Can AI create fake news that nobody can detect?

AI can create fake news that is extremely hard to detect, especially for average readers and even for many automated detectors. In some cases, it may be effectively indistinguishable from real content if you only look at the text, image, audio, or video itself.

But it is not accurate to say it can be detected by nobody, or that detection is always futile. Detection can sometimes rely on clues outside the content itself, such as inconsistent facts, timestamps, coordinated posting patterns, bot behavior, forensic traces, or provenance signals and metadata when available. Standards and platform tooling for content provenance are also being adopted, which can help in some cases.

So the best answer is: AI can create fake news that may be very hard, and sometimes nearly impossible for an individual or a detector to identify with certainty from the content alone, but it is not safe to assume it is universally undetectable.