United Kingdom: Over 90,000 automated responses of unknown origin targeted a survey on puffs

A study published in Tobacco Control reveals that 76.5% of responses to the 2023 UK consultation on creating a tobacco-free generation and tackling youth vaping were identified as automated by authorities[1]. These contributions aimed almost exclusively to oppose the ban on the sale and distribution of disposable e-cigarettes, or "puffs." Excluded from the official analysis, they would have, had they been retained, reversed the trend observed among the responses deemed valid. However, the researchers are unable to identify those responsible for or any potential funders of this campaign.

A public consultation overwhelmed by automated responses

Organized from October 2023, the British government's consultation focused on a set of measures aimed at reducing smoking and vaping among young people. It addressed, in particular, the generational ban on the sale of tobacco as well as various restrictions concerning vaping products, including disposable e-cigarettes. Of the 118,756 contributions received, the UK Department of Health and Social Care (DHSC) classified 90,835 as fraudulent and generated by automated programs, or "bots." These contributions represented 76.5% of all responses. The government report published in early 2024 excluded them, retaining only 27,921 contributions.

To examine the content of these rejected responses, Nason Maani, affiliated with the University of Edinburgh and the London School of Hygiene & Tropical Medicine, and Anna Gilmore, from the University of Bath, analyzed the public consultation documents and requested information under UK legislation on access to government documents. After several exchanges, the ministry provided them with a summary table and a data file.

Almost unanimous opposition to restrictions on disposable products

The contributions identified as automated contained no free text; they were limited to selections within multiple-choice questions. Among them, 90,833 opposed restrictions on disposable e-cigarettes taking the form of a ban on their sale and distribution, representing over 99.9% of the responses classified as automated. The same number of contributions indicated that they had discovered the consultation on social media. Furthermore, 90,832 answered "don't know" to a question about an implementation period of at least six months. Nearly all automated contributions left the other questions unanswered.

According to the authors, the scale of this influx and the high degree of homogeneity of the responses suggest a coordinated campaign. Authorities noted, in particular, numerous duplicate submissions sent frequently from a limited number of IP addresses. Some data was also inconsistent with the path normally followed by someone manually completing the form.

Without filtering, the results would have given a reverse picture of valid contributions.

The exclusion of automated responses significantly altered the interpretation of the results. Among the respondents whose contributions were retained, 79.6% supported restrictions on the sale of disposable e-cigarettes. Conversely, if all contributions had been considered genuine, 81.3% of respondents would have appeared to oppose a ban on their sale and distribution. These proportions describe contributions to a public consultation, not the results of a survey representative of the British population. They nevertheless illustrate how a massive influx of automated responses could have given the appearance of majority opposition to a measure supported by most of the participants included in the official analysis.

The researchers present this case as the first documented example of large-scale automated interference in a formal public consultation on health policy. They emphasize that the filtering carried out by the authorities prevented these contributions from influencing the published results.

Faced with a campaign of unknown origin, the authors call for strengthened controls and transparency in consultations.

The study does not allow for attributing this campaign to a company, professional organization, or other actor. The authors specify that the information obtained is insufficient to determine its origin, potential sponsors, or commercial funding. However, they draw parallels between this episode and influence strategies already documented in the tobacco sector. During previous consultations, notably the one on plain packaging in the United Kingdom, tobacco companies sought to increase contributions favorable to their positions through mobilization campaigns and the distribution of reply cards. This precedent, however, does not constitute proof of their involvement in the automated responses of 2023.

The analysis also has methodological limitations. The ministry did not provide details of its detection tools or the metadata necessary for their independent evaluation. The researchers therefore relied on its classification, without being able to verify it themselves. Finally, studying a single case does not allow for determining the frequency of such practices in other consultations.

For the authors, this episode underscores the need to adapt protection mechanisms to the evolution of automated tools. They specifically mention limiting the number of submissions from a single IP address, implementing CAPTCHA-type verification tests, and strengthening the cybersecurity capabilities of government agencies. These measures can complicate automated campaigns without eliminating the incentives to influence consultations. The researchers also recommend incorporating explicit questions about conflicts of interest and improving transparency safeguards surrounding participation. They believe that government agencies with more limited technical and human resources could be particularly vulnerable to these manipulation attempts. The challenge is to preserve the integrity of consultations and maintain trust in their use for developing public health policies.

Caroline Cerny, deputy executive director of Action on Smoking and Health, reacted:« The evidence that artificial intelligence robots were used to overwhelm a Ministry of Health consultation is deeply concerning. » « Given the tobacco industry's history of attempting to undermine public health regulations, We must be vigilant about the risk that technological advances could be used to serve commercial interests in health policy and put in place safeguards. » « Decisions that affect public health must be based on sound evidence and expertise, not on manipulated response figures. »[2].

©Generation Without Tobacco

AD

[1]Maani N., Gilmore A., Examining the nature of bot responses to public health consultations: the example of automated submissions to the 2023 UK Tobacco and Vapes Bill, Tobacco Control, published on October 27, 2026, accessed on October 28, 2026

[2]Denis Campbell, AI bots flooded 2023 UK government consultation on disposable vapes ban, The Guardian, published on October 27, 2026, accessed on October 28, 2026

National Committee Against Smoking |