What is sentiment analysis
Sentiment analysis is the automated process of reading text and judging where an opinion falls on a scale between positive and negative.
Neutral opinions fall in between. Software runs it on reviews and social posts. It also analyzes support tickets and survey answers. If you're asking what is sentiment analysis, you probably saw it as a score in a brand report. That score turns thousands of opinions into one number you can track.
Sentiment analysis defined simply
Sentiment analysis means teaching a machine to tell how a writer feels about something. The machine reads "The battery died in two hours" and marks it negative. It reads "Setup took five minutes" and marks it positive. More detailed tools also tag emotions such as anger or joy. Still, the basic sentiment analysis meaning is polarity: which side of the line an opinion falls on.
How sentiment analysis works
Sentiment analysis breaks text into words and phrases, then scores them with a word list or a trained model. Lexicon tools look up each word in a dictionary of scored terms and add up the results. Machine learning models learn from thousands of labeled examples, so they handle phrases like "not bad" better. Large language models such as ChatGPT can do sentiment analysis from a plain prompt and handle sarcasm and mixed opinions better than a word list. Whatever tool you use, check a sample by hand, because accuracy drops on slang and sarcasm.
Sentiment scoring scale
Sentiment tools report results on one of two kinds of scale: a continuous polarity range or a set of labeled buckets.
Polarity score from -1 to +1
Most lexicon and model-based tools give a number between -1 and +1. A score of -1 is fully negative, and +1 is fully positive. Values near 0 count as neutral. The width of that neutral band is a setting, and vendors set it differently. So two tools can label the same sentence in different ways. Dashboards rescale the range to 0 to 100.
Labels and star-style ratings
Other tools assign a positive or negative class instead of a number. Some also use neutral and mixed classes. Fine-grained versions use five steps, from strongly negative to strongly positive, which map neatly onto star reviews. Labels are easier to read in a report. But they hide intensity, because a mild complaint and a furious one land in the same bucket.
What a sentiment score means
A sentiment score shows the balance of opinion across a set of texts at one point in time. Say 2,000 reviews average +0.3. Positive comments outweigh negative ones, though a loud, angry minority can hide inside that average. The trend tells you more than any single reading. If the score falls from +0.4 to +0.1 after a product change, go read the comments behind the drop. Only compare scores from the same tool, since each one is calibrated differently.
Why sentiment data gets cited
Sentiment data gets cited because it turns opinion into a figure that reports and AI answers can quote. A journalist can write that reviews turned sour, and a score gives that claim something to stand on. AI assistants do something similar when they sum up how people feel about a brand, and they draw on the reviews and forum posts they've read. So what the public writes about you shapes what an answer engine repeats.
How marketers use sentiment analysis
Marketers use sentiment analysis to measure how people react to a brand or a campaign.
Before you report these numbers to leadership, confirm the tool understands your industry's vocabulary.
Sentiment analysis versus social listening
Sentiment analysis is one technique. Social listening is the wider practice of tracking and responding to what people say about a brand online. A social listening tool collects mentions and counts their volume. It also finds trending topics, while sentiment analysis labels the tone of the collected mentions. You can also run sentiment analysis on data that never touched social media, such as call transcripts.
Real-world sentiment analysis example
Picture a coffee chain that launches an oat milk latte and analyzes 5,000 posts about it in the first week. The overall score sits at +0.2, which looks fine. Split by topic, though, taste scores positive while price scores strongly negative. So the team keeps the recipe and tests a smaller, cheaper size.
Related terms
Terms used alongside sentiment analysis cover different parts of the same work.
Next, pick one tool and check 50 of its labels by hand. If they match your own judgment most of the time, the scores are worth tracking. That check gives you a working answer to what is sentiment analysis for your own data.
Sentiment is one of the four metrics in our AI visibility guide. Our AI search visibility tool records it weekly, and a competitor analysis in AI search shows how it compares with rivals.