$ 44.83 € 50.84 zł 11.63
+11° Kyiv +11° Warsaw +17° Washington

In South Africa, specialist warns of risk of manipulating AI responses — Daily Maverick

Fedir Kryshtovskyi 28 September 2026 12:07
In South Africa, specialist warns of risk of manipulating AI responses — Daily Maverick

In South Africa, Felix Norton, who works in the field of optimization for generative search systems, expressed concern about the risk that artificial intelligence may reproduce not fabricated claims, but claims deliberately placed online. As Daily Maverick writes, this concerns influence over the sources from which ChatGPT, Google AI, and other systems generate recommendations and answers.

The LinkedIn experiment

Norton said that, as a joke, he posted on LinkedIn that he was the leading expert in GEO and AEO in South Africa. According to him, three days later Google AI Overviews began repeating this claim as fact and cited his LinkedIn post.

In the author’s view, this example demonstrates a problem distinct from so-called AI hallucinations, when models confidently invent facts, quotes, or events. He believes AI can also mislead when it accurately repeats information that someone has posted in advance for subsequent use by systems.

More current news is available on the UA.News Telegram channel Telegram.

The risk of artificially manufactured consensus

As examples of possible influence, Norton cited supposedly independent rankings paid for by the company placed first, brand recommendations made by ordinary users on Reddit, as well as fake positive reviews and the publication of negative information about competitors.

In his assessment, repetition of one claim across many websites can create the impression of an established fact, although such sources are not necessarily independent. He called this artificially manufactured consensus and noted that search engines previously mostly offered users a list of links, while AI is increasingly summarizing information into a ready-made conclusion.

The author stressed that a link in an AI response does not in itself guarantee its reliability: sources may be biased, coordinated, or commercially created. He urged readers to assess not only the website, but also the origin of the information on which the system formed its answer.

Read us on
Download our app