FOLLOW US

A quiet experiment carried out during a BBC investigation has revealed something unsettling about the new generation of artificial intelligence tools shaping how millions of people get information online. In just 20 minutes, it was possible to trick major systems including ChatGPT, Gemini, and Google’s AI Overviews into repeating a false claim about a made-up identity. The test was simple, almost laughably so. The implications are not.

The findings point to a growing vulnerability in AI-driven search results, where misleading content can be injected into the information ecosystem with minimal effort. As these tools become a primary gateway to knowledge on health, finance, and everyday decisions, experts warn the risks are escalating faster than protections are being built.

A Simple Experiment Exposes a Bigger Risk in AI Search

The investigation began with a deliberately trivial claim. A fabricated online article was published stating that the author was a world-champion competitive hot-dog eater. Within a short time, that claim began appearing in responses generated by multiple AI systems, including tools embedded in major search engines.

The ease with which the false information spread highlighted a structural weakness. Rather than independently verifying facts across multiple reliable sources, AI systems often rely heavily on single web pages or isolated posts when generating answers. That design choice, meant to improve speed and relevance, also creates a pathway for manipulation.

What started as a test quickly revealed a broader issue. Researchers and SEO specialists have observed similar tactics being used far beyond internet jokes. In some cases, manipulated blog posts have been used to influence medical interpretations of supplements or distort financial guidance that users receive through AI-generated search summaries.

The concern is no longer theoretical. It is operational, widespread, and already shaping real-world decisions.

How AI Answers Are Being Manipulated Through Simple Online Content

The core vulnerability lies in how generative AI systems retrieve and summarize information. When ChatGPT, Claude, or Google’s AI tools search the web for an answer, they may prioritize a limited number of sources. If those sources are biased, misleading, or strategically engineered, the output can reflect that distortion.

According to experts monitoring AI search behavior, this can be achieved through surprisingly basic methods. A single blog post, carefully optimized for search visibility, can be enough to influence how an AI system frames an answer.

That opens the door to deliberate manipulation at scale. Companies or individuals can seed the internet with content designed not necessarily for human readers, but for machine interpretation. Once indexed, it can be picked up and amplified by AI systems presenting it as factual summary.

The BBC investigation found that these techniques are not isolated. They are being used in competitive online environments where visibility in AI-generated answers can translate into real financial gain or reputational damage.

A Growing Economic and Personal Risk for Users Worldwide

The consequences extend far beyond misinformation about novelty claims. Experts warn that distorted AI outputs can influence purchasing decisions, medical choices, and even legal assumptions.

"At the most basic level, the concern is the economic impact," says Harpreet Chatha, who runs the SEO consultancy Harps Digital. "At a more serious level, you might take medical advice that makes you sicker than you were before. Legally, you might get bad information and do something that is not legal in your state or your country."

The scale of exposure is enormous. More than a billion people now use AI chatbots regularly, while Google’s AI Overviews reportedly reach 2.5 billion users each month. With that level of reach, even small inaccuracies can scale into widespread misinformation within hours.

The concern is not only what is wrong, but what feels confidently right. AI systems present answers in a single, authoritative voice, reducing the visible uncertainty that traditionally helped users evaluate credibility.

Google’s Response and Policy Update Signals Pressure Behind the Scenes

In response to growing scrutiny, Google has updated its spam policies to explicitly address manipulation of AI-generated answers. The change indicates that attempts to influence AI outputs are now formally recognized as violations that can lead to demotion or removal from search visibility.

Google maintains that the update does not represent a shift in strategy, but rather a clarification of existing enforcement.

"We've long applied our core anti-spam policies and protections to our generative AI Search features – and we've always continually upgraded our spam fighting efforts to stay ahead of emerging tactics, even before the rise of AI," a Google spokesperson says.

The company also emphasized continuity in its approach.

"The edit to our spam policy language last week was a clarification, not any change in approach,"

Despite this, independent researchers argue that manipulation tactics continue to work in practice. In repeated tests, including one conducted by SEO researcher Lily Ray, AI systems were still tricked into repeating false or misleading claims, suggesting enforcement has not yet caught up with evolving strategies.

Experts Warn Users Remain Exposed to One Answer Thinking

For many researchers, the deeper issue is not just manipulation but structure. AI search tools are shifting users away from multiple sources of information toward a single synthesized answer. That change alters how trust is formed online.

"You should assume that you're being manipulated until they have better systems in place," says Lily Ray, founder of the search engine optimisation and AI search consultancy Algorythmic.

She highlights a fundamental shift in information consumption.

"We're moving towards this 'one true answer' world. Before, Google would give you 10 blue links and you would kind of do your own research. But AI just gives you one answer. It becomes so easy to just take things at face value. You need to be careful."

That consolidation of information increases efficiency but reduces transparency. Users are less likely to see the underlying sources shaping a conclusion, making it harder to detect bias or manipulation.

A System in Constant Conflict With Itself

Even as companies attempt to strengthen safeguards, experts say the system is already adapting in ways that complicate enforcement. As Google tightens rules around spam and manipulation, new tactics emerge in adjacent spaces such as influencer content and video platforms.

Search specialist Lily Ray notes that AI systems may begin filtering out self-promotional content more aggressively, even when it is technically legitimate. That could reduce direct manipulation but also shift incentives toward less visible forms of influence.

"So if you publish a list where you say you're the greatest hot-dog-eater, they're not going to include your name," says Ray. "They might still cite your article, but you're going to be removed from consideration."

There are also early signs of platforms introducing caution signals, labeling responses with uncertainty warnings or encouraging users to verify information elsewhere. Chatbot providers are increasingly acknowledging the presence of spam-like behavior in their systems, although companies including OpenAI and Anthropic declined to comment on specific changes.

The Next Phase of the Information Battle

For now, experts describe the situation as a cycle of adaptation. Platforms introduce safeguards, manipulation tactics evolve, and the process repeats. The result is a constant struggle between detection and deception.

Chatha argues that enforcement alone may not solve the problem.

"Google is playing whack-a-mole," he says. "They're announcing the policy update to deter people, but the tactics will just move."

He points to emerging strategies where influence shifts away from written content toward video platforms and paid promotion networks, areas where AI systems are increasingly drawing information.

As AI continues to become the default layer between users and the internet, the stakes grow higher. The challenge is no longer just about ranking search results. It is about who controls the answers people believe without question, and how easily those answers can be altered.

For users, the warning from experts is increasingly consistent. AI tools are powerful, fast, and persuasive. But without strong verification systems, they remain vulnerable to the same problem that has always shaped the internet. Anyone can publish a claim. Now, machines may repeat it at scale.