I’m working on a small experiment with Gemini where I want to extract structured information from public-information websites and turn it into a clear, step-by-step answer.
One issue I’m trying to understand is how Gemini should handle information that may change over time. For example, a page may contain registration instructions, requirements, contribution information, or FAQs, and some of those details could become outdated.
I’m currently testing this with a Philippine health-information website:philhealth-portal.ph/
My questions are:
- What is the recommended way to make Gemini prioritize the current webpage content instead of relying on older information from its training data?
- Is there a reliable way to make Gemini identify which parts of the extracted answer came directly from the webpage?
- When information on the page changes, what is the best approach for detecting and reflecting those changes in an application using the Gemini API?
I’d especially appreciate advice from anyone who has implemented web-content extraction or source-grounded answers with Gemini.