Food Alert has warned that hospitality businesses are facing a rise in AI-driven food fraud, and says it is already seeing cases in the UK.

Customers are using AI-generated images and AI-written complaint emails to seek refunds and compensation from restaurants, takeaways and other food operators. Some images appear to show mould, undercooked meat or foreign objects that were not present in the original meal.

The issue is becoming more acute as AI-generated material spreads online. Food Alert cited estimates that more than 34 million AI-generated images are now produced each day, while some complaint messages use legal language and references to regulators.

The pressure can be particularly strong for operators that rely on third-party delivery platforms. Refunds are often processed automatically by some platforms before the cost is passed on to the restaurant or takeaway involved.

That leaves food businesses dealing not only with the immediate cost of compensation, but also with the risk of reputational damage if images or allegations circulate more widely. Manufacturers and retailers may face similar risks if manipulated imagery is used to support false contamination claims.

“There have been instances where we’ve seen AI manipulation of images relating to foreign bodies or undercooked food complaints. We do think this is likely to increase, and we are aware that third-party aggregators reportedly receive a lot of suspicious complaints of this nature,” said Alasdair Dean, AI Lead, Food Alert.

Food Alert identified two main patterns in the complaints it is handling: fabricated photographic evidence and written complaints designed to intimidate businesses. Some complainants are using AI tools to draft emails that cite legislation and threaten to report operators to enforcement authorities, government agencies or legal representatives.

Annabel Kyle, Technical Director, Food Alert, described this as a growing part of the problem.

“A bigger trend for us is the use of AI to intimidate our food business clients and us in relation to food complaints,” said Kyle. “For example, if a guest disagrees with the outcome of their complaint, we will often receive an email that is clearly written with AI, quoting legislation and stating they will be reporting the matter to enforcement authorities, government agencies, legal representatives, and so on.”

Inspection risk

False complaints can have wider consequences if they are escalated to local authority officers. Even when an original allegation is fabricated, it can still prompt an Environmental Health Officer inspection.

Such visits may uncover unrelated issues at the premises, exposing businesses to regulatory action or harming their food hygiene rating. In sectors where hygiene scores are closely watched by customers and delivery partners, that can have commercial implications beyond the original complaint.

“Fraudulent complaints escalated to local authority EHO departments, whether through AI-generated images or intimidatory written correspondence, can trigger inspections. Even where the original complaint is fabricated, an inspection may uncover unrelated issues, creating real regulatory exposure. Repeated complaints on record can also, over time, affect a business’s food hygiene rating,” said Kyle.

Food Alert linked that risk to the cost of securing a new hygiene rating after problems are addressed. Its research found that 84% of local councils across England, Wales and Northern Ireland charge businesses for a food hygiene re-rating, with the average cost at £219.95.

Harder to spot

Identifying false complaints is likely to become more difficult as AI systems improve. Written complaints can still often be recognised because the language appears formulaic or unusually polished, but image-based claims are harder to verify.

That raises questions for food businesses handling large volumes of complaints and needing to distinguish quickly between genuine safety concerns and attempts to obtain refunds through manipulated evidence. Operators must still investigate every complaint seriously, adding to the burden on compliance and customer service teams.

“At the moment, email and written complaints are relatively easy to identify as AI language is currently fairly easy to distinguish. However, this is likely to change over time. Images are harder to spot, as while there are clues to look for, they are trickier to see. Again, these will become harder and harder to identify over time, and relatively rapidly,” said Dean.

Regulatory gap

Food Alert also argued that current UK law does not specifically address this type of AI-generated food complaint fraud. In its view, that leaves businesses exposed while regulation struggles to keep pace with the technology.

Kyle said the absence of clear legislation may encourage copycat behaviour by people who believe there is little risk in using AI tools to construct false claims.

“In the UK, there is currently nothing that governs the generation of images outside the intentional generation of sexually explicit images. This also means other people might see and hear of this type of fraud and the lack of legislation around it, and carry it out for themselves,” said Kyle.

Food Alert said businesses should tighten complaint investigation procedures, keep detailed food safety records and work more closely with delivery platforms when challenging suspicious refund requests. It added that operators should remain alert to possible manipulation without dismissing the possibility of genuine food safety incidents.

Only one in four workers in the UK feel their job is safe from elimination, according to ADP, which surveyed nearly 39,000 workers across 36 countries.

UK workers were slightly more confident than the global average, with 25% saying their role was safe from elimination, compared with 22% globally and 21% across Europe.

The figures point to a clear divide across the workforce. In the UK, knowledge workers were far more likely to feel secure than those in repetitive roles, with 34% of the former saying their jobs were safe, compared with 19% of the latter.

Gender differences were also pronounced. Women in the UK were less confident than men about their long-term job security, at 22% compared with 28% – the widest gender gap recorded in Europe, ADP said.

Workforce divide

The survey also found differences by employer size. Staff at mid-sized UK companies showed the highest confidence in job security, at 36%, compared with 23% at small businesses and 22% at large corporations.

That pattern runs against the wider European picture, where workers generally become more confident about job security as company size increases.

Employee perceptions of security also appear closely linked to wider workplace outcomes. Globally, workers who felt their jobs were secure were twice as likely to say they had no intention of leaving their employer.

They were also six times more likely to be fully engaged and 3.3 times more likely to report high productivity, the report found.

The findings come as employers assess how artificial intelligence, changing job design and demographic shifts are reshaping the workplace. While unemployment has remained relatively low in many markets, confidence in personal job security has not kept pace.

No market in the survey recorded a majority of workers who strongly agreed that their jobs were safe from elimination. The result suggests uncertainty is widespread rather than concentrated in a small group of countries or sectors.

Management challenge

For employers, the data highlights a management issue as well as a labour market one. Workers who do not feel secure in their roles may be less engaged, less productive and more likely to consider leaving, even when headline employment conditions appear stable.

The report draws on responses from working adults across a wide range of industries, educational backgrounds and working arrangements, including on-site and remote roles. It also includes workers in both management and individual contributor positions.

Jeff Phipps, Senior Vice-President and General Manager for ADP UK and Northern Europe, said the findings showed a gap between broader economic indicators and employee sentiment.

“The world of work is changing fast, and our findings reveal a gap between what the labour market is telling us and what employees are feeling. Employment is strong, but many UK workers are uncertain about what the future holds for their role. People want to know there’s a place for them as their organisation evolves, and that they’ll be supported to get there. The businesses that treat this seriously – investing in skills and being honest about change – are the ones seeing the payoff in how people perform and whether they stay,” he said.

The survey underlines how concerns about job loss are not spread evenly across the workforce. Workers in more repetitive roles appear especially exposed to uncertainty, while those in knowledge-based jobs report significantly stronger confidence.

That gap may become more important as businesses review staffing needs and the impact of AI on routine tasks. For employers, the challenge is likely to be not only how work changes, but how clearly those changes are explained to staff.

Across the UK labour market, confidence remains fragile even among workers who are currently employed. With just 25% saying they strongly believe their role is safe from elimination, a large majority remain unsure or unconvinced about their longer-term prospects.

SOFIAH NICHOLE SALIVIO

News Editor

DoubleVerify has published research with YouGov showing that fewer than half of UK consumers feel confident identifying AI-generated content online. It also said it had uncovered a network of AI-generated sports websites targeting UK football fans.

The survey of 2,000 UK consumers found that 41% felt confident spotting AI-generated content, while 92% said its rise was making it harder to trust information online. Another 81% said they were concerned about encountering fake or misleading AI content designed to appear genuine.

The findings point to a gap between public concern about synthetic media and people’s ability to recognise it in practice. DoubleVerify linked that gap to a cluster of more than 40 UK-based sports domains that it said were producing AI-written articles and pushing them into large online fan communities.

“As AI-generated content becomes more prevalent, the challenge now shifts to identifying it,” said Stuart Flint, Managing director, EMEA, DoubleVerify. “Consumers are increasingly sceptical, but they don’t feel equipped to distinguish between authentic and synthetic content. That creates a meaningful vulnerability across the open web and social platforms.”

Sports network

The websites identified by DoubleVerify use generic names such as sportupdatenow, soccertrend and sportmoves. Some publish more than 10 sports articles a day without clearly identified editorial staff or original reporting, according to the company.

DoubleVerify said the material ranged from fully AI-generated articles to AI-rewritten stories copied from established UK news publishers. In some cases, the stories included fabricated quotes attributed to real athletes and public figures.

The company described the distribution tactic as “FanFarming”, a process in which links to the articles are seeded into established sports fan groups on major online platforms. The aim, it said, was to draw readers back to websites that then monetise the traffic through advertising.

Based on DoubleVerify’s estimates, including distribution through fan pages, the network has already been exposed to hundreds of thousands of UK sports fans. Several of the domains also appeared to carry malicious advertising, the report said.

How it spreads

The report said sports fan communities are particularly vulnerable because they are built around breaking news, transfer rumours and match reaction. Posts often use urgent language to encourage clicks, then direct users to pages filled with display advertising.

One example involved an article on sportsrock.co.uk that attributed a fabricated quote to Sir Alex Ferguson about a Manchester United transfer target. DoubleVerify said the article was later shared in an online fan group called Fabrizio Romano Transfer News, which it said had 3.8 million members.

DoubleVerify said there was no evidence Ferguson had made the quoted statement. It added that the underlying transfer rumour was real, but the quote was invented.

Another example involved a story circulated in a Chelsea fan group that used a real match result as the basis for an alleged “locker room crisis”. According to DoubleVerify, the article relied on invented quotes, while the actual post-match comments from club captain Reece James had been routine and had not criticised teammates or the manager.

The report also described an Arsenal-related post claiming defender Myles Lewis-Skelly had said it was “time for me to join Chelsea” after being “thrown under the bus”. DoubleVerify said neither the transfer development nor the statement had occurred.

Ad risk

DoubleVerify said advertisements from major brands had appeared on some of the sites when campaigns did not include protections against low-quality AI-generated content. It argued that this exposed advertisers to reputational risk while also diverting spending from established publishers.

Some of the same accounts repeatedly posted links to several domains in the network within the same online communities, suggesting coordinated distribution across multiple sites, DoubleVerify said. It also said the model could generate thousands of pages and millions of potential ad impressions at low cost.

The research comes as concern grows across the media and advertising industries over the spread of AI-generated articles, images and videos that mimic legitimate reporting. The issue is particularly sensitive in sport, where transfer speculation and fast-moving news can make false claims seem plausible.

DoubleVerify said the challenge was not limited to football. In the first weeks of the year alone, its fraud unit identified thousands of similar sites across several languages, showing how quickly automated publishing is scaling across the wider web.

“This is a trust gap at scale,” said Flint. “Consumers know there’s a problem, but they don’t feel equipped to identify it. That creates an opening for bad actors to insert low-quality, AI-generated content into environments people already trust, like sports fan communities.”