Threads Search Scraper

Search Threads by keyword, phrase or hashtag and get the matching posts with their text, author, likes, replies, reposts, media and links. Filter by date, author, likes, language, media and more, choose between the newest posts and the most popular ones, and use monitor mode to get only what is new since the last run.

No login, cookies or API key needed. The actor talks to Threads over plain HTTP — no browser — so 1,000 posts take about a minute.

Features

• 🔎 Keyword, phrase and hashtag search — openai, world cup, #photography. Run many searches at once.
• 🕒 Newest first or most popular first — follow a topic as it happens, or find the posts that got the most attention.
• 📅 Date range — only posts between two dates, or from the last 7 days.
• 👤 Author filter — only posts written by one account.
• 🎛 Post filters — minimum likes, replies or reposts; original posts or replies only; text, photo or video posts; verified authors; language; words to exclude.
• 🔔 Monitor mode — a scheduled run returns only posts it has not returned before, so you get (and pay for) each post once.
• 🌐 Language of every post — as a two-letter code, ready to filter or group by.
• 📊 Engagement — likes, replies, reposts, quotes and shares for every post.
• 🖼 Media and links — full-size photos and videos, web links and link cards.
• 💬 Conversation context — replies come with the post they answer, quote posts with the post they quote.
• 🎯 On-topic results only — Threads sometimes pads a search with loosely related posts; the actor keeps only posts that really contain your search words.
• ♾ Goes past the first page — Threads shows a single page per search; the actor walks back through time to collect as many posts as you ask for.
• 💾 Incremental results — posts are saved as they are found, so a long run never loses what it collected.

Use cases

• Brand and competitor monitoring
• Trend and sentiment research
• Finding creators and conversations in a niche
• Tracking what an account says about a topic
• Datasets for analysis and AI

Quick start

1. Enter a search, e.g. openai.
2. Set how many posts you want.
3. Run. Posts appear in the dataset as they are found.

Input

Field	Type	Description
searchQueries	array	Required. Words, phrases or hashtags to search for. Each one is scraped separately.
maxResults	integer	Posts to return per search (1–10,000). Default 100.
sortBy	string	recent (newest first, default) or top (most popular first).
startDate	string	Only posts on or after this date: yyyy-mm-dd (UTC), or a period such as 7 days.
endDate	string	Only posts on or before this date: yyyy-mm-dd (UTC), or 2 days for "at least two days old".
fromUsername	string	Only posts by this account, e.g. zuck.
minLikes / minReplies / minReposts	integer	Only posts with at least this many. Default 0.
replies	string	include (default), exclude (original posts only) or only (replies only).
media	string	any (default), text, media (photo or video), image or video.
verifiedOnly	boolean	Only posts from verified accounts.
languages	array	Only posts in these languages, as two-letter codes: en, es, pt, fr, de, it, nl, pl, tr, id, vi, ru, uk, ar, fa, he, hi, bn, th, el, ja, ko, zh.
excludeWords	array	Leave out posts containing any of these words or phrases.
onlyNewPosts	boolean	Monitor mode: return only posts that earlier runs of the same search did not return.
proxyConfiguration	object	Proxy settings. The default is recommended.

Example

{
    "searchQueries": ["openai", "#photography"],
    "maxResults": 500,
    "sortBy": "recent",
    "startDate": "7 days",
    "minLikes": 10,
    "replies": "exclude",
    "languages": ["en"],
    "excludeWords": ["giveaway"]
}

Monitoring a topic

Turn on onlyNewPosts and put the actor on a schedule — every hour, say. The first run returns the latest posts; every later run returns only what appeared since. What has been returned is remembered per search (and per author filter) in a key-value store named threads-search-scraper-seen in your account; delete it to start over.

Output

One dataset item per post:

{
    "searchQuery": "openai",
    "id": "3740123456789012345",
    "url": "https://www.threads.com/@dev.bob/post/DeZSk4cAf-8",
    "text": "OpenAI just changed its pricing: https://example.com/pricing",
    "createdAt": "2026-10-12T12:59:51.000Z",
    "author": {
        "id": "1002",
        "username": "dev.bob",
        "fullName": "Bob Builder",
        "isVerified": true,
        "profilePicUrl": "https://scontent.cdninstagram.com/...",
        "profileUrl": "https://www.threads.com/@dev.bob"
    },
    "likeCount": 4,
    "replyCount": 1,
    "repostCount": 2,
    "quoteCount": 1,
    "shareCount": 5,
    "mediaType": "text",
    "media": [],
    "links": ["https://example.com/pricing"],
    "linkPreview": {
        "url": "https://example.com/pricing/",
        "title": "Pricing | Example",
        "displayUrl": "example.com",
        "imageUrl": "https://external.xx.fbcdn.net/..."
    },
    "hashtags": [],
    "mentions": [],
    "topic": "AI Threads",
    "isReply": true,
    "replyTo": {
        "id": "3740123456789000000",
        "url": "https://www.threads.com/@garden.anna/post/DeSL91Ekj1M",
        "username": "garden.anna",
        "text": "Which AI assistant do you use for coding?",
        "createdAt": "2026-10-12T12:46:40.000Z"
    },
    "quotedPost": null,
    "isQuote": false,
    "isEdited": false,
    "language": "en",
    "isAiLabeled": false,
    "isPaidPartnership": false,
    "replyAudience": "everyone",
    "code": "DeZSk4cAf-8",
    "scrapedAt": "2026-10-12T17:10:45.538Z"
}

mediaType is text, image, video or carousel. Each entry in media has a type, a direct url, and for videos a thumbnailUrl. Media links are issued by Threads and expire after a while — download the files soon after the run if you need to keep them.

Good to know

• Every result contains your search words. A post is kept when all the words of your search appear in its text, hashtags, topic, link title, author name or the post it quotes. A word that only appears inside a web address does not count, unless you search for an address such as openai.com. Several words are searched as a phrase, as Threads does, so world cup finds fewer posts than worldcup and world would separately.
• Busy topics are sampled, not exhausted. For each hour Threads gives its newest posts and its most popular ones — up to about 30 posts per hour. For a topic with thousands of posts an hour you get a steady sample across time rather than every single post.
• Filters work on what Threads returns. Threads' search has no filters of its own, so the actor fetches posts and keeps the ones that pass. A strict filter on a quiet topic takes longer and can return fewer posts than you asked for; a run stops on its own when thousands of posts in a row fail the filters. You only pay for posts that are saved.
• Language is worked out from the text. Threads labels few posts itself. Very short posts and posts mixing languages get language: null, and a language filter leaves those out.
• Rare topics take longer. The actor checks hour after hour for posts and widens its steps through quiet periods. Over long quiet stretches Threads may leave some posts out; set a start date close to the period you care about for the most complete results.
• Most popular first starts with the posts Threads ranks highest over the whole period, then over ever shorter slices of it. It follows Threads' own "Top" ranking, which is not strictly ordered by likes.
• Order of the dataset. Posts are saved as each slice of time finishes, so the dataset is roughly, not strictly, in order. Sort by createdAt or likeCount if you need an exact order.
• Newest posts. Threads needs a little time to make a new post searchable; posts from the last few minutes may be missing.
• No view counts. Threads does not show them to visitors who are not logged in.
• Proxies. Threads limits how many searches one IP address may run. The actor starts on datacenter proxies and switches to residential proxies by itself when that limit is reached.

Billing

You pay per post saved to the dataset. The run stops on its own when it reaches the maximum charge you set, so you never pay for more than you asked for.

Local development

npm install
npm test          # parser, input, filter, language and paging unit tests
apify run         # runs with the input in storage/key_value_stores/default/INPUT.json