Gemini helped restore YouTube email notifications for comments with Python in an hour

YouTube quietly removed email notifications about new comments, and for creators it turned out to be not a minor cosmetic change, but a real loss of a working signal. Instead of waiting for the platform to return the feature, the article's author spent one hour building their own replacement with Gemini using Python.

Why this became a problem

Until the end of June 2025, YouTube sent emails whenever a new comment appeared on a video. For many creators, this was a simple trigger: see the email, open the link, respond to the viewer. After this feature was disabled, the familiar cycle broke.

Comments didn't disappear, but noticing them became harder, and with that came an increased risk of missing questions, feedback, and engagement signals that affect both relationships with the audience and platform algorithm behavior. Formally, YouTube has other ways to track activity: built-in notifications, creator studio, third-party panels for social media management. But for the author, they didn't fit into their daily workflow.

Their work environment is email, not another dashboard that needs to be specially opened and checked. That's why the loss of the old format was noticeable. When a problem comes down not to complexity but to an inconvenient workflow, the idea of building your own small tool suddenly becomes quite rational.

How Gemini assembled the solution

First, the author took the shortest path and asked Gemini if YouTube had a ready-made RSS for comments. The answer was negative: such RSS feeds for comments were closed back in 2015. The next question turned out to be key — does an official interface exist through which you can track comments programmatically. Here Gemini pointed to YouTube Data API v3 and immediately offered to write a basic Python example that checks for new comments every hour and sends emails with links to them. The working version required very little infrastructure:

• API key for YouTube Data API v3 from Google Cloud Console
• The ID of the needed YouTube channel
• App password from the Google account for sending emails
• A small requirements file for Python dependencies
• A Dockerfile or local run if a container isn't needed

Then AI didn't just write the code, but guided the author through the setup. Gemini suggested where to enable the API, how to find the channel ID, and how to create an app password in the Google account. Then it helped package the script in Docker and run it on a Linux machine where other containers were already running. Essentially, a background service was created that could be kept running 24/7 on a home server, Raspberry Pi, or a cheap cloud instance.

What was added on top

The story didn't end there. The author immediately thought about a second problem: even if the script works today, YouTube could change the API, restrict access, or temporarily break the server response. So he asked Gemini not to ignore network errors, but to turn them into a separate notification channel. The model suggested counting consecutive failures and sending a separate email if the problem persists long enough to look like a real integration breakdown, not just a random timeout.

"The script will keep a count of consecutive failures: checking once

an hour, 48 errors in a row — that's exactly two days of API unavailability."

So a simple automation gained basic self-monitoring. As a result, the author not only got their old email alerts back, but made the system more reliable than before. The next morning they received the first email about a new comment and confirmed that the setup works as intended. The most telling conclusion in this story isn't that Python or Docker turned out to be useful again, but that Gemini Pro helped bring the idea to a working state without a multi-weekend project.

What this means

Such cases show well where AI-powered coding already brings practical value: not in flashy demos, but in small one-off automations that previously it was either too lazy or too expensive to do manually. If you have a clear pain point, access to an API, and an hour of time, models like Gemini can already close the task to a working tool.