---
title: "Insights"
description: "What Speak AI extracts from every recording, what each insight type means, and how to read them without opening the transcript."
---

> Documentation Index
> Fetch the complete documentation index at: https://docs.speakai.co/llms.txt
> Use this file to discover all available pages before exploring further.

# Insights

Every recording is analyzed the moment it's transcribed: keywords, sentiment, entities, and
topics extracted automatically, no prompt required. Read one file's insights on its page, tune the
[categories](/help/insights/categories/) to your domain, set [keyword alerts](/help/insights/keywords/)
for the phrases that matter, define [AI fields](/help/insights/fields/) for structured extraction,
and use [Explore](/help/insights/explore/) to compare it all across your library, where
[themes](/help/insights/themes/) show what recurs and what's changing.

![The Insights tab of Analysis and Data Visualization, summarizing the last 30 days. Recording Activity reports 8 team recordings and 7k words transcribed across 8 speakers, Team Activity lists the account and its recordings, and Sentiment splits the results into positive, neutral and negative with a count and a percentage each.](/help/media/insights/insights-index.jpg)
- **[AI fields](/help/insights/fields/)**
- **[Categories](/help/insights/categories/)**
- **[Explore](/help/insights/explore/)**
- **[Keywords](/help/insights/keywords/)**
- **[Sentiment scores](/help/insights/sentiment/)**
- **[Themes](/help/insights/themes/)**

## Keywords and topics

The terms and subjects mentioned most often in a recording. Read them first to tell what a
conversation was about before you open the transcript. Explore and folder statistics also render
keywords as a [word cloud](/help/exports/word-clouds/), where the largest words are the most
frequent. Topics are extracted from audio and video only.

## Sentiment

Speak scores the emotional tone of your content at both the document level and the individual
sentence level:

- **Positive:** optimistic, supportive, or enthusiastic language
- **Negative:** critical, frustrated, or concerning language
- **Neutral:** factual, informational statements
- **Compound score:** an overall score from -1 (most negative) to +1 (most positive)

[Sentiment scores](/help/insights/sentiment/) explains the bands each score falls into.

## Named entities

Speak identifies and labels the specific things mentioned in your content:

- **People:** names of individuals mentioned
- **Organizations:** companies, institutions, brands
- **Locations:** cities, countries, addresses
- **Products:** product and brand names
- **Dates and times:** specific dates, deadlines, time references
- **Money:** dollar amounts, pricing, financial figures

Entities are useful as a filter. Narrow your library to files that mention a competitor, then read
the sentiment on just those files to see how people talk about them.

## Speaker analytics

For recordings with more than one speaker, Speak tracks:

- **Speaking time:** how long each person spoke
- **Word count:** total words per speaker
- **Words per minute:** speaking pace for each person
- **Speaking percentage:** share of the conversation

## Categories

Speak sorts content into a set of default categories that work with no configuration. You can add
your own keywords to any of them so specific terms are always captured. See
[Categories](/help/insights/categories/) for the full list and the steps.

You can also create categories of your own, such as "Product Feedback", "Action Items", or
"Customer Complaints":

1. Go to your [account settings](https://app.speakai.co/profile/usage)
1. Find **Custom Categories**
1. Add your categories and define what each one means
1. New media is analyzed against your custom categories

Speak can also suggest categories based on the content you already have.

## Where insights appear

- **Media detail page:** the full breakdown for one file
- **Folder statistics:** aggregate insights across every file in a folder
- **[Explore](/help/insights/explore/):** cross-media analytics and trends
- **[AI Chat](/help/ai-chat/):** ask questions about insights, such as "What were the most negative moments?"
- **[Exports](/help/exports/insights/):** include insights in CSV, PDF, and other formats

Analyzing a real dataset? [Book a demo and see your own recordings analyzed](https://calendly.com/speak-ai/demo?utm_source=docs&utm_campaign=book-demo).

Source: https://docs.speakai.co/help/insights/index.mdx