Social Media Sentiment Analyzer — AI Emotion & Brand Health

Analyze public social sentiment for any keyword, hashtag, or brand across Reddit, YouTube, X/Twitter, and Instagram in one Apify run. It accepts a search query plus optional platform, time window, engine, and credential settings, then returns one aggregate brand-health record and one record per analyzed post or comment. The useful fields include sentiment, sentiment score, confidence, emotions, aspects, language, toxicity, source URL, author, title, engagement, and timestamps. The outcome is a structured sentiment snapshot plus an AI brand-health report with overall score, headline, summary, sentiment split, praise themes, complaints, crisis signals, and recommendations.

Best fit and connected workflows

This Actor fits workflows where a search term needs to become a cross-platform sentiment summary with evidence behind it. It connects naturally to:

• social listening for products, campaigns, competitors, or public topics
• launch and campaign tracking across Reddit, YouTube, X/Twitter, and Instagram
• brand-health review for teams that want both per-post detail and an aggregate report
• AI agent workflows that need one tool call for structured sentiment analysis
• moderation and brand-safety checks when includeToxicity is enabled

For related enrichment after you identify relevant creators or audiences, pair this Actor with Social Media Influencer Scraper.

Example scenario

Maya is a brand analyst preparing a weekly check on Tesla. She starts with query: "Tesla", selects Reddit and YouTube, and keeps the default 7-day window. The Actor returns the report record with brandHealthScore, headline, summary, topPraise, topComplaints, and crisisSignals, followed by post records with platform, sentiment, sentimentScore, confidence, emotions, language, and sourceUrl. Maya sees representative praise and complaint themes in the report, then uses the source URLs to review the original public posts before sharing the findings with her team.

Input

The required input is a search query. Platform selection, time range, engine choice, report generation, and optional credentials refine how the Actor searches and scores public content.

Field	Type	Description
query	string	Search term, hashtag, brand, or topic to analyze. Defaults to ChatGPT.
platforms	array	Platforms to analyze: reddit, youtube, twitter, instagram. Defaults to Reddit and YouTube.
maxItems	integer	Hard cap on total analyzed posts or comments across selected platforms. Range 10-1000.
timeRange	string	Time window: 24h, 7d, 30d, or all.
engine	string	llm for AI scoring or lexicon for fast polarity-only scoring.
llmProvider	string	Preferred provider when using llm: anthropic or openai.
llmApiKey	string	Optional bring-your-own key for the selected LLM provider.
includeToxicity	boolean	Adds a per-post toxicity score when the AI engine is used.
includeReport	boolean	Generates the aggregate AI brand-health report.
responseFormat	string	concise or detailed per-post output shape.
xAuthToken	string	X auth_token cookie for X/Twitter runs.
xCt0	string	X ct0 cookie for X/Twitter runs.
instagramSessionCookies	string	Instagram session cookies for Instagram runs.
maxCostUsd	number	Cost ceiling for this Actor's own pay-per-event charges.

Focused input example

{
  "query": "ChatGPT",
  "platforms": ["reddit", "youtube"],
  "maxItems": 100,
  "timeRange": "7d",
  "engine": "llm",
  "llmProvider": "anthropic",
  "includeToxicity": true,
  "includeReport": true,
  "responseFormat": "detailed",
  "maxCostUsd": 0.5
}

Output

The dataset starts with one report record, followed by individual analyzed post records. The report captures the aggregate view, while post records preserve the evidence behind that summary.

Field	Type	Description
recordType	string	report for the first record, post for each analyzed item.
_summary	boolean or null	True on the report record.
query	string or null	The analyzed search query.
brandHealthScore	number or null	Overall 0-100 health score on the report record.
headline	string or null	One-line AI headline on the report record.
summary	string or null	Plain-English narrative summary on the report record.
sentimentBreakdown	object or null	Aggregate sentiment counts and percentages.
emotionDistribution	object or null	Aggregate emotion counts.
topPraise	array or null	Top positive themes with example quotes.
topComplaints	array or null	Top negative themes with example quotes.
crisisSignals	array or null	Detected reputational risks with severity and evidence.
recommendations	array or null	Suggested next steps from the AI report.
platformCounts	object or null	Posts analyzed per platform.
platform	string or null	Source platform on post records.
sourceUrl	string or null	Direct URL to the post or comment.
author	string or null	Author handle or username.
text	string or null	The analyzed post or comment text.
title	string or null	Parent post or video title, where available.
publishedAt	string or null	ISO 8601 timestamp when available.
engagement	object or null	Engagement metrics such as likes, comments, shares, score, or upvotes.
sentiment	string or null	positive, neutral, negative, or mixed.
sentimentScore	number or null	Sentiment score from -1.0 to +1.0.
confidence	number or null	Confidence score from 0 to 1.
emotions	array or null	Detected emotions for LLM runs.
aspects	array or null	Aspect-based sentiment for LLM runs.
language	string or null	Detected ISO 639-1 language code for LLM runs.
toxicity	number or null	Toxicity score from 0 to 1 for LLM runs when enabled.
warnings	array or null	Non-fatal warnings, such as a platform being skipped.
scrapedAt	string or null	ISO 8601 timestamp when the record was produced.

Illustrative output record

{
  "recordType": "post",
  "platform": "reddit",
  "sourceUrl": "https://www.reddit.com/r/example/comments/abc123",
  "author": "sample_user",
  "text": "I like the new release, but the pricing is still a concern.",
  "title": "Discussion thread about the launch",
  "publishedAt": "2026-06-01T10:30:00Z",
  "sentiment": "mixed",
  "sentimentScore": 0.1,
  "confidence": 0.92,
  "emotions": ["anticipation", "trust"],
  "aspects": [
    {
      "aspect": "release",
      "sentiment": "positive"
    },
    {
      "aspect": "pricing",
      "sentiment": "negative"
    }
  ],
  "language": "en",
  "toxicity": 0.02,
  "scrapedAt": "2026-06-01T10:35:00Z"
}

How it works

This Actor orchestrates dedicated scrapers for the selected platforms, then normalizes their results into one schema. It applies the chosen sentiment engine to each post, with llm providing sentiment, emotion, aspects, language detection, and optional toxicity, and lexicon providing a fast polarity-only score. When includeReport is enabled, it also produces one aggregate AI brand-health record with overall score, breakdowns, themes, crisis signals, and recommendations. The first dataset item is the report record, followed by the post records.

Pricing

This Actor uses Pay per event plus Apify platform usage. Its charged events are:

• Actor Start
• Post analyzed
• AI brand-health report

The primary event is Post analyzed, which is charged for each analyzed post or comment. The report event is charged once per run when the aggregate report is generated. Apify platform usage is billed separately from these event charges. For the current live values and any updates, open the Pricing tab on the Actor page.

For example, a run that analyzes one hundred posts and generates the report uses one start event, one hundred post events, and one report event. That means the Actor charges cover the start charge, one hundred per-post charges, and one report charge, plus separate Apify platform usage.

Use with AI agents (MCP)

This Actor is usable through Apify MCP. It exposes a structured tool for turning a search query into a brand-health report and a dataset of analyzed posts that an AI agent can inspect, summarize, compare, or cite.

Exact Actor identity: khadinakbar/social-media-sentiment-analyzer

Tool description: analyze public social sentiment for a keyword, hashtag, or brand across Reddit, YouTube, X/Twitter, and Instagram, then return per-post sentiment, emotion, aspects, language, toxicity, and an AI brand-health report.

Analyze the public sentiment around "ChatGPT" across Reddit and YouTube for the last 7 days. Return the brand-health report first, then list a few representative posts with sentiment, emotions, and source URLs.

Output interpretation:

• The first dataset record is the aggregate report (recordType: "report").
• Remaining records are analyzed posts or comments (recordType: "post").
• Use sentimentScore and confidence for machine comparison, and sourceUrl for provenance.
• warnings captures non-fatal platform notes, such as a selected source being skipped.

Scope, pagination, and cost guidance:

• query defines the topic searched across selected platforms.
• platforms controls which sources are included.
• maxItems caps the total analyzed posts or comments.
• includeReport adds the aggregate AI report.
• responseFormat: "concise" is the lighter option when an agent needs compact per-post records.
• maxCostUsd can bound this Actor's own pay-per-event charges.

Run via the Apify API

JavaScript

import { ApifyClient } from 'apify-client';

const client = new ApifyClient({
  token: process.env.APIFY_TOKEN,
});

const run = await client.actor('khadinakbar/social-media-sentiment-analyzer').call({
  query: 'Tesla',
  platforms: ['reddit', 'youtube'],
  maxItems: 50,
  timeRange: '7d',
  engine: 'llm',
  includeReport: true,
  responseFormat: 'detailed',
});

const { items } = await client.dataset(run.defaultDatasetId).listItems();
console.log(items[0]);

Python

from apify_client import ApifyClient
import os

client = ApifyClient(os.environ["APIFY_TOKEN"])

run = client.actor("khadinakbar/social-media-sentiment-analyzer").call(run_input={
    "query": "Tesla",
    "platforms": ["reddit", "youtube"],
    "maxItems": 50,
    "timeRange": "7d",
    "engine": "llm",
    "includeReport": True,
    "responseFormat": "detailed",
})

items = client.dataset(run["defaultDatasetId"]).list_items().items
print(items[0])

Best results and outcome guidance

For a broad brand snapshot, start with a short search term and the default Reddit plus YouTube sources. For a launch or crisis watch, use timeRange: "24h" or timeRange: "7d" and keep includeReport enabled so the aggregate record captures the overall direction plus the strongest themes. When you want a compact agent-friendly payload, set responseFormat: "concise"; when you want richer review data for dashboards, use responseFormat: "detailed". If you are comparing runs, keep maxItems consistent so the dataset shape stays easier to compare.

Design note

I found that the dataset contract puts the aggregate report first and marks it with recordType: "report" and _summary: true, while each analyzed item uses recordType: "post". That visible contract makes it straightforward to separate the executive summary from the evidence layer in downstream workflows.

FAQ

Which input should I use for a brand check?
Use a keyword, hashtag, or brand name in query. The Actor searches public posts and comments across the selected platforms.

How does this Actor split results across platforms?
maxItems sets the total cap across all selected platforms, and the budget is divided across the chosen sources.

What if I only want post-level sentiment?
Set includeReport to false to receive only analyzed post records.

Which platforms work without credentials?
Reddit and YouTube work without platform credentials. X/Twitter and Instagram use encrypted session cookies when selected.

How should I read the first dataset item?
Treat it as the aggregate report. It summarizes sentiment, emotions, themes, risks, and recommendations for the full run.

Related Actors

• Social Media Influencer Scraper - use this after sentiment analysis when you want to enrich creator or audience research from a relevant social source.

Responsible use

Use this Actor for analysis of publicly available social content in line with platform terms and applicable law. Provide only credentials you are authorized to use. Treat sentiment, emotion, toxicity, and crisis signals as automated analytical outputs that support review and decision-making. When sharing results, preserve the source URLs and context of the original posts.