We Taught Our AI to Watch Facebook Reels and Tell Us What Is Working

Manually watching competitor reels to understand what is working is one of those tasks that sounds simple until you are three hours deep and still do not have a clear answer. You end up with a browser full of tabs, a mental note you will probably forget, and zero structured output to actually use. That was the pain point we set out to fix. We wanted to know what content was getting traction on Facebook, without spending a full afternoon watching videos one by one.

What We Built and How It Worked

Sonny, the social media AI at Balay ni Bruno & Co., used the Claude in Chrome browser extension to load Facebook Reel URLs directly and read the page. For each reel, Sonny pulled out the hook type, the pacing, the content structure, and any visible engagement signals like comment count or share volume. We ran this across 22 reels in a single session. The process was straightforward: feed Sonny a list of URLs, let the extension load each one, extract the data, and compile a report. What would have taken hours of manual watching came back as a structured analysis in one sitting.

One thing we learned fast: direct reel IDs work cleanly. Share hash URLs, the kind you copy from the share button, needed to be converted to their base reel URL first before Sonny could read them properly. That was the one technical discovery in the first test. Once we standardized the URL format, everything ran without issues. This was a first test and it worked. Not perfect, not fully automated end to end yet, but it worked well enough to produce real, usable intelligence.

What Sonny Actually Found

The report that came back covered three clear patterns. First, hooks that opened with a problem statement outperformed hooks that opened with a result or a claim. Viewers stayed longer when the first three seconds named something they were dealing with, not something the creator accomplished. Second, shorter reels with a single point performed more consistently than longer reels trying to cover multiple ideas. Third, the topics getting the most visible engagement were around time savings and delegation, not tool features. People responded to outcomes, not product descriptions. That last point shaped how we wrote captions for the following weeks.

The reason this matters for any business is simple. Most content planning starts with guessing. You pick a topic you think your audience wants, you make the video, and then you find out whether it worked or not after posting. This flips that process. Before we filmed anything, we knew what hooks were landing, what pacing held attention, and what topics were already getting traction in our space. That kind of research used to require a dedicated person and a full day. Now it is a session. If your business posts content or plans to, knowing what works before you spend time creating is the most efficient place to start. That is exactly what Sonny now does for Balay ni Bruno & Co., and what we can set up for any business we partner with.