Here is a short story about the afternoon an AI assistant confidently told us the wrong thing, backed it up with “sources,” and nearly had us publish it.
We were researching a guide on Making Tax Digital for landlords. We asked an AI assistant to gather supporting articles, and it produced a neat, convincing list. The trouble appeared when we started opening them. Several of those “news articles” were actually company blog posts, and those posts had themselves been written by AI. The assistant was quoting other AI as its evidence, and some of the MTD information it served up was simply wrong.
We only caught it the slow way. We opened Google, searched by hand, and clicked through to an official government page to confirm the real rules. Even when we asked the assistant to rely on government websites alone, it could not reliably do it. That one afternoon taught us more about the limits of AI than any headline could.
What Is the AI Echo Chamber?
The AI echo chamber is what happens when artificial intelligence increasingly repeats content that other AI has produced. Instead of tracing a fact back to original, verified, human research, one system picks up a claim from another, which picked it up from another. Nobody in the chain necessarily checked whether it was true, yet the answer arrives sounding polished and authoritative.
There are actually two related problems here, and it helps to keep them apart.
The first is what we ran into: an information problem. The web is now filling up with AI-written articles, and when an AI tool goes looking for sources, it can end up citing that machine-made content as if it were independent reporting.
The second is a deeper, longer-term concern that researchers call model collapse. This happens when AI models are trained on data that is itself saturated with machine-generated content rather than genuine human input. Over successive generations, small errors and biases get amplified, outputs grow more uniform, and the richness of real human perspective fades. It is a training problem rather than a search problem, but it comes from the same root: AI learning from AI.
How Misinformation Gets Laundered into “Fact”
The mechanism is quietly clever. A wave of AI-written articles appears across dozens of separate websites, each one lightly reworded, all repeating the same underlying claim. These mass-produced articles and posts appear to come from multiple sources, but they essentially push the same narrative.
To a search engine, and to an AI gathering evidence, that looks like agreement. Twenty websites saying the same thing feels like solid proof. In practice it can be a single unchecked claim, copied twenty times. The repetition manufactures confidence, and as one researcher on the subject put it, the more you are exposed to certain information, the more you believe it is true, even when it is inaccurate.
That is precisely what happened to us with MTD. The claim was everywhere, so it looked settled. It was not.
Why This Matters for Landlords, Guests, and Tenants
You might wonder why an accommodation company is writing about artificial intelligence. The answer is straightforward. The subjects where AI most often slips up are the ones that cost real money and cause real stress: tax rules, tenancy law, deposit protection, mortgage criteria, travel requirements and local regulations.
If you are a landlord leaning on a chatbot to explain your obligations, or a guest checking the rules for a trip, a confident but wrong answer can send you in the wrong direction entirely. AI is a superb place to start gathering ideas and getting your bearings. It is a weak place to stop when the question carries a financial, legal or safety consequence.
How to Protect Yourself
You do not need to give up on AI. You just need a simple habit of checking.
- Go to the source. For tax, benefits or legal questions, confirm the answer on GOV.UK or the relevant official body. One government page outweighs a hundred blogs.
- Check the author, not just the claim. If an article has no named author, no date and no publisher you recognise, treat it as unverified.
- Be wary of suspicious agreement. If ten sites word a point almost identically, that is a red flag for copied AI content, not proof that it is right.
- Ask AI for its sources, then actually open them. The claim and the source do not always match, as we found out.
- Use AI to draft, not to decide. Let it speed up your thinking, then verify anything that matters before you act.
Our Promise on the Content We Publish
This experience reinforced something we already held to. Every guide we publish, from tax explainers to local tips, is checked against primary sources before it goes live. We use modern tools to work faster, and we still confirm the facts ourselves, because your trust is worth more to us than a quick turnaround.
AI is a genuinely useful assistant. It is not yet a reliable referee, and knowing the difference is what keeps you, and us, on the right side of the facts.
If you ever spot something in one of our articles that does not look right, please tell us. We would far rather hear it from you and put it straight than let a single wrong word stand.
