Imagine the scene: you run a small business, you’re already swamped, and you need to know when to send a VAT return. You do what millions of people now do: you ask a chatbot, or type a question into Google and glance at the AI summary at the top. The answer arrives quickly, confidently, with no warning signs. But what if it is completely wrong? That is the nightmare scenario that has pushed Her Majesty’s Revenue and Customs into action. HMRC has become increasingly alarmed at the way artificial intelligence absorbs misinformation from social media and passes it off as official advice. Earlier this year, the department uncovered a case where search engines and AI-generated summaries were telling businesses the wrong deadlines for VAT payments. An investigation traced the error to a website that had been copied by other sites, amplified on social media, and then swallowed whole by AI models. The result could easily have been missed deadlines and automatic penalties for taxpayers who thought they were following official rules. As James Baker, HMRC’s head of digital communications, told Bloomberg, “Where sources conflict, trust falls and the risk of costly errors rises.” For a small business owner, a fine from the taxman is not just an annoyance; it is a direct threat to their livelihood. And with AI now woven into the fabric of everyday life, this kind of confusion is only going to get worse unless someone intervenes. That is why HMRC has concluded that simply publishing correct information on its own website is no longer enough. The modern information journey is fragmented, with people moving from search engines to Reddit to TikTok to chatbots in a matter of minutes. Baker says proactive intervention is needed, “with the right content in the right places,” to shape how the wider information ecosystem responds to and corrects misinformation quickly. In other words, the tax authority cannot just wait for people to come to it. It has to go where the confusion actually lives.
How exactly does a piece of tax misinformation go viral? It starts with a half-true tip, often posted by an online influencer trying to attract clicks. On TikTok, Instagram, and YouTube, there is no shortage of videos promising to reveal secret ways to cut your tax bill. Some are harmless; others are dangerously misleading. AI models, designed to scrape the internet and summarise what they find, have no reliable way of knowing which sources are credible. They see the same claim repeated across multiple websites and treat it as consensus. HMRC’s official guidance, buried on a government domain and written in cautious bureaucratic language, is far less likely to be quoted. That mismatch is a growing headache for the tax authority, especially because Britons are turning to AI for financial help in record numbers. A survey by the comparison website Finder found that roughly three in ten Britons have used chatbots or similar tools for financial advice. That means millions of people are making decisions based on answers generated by systems that are prone to “hallucinations”—the polite term for when an AI simply makes something up. Add in the influence of social media personalities, who are not tax experts but sound very sure of themselves, and you have a perfect breeding ground for confusion. The danger is worst for small businesses, which often lack the time or money to hire professional advisers. They search for a quick answer, find one that sounds authoritative, and act on it. If the answer is wrong, they may not realise until the penalty notice lands in the post. This is particularly alarming for HMRC because small businesses account for a large share of the UK’s tax gap—the £59 billion difference between what the government is owed and what it actually collects. When businesses are misled, the money doesn’t come in, and the burden falls on everyone else. The tax authority cannot afford to let misinformation become a permanent feature of the digital landscape.
To fight this, HMRC is doing something it has never done before: it is sending human officials into the wilds of the internet to act as trusted guides. The plan is to deploy what it calls “digital ambassadors” to online forums, social media threads, and comment sections where tax questions are being asked. These are not bots or automated replies. They are real people, trained in both tax rules and digital communication, whose job is to show up in conversations, flag correct information, and help users distinguish fact from fiction. Baker describes them as “a proactive digital equivalent to HMRC’s broadcast spokespeople”—the people you normally see on television defending a new policy or explaining a tax change. The difference is that these ambassadors will be embedded in the places where ordinary people actually gather: Reddit threads, Facebook groups, perhaps even TikTok comment sections. The idea is not to lecture or police, but to be genuinely helpful. If someone asks whether they need to register for VAT, an ambassador can drop a clear answer with a link to the authoritative page. If a thread contains a misleading video, the ambassador can gently explain why and offer the correct version. This is a subtle but important shift in how government communication works. It moves from “we publish information and hope you find it” to “we go to where you are and help you understand.” It is also a recognition that authority is no longer guaranteed by a domain name ending in gov.uk. As Baker puts it, “Authority no longer lives on your website alone. It lives in how faithfully the entire machine-readable ecosystem, from Google’s sidebar to a chatbot’s summary to a Reddit thread, repeats what you actually said.” The ambassadors will not be there to argue or act as online police. Their success will depend on trust, and trust is built through consistency and openness. If HMRC gets this right, it could set a valuable precedent for other government agencies struggling with the same problem.
Alongside the human touch, HMRC is building a more technical defence. The department has launched something called “HMRC Verified”—an alert system designed to counter online misinformation and, crucially, to guide large language models toward official advice. The idea is to create a clear, machine-readable trail that AI systems can follow when searching for tax information. If a chatbot is asked about a tax rule, it should find HMRC’s verified answer first, rather than a viral TikTok that happens to be wrong. According to HMRC, large language models have already responded to these changes and highlighted correct information more often. But the work does not stop there. HMRC is also scoping out a new AI-assisted tool that would take a more active role in the fight against misinformation. Imagine a system that constantly scans the internet for tax-related claims, compares them against official guidance, and identifies where they conflict. It would then draft a response for a human reviewer to check and approve before it is posted. This combines the speed of artificial intelligence with the judgement of a real tax expert. It is a recognition that the problem is too large and fast-moving for human beings alone, but also too sensitive for AI to handle without oversight. At the same time, HMRC has issued guidance to its own staff on how to write content that is “LLM ready,” meaning it can be easily understood and cited by large language models. This is a quietly revolutionary move. Government websites have traditionally been written for lawyers and accountants, not for algorithms. The new approach is different: make the official answer so clear, structured, and accessible that any AI system—and any human—can find it and understand it. It is like leaving a trail of digital breadcrumbs through a forest of misinformation. And it is not just about correcting errors after they spread; it is about preventing them from spreading in the first place.
Why is HMRC taking this so personally? The answer lies in the numbers. The UK has a tax gap of £59 billion, around $79 billion, which is the difference between what the government should receive and what it actually gets. Small businesses are by far the largest contributors to that gap. That does not mean they are deliberate tax dodgers; many are simply confused. When a freelancer or a new company misses a deadline or misreports a liability, it often isn’t because of greed. It is because they didn’t know any better. In the old days, they might have asked an accountant or called HMRC directly. Now they ask a chatbot, and the chatbot gives them an answer that sounds official but isn’t. This is a serious challenge for tax collection, and HMRC knows that it has to adapt. The department’s conclusion, in the words of James Baker, is that “proactive intervention, with the right content in the right places, can shape how the wider information ecosystem responds to and corrects misinformation quickly.” That is a carefully worded way of saying: we need to be faster and smarter than the algorithm. The strategy has three pillars: digital ambassadors to build trust, verified content that AI can find, and AI tools to monitor and correct the ever-changing landscape of online advice. It is a recognition that in a world of deepfakes, viral videos, and instant answers, the taxman’s official website is no longer the default source of truth. If people are going to trust the tax system, they need to see the tax system showing up where they actually live—online. There is also a fairness issue. When one business pays less than it should because it was misled online, the rest of the population has to make up the difference. And when people lose trust in HMRC’s advice, they may start believing dangerous myths: that tax is optional, that HMRC is a scam, that the rules don’t apply. That kind of cynicism is corrosive, and it doesn’t happen overnight. It creeps in when every search for help returns a different answer.
None of this is a perfect solution. The internet is an ocean of misinformation, and even the best digital ambassadors cannot be everywhere at once. AI tools are improving, but they are not magic. HMRC will inevitably continue to encounter wrong answers, misleading videos, and chatbots that confidently repeat nonsense. What matters is that the authority has recognised the problem and is willing to change its approach. Instead of standing on the sidelines and watching the information ecosystem drift further from reality, HMRC is getting into the arena. It is experimenting with new tools, new roles, and new ways of writing official content. It is trying to be where people are, speaking the way they speak, and making the correct answer the easiest answer to find. For ordinary people, the message is simple: when you ask a machine for tax advice, remember that the machine is not the taxman, and it is not a qualified accountant. It is a pattern-matching tool that has learned from the mess of the internet. HMRC’s efforts may help clean up that mess, but they are not a substitute for your own caution. If something sounds too good to be true—especially on social media—check it against official sources. And if you are not sure, ask someone who actually knows. The challenge is not just correcting one piece of bad advice; it is rebuilding the relationship between citizens and the state in a digital age. For decades, official information flowed through newspapers and letters. Now it flows through feeds and algorithms. HMRC’s response shows an understanding that institutions must earn attention, not assume it. The shift may be awkward at times—tax officials posting on Reddit is still a strange image—but it is necessary. If the tax authority can show that it is genuinely helpful, honest, and human, it may restore some of the confidence lost to years of online confusion. In the meantime, the digital ambassadors will be out there, helping to separate fact from fiction, one post and one verified answer at a time. Taxes are complicated enough. Nobody should have to worry that the advice they were given was invented by a machine and repeated by a stranger on Reddit.