.

    Premisecitation

    ChatGPT still sometimes writes plausible-sounding but incorrect or nonsensical answers due to its inability to fact-check and its knowledge limited until 2021.

    “Despite these advancements, ChatGPT still sometimes writes plausible-sounding but incorrect or nonsensical answers, due to its inability of fact-checking and its knowledge limited until 2021 (1).”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [1] ChatGPT: Optimizing language models for dialogue [internet]

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S4 (based).

    Interpretationreasoning

    LLMs could be used to create content specifically designed for target population groups and in order to go viral and foster misinformation spread.

    “Given the nature of social media content diffusion, LLMs could be used to create content specifically designed for target population groups and in order to go viral and foster misinformation spread.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S2 (checkable), S4 (based). May combine two claims.

    Interpretationreasoning

    Human-like AI-produced contents will dramatically exacerbate every future health threat that can generate infodemics, which from now on will be AI-driven.

    “We foresee a scenario in which human-like AI-produced contents will dramatically exacerbate every future health threat that can generate infodemics, that from now on will be AI-driven.”
    Quote in the paperNot a substantial claim

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

    Interpretationnone given

    ChatGPT's ability to follow user instructions makes it great at interacting with humans.

    “ChatGPT’s ability to follow user’s instruction is a double-edged sword: on one hand, this approach makes it great at interacting with humans”
    Quote in the paperNot a substantial claim
29 of 29

  • The paper relies on this without support.

    The scientific community has not yet determined the principles to follow for a helpful and safe use of this disruptive technology.

    “Given that the scientific community has not yet determined the principles to follow for a helpful and safe use of this disruptive technology”

  • The paper relies on this without support.

    The AI-driven infodemic threat is forthcoming.

    “as the scenario is rapidly evolving and the AI-driven infodemic threat is forthcoming.”

  • The paper relies on this without support.

    The scenario is rapidly evolving.

    “as the scenario is rapidly evolving and the AI-driven infodemic threat is forthcoming.”
3 of 3

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.

  • ChatGPT is not first author of all four cited papers

    Minor

    The text says ChatGPT is listed as first author of four cited papers, but reference 26 lists Kung as first author and does not list ChatGPT as an author.

1 of 1

The citation lists haven’t been fetched yet

References 66 · Cited by 708

Research works printed but not linked: 3

  1. 2023, ArXiv
  2. Becker, 2020, ArXiv.
  3. Elsevier declaration of generative AI in scientific writing [internet]

Authors
De Angelis, L., Baglivo, F., Arzilli, G., Privitera, G. P., Ferragina, P., Tozzi, A. E., & Rizzo, C.
Affiliations
  • University of Pisa (IT)
  • Bambino Gesù Children's Hospital (IT)
Corresponding author
Baglivo, Francesco
ORCID
Venue
Frontiers in Public Health
Publisher
Frontiers Media SA
Published
2023-04-25
Volume
11
Pages
1166120-1166120
ISSN
2296-2565 (ISSN-L)
DOI
10.3389/fpubh.2023.1166120
PMC
PMC10166793
OpenAlex
W4366989525
Cited by
708
References
66
Citations by year
2026: 102 · 2025: 222 · 2024: 260 · 2023: 121 · 2022: 2
FWCI
302.95
Type
journal article
Language
en
Open access
Licence
Locations
  • Frontiers in Public Health (publishedVersion, cc-by)
  • SSRN Electronic Journal (acceptedVersion)
  • PubMed Central (publishedVersion, cc-by)
  • PubMed (publishedVersion)
  • CINECA IRIS Institutial research information system (University of Pisa) (submittedVersion, cc-by)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
  • CINECA IRIS Institutional Research Information System (Sant'Anna School of Advanced Studies) (submittedVersion, cc-by)
Funders
  • Ospedale Pediatrico Bambino Gesù
  • Ministero della Salute
Topics
  • Social Sciences › Sociology and Political Science › Misinformation and Its Impacts
  • Health Sciences › Medicine › Health Informatics › Artificial Intelligence in Healthcare and Education
  • Social Sciences › Safety Research › Ethics and Social Impacts of AI
Keywords
ChatGPT, large language models, infodemic, misinformation, artificial intelligence in healthcare, public health risk, AI-generated text detection, AI ethics, AI-assisted scientific writing, AI authorship, natural language processing
MeSH terms
Infodemic, Artificial Intelligence, Language, Natural Language Processing, Public Health

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 2026-10-08.

Classified by jev-1.13-20260917 (gold-guideline-v0.2) on 2026-10-08.

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