Google’s AI just trusted a liar. You’re next. That’s the unsettling conclusion of a BBC investigation that revealed how easily AI-powered search tools, including Google’s AI Overviews and ChatGPT, can be manipulated into spreading false claims. In February 2026, a journalist published a single fake blog post claiming to be a world-champion hot-dog eater. Within 20 hours, Google’s AI and ChatGPT both repeated the lie as fact. The same technique is already being exploited at scale by spammers targeting medical advice, financial information, and product recommendations. Google updated its spam policies in response. Researchers say the problem hasn’t stopped.

How the Manipulation Works

The technique is surprisingly simple. When AI chatbots and search tools pull information from the live internet, rather than their built-in training data, they often rely on a single webpage or social media post as a source. A well-crafted blog post published almost anywhere online can poison those results within hours.

In February 2026, a BBC journalist demonstrated the problem by publishing a fake article on a personal website claiming to be a world-champion competitive hot-dog eater. Within 20 hours, Google’s AI Overviews and ChatGPT both repeated the false claim as fact when queried.

The same investigation found the technique already in use at scale. Manipulated AI answers appeared in searches about supplement safety, retirement finances, and medical treatment areas where bad information carries serious consequences.

---

The Policy Response

Google updated its spam policies to explicitly state that manipulating AI-generated search results violates the company’s rules. A Google spokesperson characterized the update as a “clarification” of existing anti-spam efforts, not a change in approach.

“We’ve long applied our core anti-spam policies and protections to our generative AI Search features,” the spokesperson said. Google maintains it has continuously upgraded its spam-fighting systems.

But researchers and SEO specialists who spoke to the BBC argue the policy update signals something else: a recognition that existing safeguards failed.

Lily Ray, founder of the SEO and AI search consultancy Algorythmic, says users should operate under a simple assumption. “You should assume you’re being manipulated until they have better systems in place,” Ray told the BBC.

---

The Evidence: Policy Updates Haven’t Stopped the Problem

The gap between policy and reality became clear almost immediately.

After Google’s clarification, another SEO specialist replicated the BBC’s experiment. The result? Google’s AI told users the individual was an expert sandcastle builder, based, again, on a single fabricated source.

“It’s whack-a-mole,” said Harpreet Chatha, who runs the SEO consultancy Harps Digital. “They’re announcing the policy update to deter people, but the tactics will just move.”

Chatha points to an emerging shift: as Google cracks down on manipulative blog posts, bad actors are paying YouTube influencers to make claims that AI search tools then cite. “There’s nothing stopping them from paying 20 YouTube influencers to say their product is the best,” he said.

---

The Bigger Shift: From Ten Links to One Answer

Underneath the cat-and-mouse game lies a fundamental change in how information reaches the public.

For two decades, Google returned ten blue links. Users clicked, compared, and made judgments. The new AI model returns one answer, confident, declarative, and increasingly treated as definitive.

More than a billion people use AI chatbots regularly. Google’s AI Overviews reach 2.5 billion users monthly, according to company data.

“We’re moving towards this ‘one true answer’ world,” Ray said. “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.”

---

What the Companies Aren’t Saying

Behind the scenes, subtle changes suggest AI companies recognize the scale of the problem.

Ray and other researchers note that Google and ChatGPT now appear to quietly remove companies from AI-generated answers when self-promotion is suspected. Some AI tools have begun adding caveats or labels indicating low confidence or recommending third-party reviews for purchasing decisions.

ChatGPT and Claude, an AI made by Anthropic, now sometimes explicitly tell users they are filtering spam from responses.

But none of the companies will discuss these changes publicly. OpenAI and Anthropic declined to comment. Google’s spokesperson did not respond to questions about the behind-the-scenes adjustments.

---

What Comes Next

The manipulation tactics will evolve. Blog posts will become videos. Videos will become AI-generated podcasts. The sources will multiply faster than spam policies can adapt.

For users, the practical lesson is straightforward. AI answers deserve the same scrutiny once applied to search results, more so, given the single-source nature of the responses.

“You need to be careful,” Ray said.

---

Frequently Asked Questions

Can AI chatbots like ChatGPT and Google AI Overviews be manipulated?

Yes. A BBC investigation in early 2026 showed that publishing false information on a single webpage can cause AI search tools to repeat those claims as fact within hours. The technique works because AI systems often scrape answers from individual web pages rather than cross-referencing multiple sources.

How does AI search manipulation work?

When AI tools search the live internet for answers, they sometimes rely on a single blog post, social media update, or YouTube video as a source. Publishing optimized content with specific keywords can cause AI systems to surface that content in their responses—even if the information is false.

What has Google done to stop AI search manipulation?

Google updated its spam policies to explicitly prohibit manipulating AI-generated search results. However, researchers demonstrated the same vulnerability after the policy change, and Google has not publicly detailed specific technical fixes.

What kinds of searches are being manipulated?

Researchers have documented manipulated AI answers in searches about supplement safety, retirement finances, medical treatments, and commercial product recommendations, areas where bad information can cause financial loss or health risks.

How can I protect myself from AI misinformation?

Experts recommend treating AI-generated answers with skepticism, cross-referencing important information with multiple sources, and paying attention to caveats or confidence labels that some AI tools now include in their responses.

---

About the Author

Anova Stream covers technology policy and digital information integrity. Their reporting on AI and search manipulation has appeared in multiple international publications.