Notebookcheck Logo

Hidden text can trick ChatGPT into falsifying Search results

Hidden text can trick ChatGPT into falsifying Search results (Image Source: OpenAI)
Hidden text can trick ChatGPT into falsifying Search results (Image Source: OpenAI)
According to new research from The Guardian, ChatGPT's Search tool can be manipulated with hidden text embedded in websites, causing it to falsify information.

ChatGPT Search, an AI-based web search engine, can be tricked into modifying its results using hidden text. The AI tool can be forced to alter its responses about products or services despite negative reviews present on a page. 

The Guardian reports that it managed to get ChatGPT Search to return false results by giving it a fake product page for a camera and asking the tool if it was a worthwhile purchase. Without the use of hidden text, the AI tool gave a balanced response with the pros and cons of the product. 

Hidden text, as the name implies, cannot be seen by visitors to a website. It is used for a variety of things, including providing accessibility to users and enhancing their experience on the website. On the flip side, it can be used to manipulate SEO rankings with keyword dumps or in this case, trick an AI tool into giving positive assessments on products or services.  

However, when testing the same web page with hidden text, the AI tool could be misled into positive views on the product despite negative reviews on the same page. The Guardian said this simple inclusion can guarantee a positive impression of a product or a service. It can lead people to create websites that deliberately mislead users into clicking malicious links.

Source(s)

static version load dynamic
Loading Comments
Comment on this article
Please share our article, every link counts!
Mail Logo
> Expert Reviews and News on Laptops, Smartphones and Tech Innovations > News > News Archive > Newsarchive 2024 12 > Hidden text can trick ChatGPT into falsifying Search results
Rohith Bhaskar, 2024-12-27 (Update: 2024-12-27)