Bots supplied 76% of responses to UK vaping consultation, researchers find

A UK consultation meant to gather public views received more than three automated submissions for every other response. Research published on October 27 found that bots supplied 90,835 responses to the government’s 2023 tobacco and vapes consultation. Officials excluded them from decision-making, but researchers warn that future attempts could be harder to detect.

The findings appeared in Tobacco Control, in research by Nason Maani of the University of Edinburgh and Anna Gilmore of the University of Bath. As the Guardian reported, the Department of Health and Social Care identified 76% of the consultation’s 118,756 responses as machine-generated.

Repetition, speed and a narrow target

The submissions left a pattern that officials could screen for. According to the department, responses were frequently duplicated, arrived at high speed and came from a small number of IP addresses—the internet addresses associated with their connections. The department used bot-detection methods to identify the automated traffic.

The activity was also concentrated on one policy choice. All but two of the automated responses answered the same question in the “smoke free generation” consultation, opposing the proposed ban on disposable vapes. The target was a proposal from Rishi Sunak’s Conservative government, rather than a spread of unrelated consultation questions.

Those details establish the scale and pattern of automation. They do not identify a particular generative-AI model behind the responses. The researchers described the incident more specifically as what “appears to be the first documented instance” of large-scale automated interference in a formal public-health policy consultation—a qualified assessment, not proof that no earlier attempt occurred.

A detected attempt, not a policy victory

The department excluded all 90,835 automated responses from its decision-making. Its spokesperson said the controls worked effectively and that, after the fake responses were removed, the consultation helped inform the government’s action toward a smoke-free generation.

The government also said it would remain alert to interference in future consultations. Maani’s warning looks beyond this successful detection: he cautioned that campaigns could become harder to catch, leaving officials unaware of activity their checks miss.

Such campaigns are a threat because they raise the possibility of a handful of powerful interests looking like a mass movement.

Nason Maani, University of Edinburgh researcher

Stronger barriers can block real people, too

Azzurra Moores, associate director for information ecosystems at the thinktank Demos, identified a tension in the response. Barriers designed to keep bots out can also make participation harder for real people, she said. Stronger screening therefore carries a cost for the openness that makes public consultations useful.

Moores described consultations as one of the main ways government opens its decision-making to the public. Swamping them, she argued, can stop them from conveying what people actually think and undermine trust in the process. She also warned that AI makes fake responses harder to spot and that interference can adapt faster than departments update their checks.

Her proposed response combines better screening with greater transparency, rather than treating either as a complete fix. Caroline Cerny, deputy chief executive of Action on Smoking and Health, put the decision-making principle plainly: public-health decisions should rest on sound evidence and expertise, not manipulated response numbers.