With the rise of AI, so too comes the ability to generate human-like text. But can we reliably detect if a piece of Hindi text was actually written by a machine? Researchers recently put this to the test, creating a "Counter Turing Test" specifically for Hindi text. They fed headlines from real Hindi news articles to 26 different Large Language Models (LLMs), including big names like GPT-4 and Bard, as well as open-source models like Gemma. The goal was to see how convincingly these AIs could generate news stories based on these headlines, then to determine if existing AI detection tools could spot the fakes. They introduced a new dataset, AGhi, composed of thousands of these AI-generated Hindi news articles (along with real ones for comparison) as testing grounds. The researchers analyzed five current AI detection techniques to uncover how well they perform on this dataset and the outcome revealed both strengths and weaknesses. Established methods struggled to flag sophisticated AI-generated text. Surprisingly, text from open-source models was easier to identify than output from the larger, more advanced AIs like GPT-4. To measure the slipperiness of these AI texts, they also introduced the Hindi AI Detectability Index (ADIhi), offering a new tool to benchmark how easily different LLMs can be identified. This work highlights the cat-and-mouse game between AI text generation and detection. While it sheds light on the current limitations in reliably spotting Hindi AI text, it also sets the stage for stronger detection techniques and establishes open-source resources for other researchers to build upon. The future likely holds more sophisticated tools to determine AI authorship. However, the challenge will continue as LLMs evolve and produce even more nuanced text that blurs the line between human and machine creation.