In a single Reel, people show products, review them out loud and reveal their lifestyles in the background. Much of that never reaches brand reports: if the caption doesn’t name the brand, a product clearly on screen simply doesn’t count. At Korea’s MAX SUMMIT 2026, Lee Seung-gon, Korea head of AI social listening company Syncly, presented a session with LG Electronics’ customer insight team on finding the “hidden 60%.” His point: the data isn’t missing; it exists but goes unread. That’s the case for multimodal social listening.
A Reel carries three kinds of information:
Most listening tools read only the first. Yet consumers put most of the information in the other two — a competitor’s product sitting in frame, the time and place of use, or a spoken line like “great after my serum in the morning.” Lee compares it to understanding a film from subtitles alone.
Clients often say a direct platform search shows hundreds of posts while their report shows a handful. Two causes: data collection on video platforms is limited, and what is collected is analyzed mostly by caption. So the first bottleneck to check is collection — specifically, query conditions. Narrowing Boolean queries to cut noise also cuts relevant conversations, a loop Lee calls “query-building hell.”
Syncly’s pitch is natural-language input with AI filtering by context — separating the cosmetics brand Missha from a football coach named Misha without long exclusion lists — plus speech recognition, on-screen text recognition and computer vision to pull signals from the video itself. The company says this cut setup and cleanup effort by about 80% and compressed insight-to-action from two weeks to hours; those are vendor figures.
LG’s customer insight team analyzed a year of interior-design Instagram Reels — content that rarely names appliance brands in captions. By reading visuals and audio, the team quantified how TVs are installed: wall-mounted, on stands or hidden inside walls. It measured how common “hiding appliances” really is, and which life stages and motivations drive people to use living rooms as studies. Questions that lived in intuition gained quantitative answers: how many, who and why.
The session closed with “Stop Listening, Start Moving.” In many organizations, social listening ends as a monthly report; if nothing changes, the listening hasn’t become work yet. Syncly says it is now connecting insights directly to creator selection and ad creative.
Search your brand directly on Instagram or YouTube and compare the volume with your report numbers. A big gap doesn’t mean you’re using the tool wrong; it means consumers moved where they talk. The goal is turning what you hear into action, as IHOP did by reviving a fan-requested menu item and as mining community questions for GEO category entry points shows.
Much of what consumers say about brands lives in video visuals and audio, which reports miss when captions don’t name the brand. The data exists but isn’t being read.
It analyzes on-screen products, spaces and logos and spoken reviews and emotion, not just captions and hashtags, extracting usage context from each video.
Compare direct search results for your brand on Instagram or YouTube with your report numbers. A large gap means you should review collection scope and query conditions first.
To apply what you just read to your own site, start with a free audit of where things are now.
A strategist replies within 24 hours on business days.
TikTok’s Road to Black Friday series shows how top TikTok Shop sellers prep for BFCM: Princess Polly’s 6,000 creator videos a month, Dossier’s sampling engine and SharkNinja’s LIVE strategy.
Meta shared how small businesses use its Muse AI agents: analyzing ad performance, building campaigns through conversation and automating workflows with Shopify, Canva and HighLevel. Prompts included.
Reddit outlined its U.S. midterm election plans, including the Up the Vote hub, AI detection that cut spam exposure 20% and removes nearly 2 million fake upvotes daily, and AMA rules for political ads.