Key facts:

• A peer-reviewed analysis has identified the first documented case of large-scale automated bot interference in a formal public health consultation.
• During a 2023 consultation on disposable vapes and related policies, 118,756 responses were submitted to the Department of Health and Social Care.
• About 76% of responses were ultimately flagged as fraudulent, and 90,833 of 90,835 bot-like submissions answered only the question on restricting disposable vape sales.
• The automated responses overwhelmingly opposed restrictions, but once they were removed, 79.6% of legitimate respondents favored tighter rules.
• Researchers say the pattern resembles coordinated tobacco-industry lobbying and raises new concerns about AI-driven manipulation of public policy.

AI-generated submissions flood a health consultation

Public consultations are supposed to give citizens a structured way to shape decisions that affect their health, environment, and everyday choices. They rely on a simple premise: when a government asks for public input, the responses come from people who have a stake in the outcome. That premise is now under pressure from automated systems that can generate persuasive text at enormous scale, mimic ordinary citizens, and overwhelm the signal that policymakers are trying to hear.

A new analysis has found what researchers describe as the first documented instance of large-scale automated interference in a formal public health consultation. The case involves a 2023 consultation on disposable vapes and related tobacco and nicotine policies. According to the analysis, a large share of the submissions were not genuine public comments at all. They were generated by bots and submitted in a way that made them appear to come from real people.

The consultation drew 118,756 public responses, submitted through multiple-choice questions. That volume alone would have made the exercise significant. But when researchers examined the submissions, they found something far more troubling. Roughly 76% of the responses were eventually flagged as fraudulent. In other words, more than three out of every four submissions were not legitimate expressions of public opinion.

What the numbers reveal

The scale of the interference is striking, but the pattern is even more revealing. Of the 90,835 responses that were flagged as bot-generated, 90,833 answered only one question. That question asked for an opinion on restricting disposable vape sales. The near-total focus on a single policy question suggests that the automated campaign was not a random flood of spam. It was targeted at the specific issue that regulators were considering.

The direction of the responses was equally consistent. The bot-generated submissions overwhelmingly opposed restrictions on disposable vape sales. That uniformity raised suspicion among researchers, who noted that genuine public consultations usually produce a wider range of views. When the fraudulent responses were removed from the consultation data, the picture changed dramatically. Among legitimate respondents, 79.6% favored implementing restrictions on the sale of disposable vapes.

That reversal matters. It means the automated campaign did not merely add noise to the consultation. It threatened to distort the apparent balance of public opinion. If the bot responses had been accepted at face value, policymakers could have concluded that most people opposed restrictions. After the fraudulent submissions were discarded, the evidence pointed in the opposite direction.

Why disposable vapes became a policy flashpoint

Disposable vapes have become one of the most contested areas of public health policy. Health officials have raised concerns about youth uptake, nicotine addiction, and the environmental impact of single-use plastic and battery products. Many disposable vapes are sold in bright packaging and sweet flavors that critics say appeal to young people. At the same time, some adults use vapes as a smoking cessation tool, which has made the policy debate more complicated.

Governments considering restrictions must weigh public health benefits against concerns about adult access, illicit markets, and enforcement. That is exactly the kind of decision where public consultations are meant to provide useful evidence. They give regulators a sense of how citizens view trade-offs, what concerns are widespread, and where policy design might need adjustment. When automated accounts flood that process, they can undermine the credibility of the consultation itself.

The disposable vape case also illustrates why public health policy is vulnerable to coordinated manipulation. The decisions at stake can affect the revenue and reputation of large industries. They can shape consumer behavior for years. They can influence how quickly new products enter the market and how they are marketed to young people. For actors who want to delay or weaken regulation, a mass submission campaign can look like a low-cost way to create the appearance of public opposition.

Automated lobbying echoes the tobacco playbook

Researchers have pointed out that the pattern is eerily similar to earlier lobbying tactics used by tobacco companies and other industries. For decades, tobacco interests faced accusations of coordinated campaigns to influence public consultations through mass submissions. Those campaigns often relied on templates, form letters, and organized turnout. The goal was to make industry-friendly views appear widespread and grassroots.

AI-generated bots take that playbook to a new level. Instead of relying on human volunteers to submit form letters, automated systems can generate thousands of responses quickly, vary the wording, and adapt to the questions being asked. They can also operate at a scale that would be difficult for under-resourced regulators to monitor in real time. The result is a form of automated lobbying that is cheaper, faster, and harder to trace than traditional mass campaigns.

The analysis notes that the scale and directional consistency of the bot responses suggest a coordinated campaign. That does not prove who was behind the submissions. Researchers could not pinpoint the origin of the malicious responses. But the pattern is consistent with prior evidence that industry actors have tried to manipulate public consultations through mass coordinated submissions. The difference is that the latest campaign appears to have used AI-generated text rather than human-written templates.

Detection challenges and consultation integrity

Detecting bot interference is not always straightforward. A public consultation can attract genuine passion from thousands of people, and high response volumes are not inherently suspicious. The challenge for officials is to separate legitimate mobilization from fraudulent automation. In this case, researchers identified suspicious patterns: an unusually high share of responses, a narrow focus on a single question, and overwhelming opposition to restrictions that did not match other evidence.

Those clues helped researchers flag the fraudulent submissions after the fact. But the episode raises a larger question about how consultations should be designed in an era of generative AI. Governments may need stronger verification measures, such as identity checks, duplicate detection, rate limits, and analysis of submission metadata. They may also need to publish more transparent methods for detecting and removing fraudulent responses, so that the public can trust the final results.

There is a tension here. Adding too many barriers to participation could discourage ordinary citizens, especially those who are already skeptical of government. Adding too few barriers could leave consultations open to manipulation. The right balance will depend on the policy at stake, the expected volume of responses, and the resources available to administrators. But the disposable vape case shows that the risk is no longer theoretical.

Implications for public health policymaking

The broader significance of this case goes beyond vaping. Public health policy increasingly depends on data, public input, and expert evidence. If any of those inputs can be flooded by automated systems, the integrity of the whole process is weakened. A consultation that appears to show public opposition may actually show the opposite. A regulator that relies on the apparent consensus may make a decision that does not reflect public sentiment at all.

That risk extends to other areas where automated text can be generated cheaply. Consultations on food labeling, air quality, vaccine policy, drug regulation, and environmental rules could all become targets. The same techniques could be used to manufacture support for a policy, not just opposition. They could also be used to create chaos and erode trust in government, regardless of the policy outcome.

The researchers could not identify who was behind the submissions, but they warned that the incentives to manipulate public health policy are unlikely to disappear. As AI tools become cheaper and more accessible, the next consultation may face an even more sophisticated test. The case offers an early warning: public consultations need to be protected as carefully as other parts of the policymaking process, or the voice of the public can be drowned out by machines pretending to be people.

Source: Digital Trends News