Social listening is software that reads what people say about a brand on social media, and shows the brand what is being said, by whom, and how people feel about it. A closely related job, media monitoring, does the same for news websites, newspapers, television and radio. Most products now do both. They read sources in dozens of languages, and then they help the brand answer.
The software first collects. It gathers public posts and comments from social media platforms, articles from news websites, and transcripts of television and radio. A transcript is the spoken words written out as text. It keeps every item that mentions the brand, its products or its executives, along with chosen keywords.
Then it reads each item. Older products counted mentions, and guessed whether a post was pleased or angry from keywords such as "love" or "worst". This guessing is called sentiment analysis. Keywords often went wrong. Now a large language model reads every post. It recognises sarcasm far more often than keyword counting did, but not every time. It reads "Wonderful, another three-hour delay. Thank you so much!" as a complaint. It also reads Hinglish, a mix of Hindi and English words, usually written in English letters. The model groups posts on the same subject, and writes a short summary of what changed since the morning.
Imagine an airline on a stormy evening at Delhi airport. Twelve flights are delayed. Within an hour, 4,000 posts mention the airline, in English, Hindi and Hinglish. The dashboard shows that 80% of them are negative, and that they come from a few topics: no announcements at Gate 42, and long queues at the help desk. One passenger's video of the crowded gate has been shared 9,000 times. The software sends an alert to the communications team. The team posts an update on the delays. It also sends the angry posts from passengers who shared a booking number to the contact centre, where support staff reply to each one.
Brands also use listening for research. Before launching a product, a brand can read what people already say about the category. Survey software can then ask a sample of people directly. Brands also use it to find creators who already post about their products, for campaigns run with affiliate links for creators.
The brand's own posts, and the replies to them, are handled in social media management software. Reviews on shopping sites and on maps are covered by ratings and reviews software and online reputation management.
The models that read the posts also write many of them. So listening products now also try to spot posts from fake accounts and bots, which can make a complaint look bigger than it is.