interface
history
output with information and links of news reports online

I got this idea mostly from scrolling through social media all the time. Fake news pops up everywhere and spreads so fast, it confuses everyone. People just share stuff without checking if its true or not. That got me thinking, maybe I could make something to help with that. So I decided to build an AI tool that checks news and figures out if its probably real or fake.

The tool is called RealNews.AI. It takes whatever news you paste in, like a headline or a whole article, and uses AI to analyze it. Then it says if the news is likely real, likely fake, or uncertain. It even has this confidence meter thing, and a basic explanation about why it thinks that. I tried to keep it simple so anyone can get it without getting lost in details.

For building it, I used Python and Flask on the backend side. The frontend is just HTML and CSS, nothing fancy. The main part is the Google Gemini AI API, which does the actual analyzing. The way it works is, user puts in the news, app sends it to Gemini, AI looks at it and sends back the result with some reasoning and how confident it is. Then it shows up on the screen for the user. It was my first time doing all that together.

One big issue I ran into was getting the prompts right for the AI. I wanted the responses to be clear, not some complicated mess that no one understands. It took a few tries to make it simple. Also, keeping the whole app easy for beginners while making sure its accurate enough, that was tricky too. I did not want it to feel overwhelming.

Im pretty proud of how it turned out. I managed to hook up the Gemini AI into a real web app that actually works. And I did it all by myself as a student, no team or anything. Its cool that it tackles a problem like fake news in the real world. Plus, I learned a ton about putting together and deploying an AI project from start to finish.

From doing this, I picked up how to use AI APIs like Gemini. Building a full stack web app was new for me too. Prompt engineering, thats the part where you craft what you ask the AI to get better answers, I got better at that. And handling a project for a hackathon, presenting it and all, that was useful.

Next, I want to add checking sources and maybe links to verify stuff. The confidence meter could be more accurate, I think. Supporting other languages would be good too. And a browser extension so it can check social media in real time, that sounds useful. Not sure exactly when, but its on my list.