AI-powered Web Search Service: Automatic Trend Monitoring

The editor logs in and opens the “Charts” ranking. Here, the most popular tracks, albums, singles, and music videos in the Asian music scene are displayed.

For each trend, full information can be viewed: chart positions and ratings, artist, release date, group members, music genre, and much more. A separate item shows the country where the trend originated.

This list includes not only facts but also links to sources. If desired, the editor can follow a link and quickly perform fact-checking.

To identify trends, we parse dozens of platforms every few hours. These include social media, charts, streaming services, and music news websites. We also analyze Asia-specific platforms that have no equivalents in the English-language internet.

When a new album or single appears simultaneously on multiple platforms, we consolidate all the information into a single trend. This prevents users from being overwhelmed by duplicate trends.

Trends that are losing popularity but are still relevant are given the status “Fading.”

The service not only identifies trends and gathers information about them. The editor immediately sees several post options for each trend. They can use the text as-is or edit it. In any case, this takes significantly less time than writing the text from scratch.

The editor can like or dislike a trend. After that, copywriters and SMM specialists can immediately see which news items the editor approved and prepare the final post.

In addition to likes and dislikes, the editor can assign different priorities to trends.

Trends that have lost relevance are moved to the archive. This happens 24 hours after the system registers that the trend is no longer in the charts. The archive has the same functionality: if desired, editors can revisit “outdated” trends to highlight them in social media channels.

The system operates as a funnel of AI agents. One agent monitors platforms and identifies trending topics. A second agent searches the internet for information on each trend. A third agent generates original text based on the search data.

Then three more agents come into play, each creating post options — every agent proposes its own text. Each of these agents has its own prompt and separate database, and each is trained on examples from different posts.

Manual information search is slow and therefore costly. Editors need to constantly keep dozens of tabs open: charts, social media, streaming platforms, and news sites. On top of that, the most up-to-date information is on Asian platforms, where users publish content in their native languages, making it even harder to navigate.

In our service, this entire process is 100% automated. All information is presented in a convenient format, with verification links right alongside. Everything an editor needs is accessible in just one minute.