Survey Paper DDoS Attack: AI's Impact on Research

• Survey paper DDoS attack is the unchecked proliferation of AI-generated survey manuscripts that flood academic repositories and degrade the quality of scholarly synthesis.
• This phenomenon, enabled by large language models, leads to redundancy, citation errors, and strains the peer-review process with overwhelming submission volumes.
• Proposed remedies include dynamic live surveys, mandatory AI contribution disclosures, and rigorous review standards to restore integrity and efficiency in academic publishing.

The concept of a "survey paper DDoS attack" draws an analogy between the disruptive flooding of Internet resources in distributed denial-of-service (DDoS) attacks and the inundation of academic research repositories with a large volume of AI-generated survey articles. With the widespread availability of LLMs, the barriers to producing survey manuscripts have decreased, leading to exponential growth in their number and the emergence of new challenges for scholarly communication and research quality control.

1. Definition and Conceptual Analogy

A "survey paper DDoS attack" refers to the unchecked proliferation of superficially comprehensive, often redundant, or error-prone AI-generated survey manuscripts on preprint platforms (notably arXiv). This mirrors classical DDoS attacks in that a flood of submissions degrades the accessibility and utility of genuine, well-curated scholarly syntheses. The effect is "literature clutter," impeding the ability of researchers to efficiently synthesize knowledge and undermining trust in the academic record (Lin et al., 9 Oct 2025).

In this analogy:

• The attack vector is the volume of survey papers enabled by LLM-based automation.
• The "victims" are both research consumers (who must sift through redundant literature) and the integrity of scholarly communication itself.
• The "damage" is manifested as decreased signal-to-noise ratio in search results, increased burden on editorial/review systems, and the dilution of expert-driven synthesis.

2. LLMs as an Enabler

LLMs enable rapid, low-effort production of survey papers, facilitating the assembly of literature overviews with minimal human expertise or original critical analysis. Distinct characteristics of this phenomenon include:

• Lack of deep analytical content and original taxonomies in automated survey output.
• High redundancy where surveys replicate and rearrange existing reviews rather than contribute novel synthesis.
• Increased prevalence of citation errors and factual inaccuracies due to model hallucination.

Despite their capacity to accelerate survey writing and reduce entry barriers, the indiscriminate use of LLMs in this context results in template-based content that lacks substantive scientific contribution (Lin et al., 9 Oct 2025).

3. Measurable Impact on the Academic Community

The influx of AI-generated surveys exerts multiple pressures on the research community:

• Quantitative trend analysis (see Figure 1 of (Lin et al., 9 Oct 2025)) reveals an exponential rise in survey paper submissions after the public availability of tools such as ChatGPT in 2022. Average AI-generated content scores for survey papers increased notably post-2022.
• Anomalous author behavior is documented, including multiple surveys submitted per month by single individuals, raising the specter of survey paper "content farms."
• Peer-review and curation processes become increasingly strained, as editors and reviewers must disentangle authentic, insightful reviews from repetitive or low-quality LLM-generated manuscripts.
• The consequence is "denial of service" at the literature layer—for example, important or expert-curated surveys are obscured by sheer volume, and overall trust in systematic reviews erodes.

Table \ref{tab:two growth ratio} in the paper quantifies growth ratios in AI-generated survey content, supporting the assertion of an ongoing survey paper DDoS phenomenon.

4. Remedial Measures and Norms

To mitigate the risks associated with the survey paper DDoS attack, several measures are proposed:

• Mandatory disclosure of all AI assistance used in the production of survey papers, such as footnotes or explicit "AI Contribution" sections, to increase transparency and aid reviewers in critical assessment.
• More stringent peer review regimes with defined quality thresholds for surveys, overseen by experienced area chairs or editorial boards.
• Monitoring and flagging of suspicious submission patterns, for instance, authors or research groups with unusually high survey output.
• Routine application of AI-detection tools (e.g., DeTeCtive, MAGE) for initial screening, allowing the identification and targeted scrutiny of possible LLM-generated manuscripts.

These interventions aim to realign survey publishing practices with established scientific norms and to defend against a potential erosion of scholarly standards (Lin et al., 9 Oct 2025).

5. Dynamic Live Surveys: An Infrastructure Proposal

As a systemic response, the paper proposes the creation of "Dynamic Live Surveys," envisioned as continuously updated, community-curated, version-controlled repositories:

• Automated agents monitor literature for new publications and extract data (abstracts, results, figures), populating a citation knowledge graph.
• Human domain experts curate structural synthesis, maintain taxonomy integrity, and ensure analytical depth by refining automated drafts.
• Version control enables branching, merging, and easy incremental updates, mitigating redundancy and obsolescence.
• Contributor recognition mechanisms provide incentives for sustained, expert-driven curation.

Dynamic Live Surveys seek to blend the efficiency of AI-based collation with the judgment and oversight of human experts, offering a scalable, up-to-date, and high-fidelity alternative to static, one-off survey manuscripts. Challenges include ensuring adequate human oversight, managing version histories, and fostering a collaborative maintenance culture.

6. Cultural, Ethical, and Policy Dimensions

Unchecked proliferation of AI-generated survey content raises important ethical issues:

• Diminished scholarly accountability and the threat of "citation poisoning," with low-quality surveys inappropriately influencing the literature graph.
• The challenge of balancing automated knowledge collation with the necessity for deep, critical human synthesis.
• The need for clear ethical guidelines, educational outreach on responsible AI use in academic publishing, and ongoing cultural dialogue within research communities.

In recognition of these issues, policy recommendations include:

• Enhanced quantitative research into content detection metrics (e.g., citation overlap analysis, semantic similarity, Jaccard Index on reference lists).
• The adjustment of conference/journal guidelines to enshrine transparency and quality in review articles.
• Regulatory support for infrastructures such as Dynamic Live Surveys, and the evolution of community norms to uphold the rigor and integrity of survey scholarship (Lin et al., 9 Oct 2025).

7. Quantitative Trends and Visual Evidence

Empirical evidence, as presented in Figure \ref{fig:trend} of the paper, underscores the scale of the problem:

• The left panel displays the rapid growth in survey paper submission volumes (2020–2024).
• The middle panel indicates a sharp increase in AI-generated content scores beginning in 2022.
• The right panel tallies the number of "abnormal" authors (those with >3 surveys/month). Combined with detection tool metrics (Table \ref{tab:two growth ratio}), these findings substantiate concerns over a systemic shift in survey paper authorship post-AI-adoption.

---

The "survey paper DDoS attack" introduced in (Lin et al., 9 Oct 2025) constitutes a new research community risk: massive LLM-enabled generation of low-quality survey literature threatens both the utility and epistemic integrity of academic knowledge synthesis. Mitigation requires institutional transparency, rigorous review, innovative infrastructure, and community-driven ethical standards that restore the essential balance between automation and human expertise.