Best Social Media Scrapers (2026) + What's Legal to Collect
Eight social media scrapers we ran or verified for October 2026, with real prices, live-test results, and the legal lines on public data, logins, GDPR.
On this page
Every social media scraper listicle I read this month recommends tools that quietly died. Nitter got a cease-and-desist from X Corp in August 2026 and went offline the next day. Reddit’s free .json endpoints started returning 403s in May. X discontinued its free API tier in February. So this list has dates on it: eight tools we ran or verified against live platforms in October 2026, what each one still reaches, what it costs, and where the legal lines actually sit.
Disclosure first: we build Magpie. It covers three social platforms, and none of the five you’re probably thinking of. This article covers the rest.
What a social media scraper can reach in 2026
The difficulty is not uniform. Hacker News is genuinely open: a public API, no login, no games. Instagram and LinkedIn serve some data to logged-out visitors and fingerprint everything else. Facebook hides nearly everything behind its wall. The table is the map as of mid-October 2026, from our own runs and dated repo checks:
| Platform | Reachable logged out? | Route as of Oct 2026 | Main risk |
|---|---|---|---|
| Hacker News | Yes, everything | Algolia HN API; magpie vertical runs it locally in ~3 s |
None. Genuinely open. |
| YouTube | Yes, almost everything | yt-dlp; watch-page player response | Breaks when YouTube changes internals |
| Posts: no longer | Arctic Shift dumps; OAuth apps (closed to new ones) | Platform hostility: .json 403s, API winding down |
|
| Public profiles and posts | Instaloader, no login | No keyword search; internals rotate | |
| TikTok | Public videos/profiles | TikTokApi + headless browser + proxies | Empty responses once flagged |
| X (Twitter) | No. Login wall since 2023 | Official pay-per-use API; twscrape with your accounts | Account bans; X litigates (Nitter C&D) |
| Some public pages | Managed actors; session automations | LinkedIn litigates; account bans | |
| Almost nothing | Managed actors (Apify, Bright Data) | Heaviest anti-bot; Meta sues |
Two dates matter more than the rest. Reddit’s unauthenticated .json endpoints have returned HTTP 403 since late May 2026, and its public Data API is closing to new apps; existing apps had to register by September 30, 2026. If a tutorial still tells you to append .json to a subreddit URL, it was written before that.
Social media data extraction tools, compared
| Tool | Platforms | Price (checked Oct 2026) | How it gets data | The catch |
|---|---|---|---|---|
| Apify | IG, TikTok, X, LinkedIn, YT, Reddit, more | Free tier with $5 credit; Starter $19/mo; Scale $199/mo; Business $999/mo | Hosted actors, one per job | Actors break when platforms change; you pay per run |
| PhantomBuster | LinkedIn, Instagram, X, TikTok | $69 / $159 / $439 per month by tier | Cloud automations drive your own logged-in session | Your account can be banned |
| Bright Data | All major platforms | Per-record pricing, custom quotes | Proxy network + scraper APIs + ready datasets | Enterprise pricing; compliance work is yours |
| Instaloader | Free, MIT | Logged-out profile/post reads | No keyword search across Instagram | |
| yt-dlp | YouTube | Free (public domain) | Player response + innertube endpoints | Needs frequent updates; metadata-first mindset |
| twscrape | X | Free, needs your X accounts | Account cookies, rotates a pool | Accounts get rate-limited and banned |
| Arctic Shift | Free | Monthly dumps + download API | Batch jobs, not live queries | |
| Magpie | Reddit, Hacker News, YouTube | CLI free (AGPL-3.0-or-later / commercial); Desktop $49/mo after 5 runs | Local binary, zero-LLM extractors | Three social platforms; no login-walled sites |
Ranking logic: managed platforms first (they cover the most platforms), then free tools by platform, then Magpie last because its scope is the smallest, stated up front. Best is a function of which platform you’re hitting.
How we picked these social media scraping tools
Three rules. We ran every tool we could from this machine and verified the rest against project repos, vendor pricing pages, and dated announcements from September and October 2026. Release dates below come from those repos. And every entry keeps the cons that would actually stop a purchase, because “cons: expensive” is not a con, it’s a price tag.
The managed platforms
1. Apify: best coverage if you don’t want to run anything
Apify is an actor marketplace: someone maintains an Instagram actor, a TikTok scraper, a LinkedIn jobs actor, and you rent runs. The free plan includes $5 of platform credit each month; paid tiers run $19, $199, and $999 per month, each with its credit rolled in. At recent free-tier rates, $5 covered roughly 1,850 Instagram results.
The appeal is real: you never touch a proxy, a fingerprint, or a broken selector at 2 a.m. The catch is structural. When a platform ships a redesign, the actor breaks and you wait for its maintainer, because it’s still their actor, not your code. Costs also scale with success: a growing job costs more every month, forever. We compared the flat-price and local alternatives in our Apify alternatives piece.
Pros: widest platform coverage in one account; actors for niche targets; scheduling built in; free credit to test.
Cons: per-run pricing compounds; actor quality varies; your pipeline depends on strangers’ maintenance habits; ToS risk transfers to you, not away from you.
2. PhantomBuster: best for LinkedIn and Instagram sales automation, highest account risk
PhantomBuster runs cloud “Phantoms” that automate actions through your own logged-in sessions: connection requests, profile scrapes, follower exports. Plans cost $69, $159, and $439 per month depending on execution hours.
This is the tool people mean when they ask about “LinkedIn scrapers that actually work,” and it does work. For a while. It works because it pretends to be you. That is also the whole risk: LinkedIn and Instagram explicitly forbid automation on your account, and the ban lands on the account you use for real work. If the account is your livelihood, read our LinkedIn scraping tools roundup and the LinkedIn legality guide before you point a Phantom at it.
Pros: reaches LinkedIn data nothing logged-out can; deep sales-workflow recipes; no infrastructure to run.
Cons: one ToS strike from losing a real account; hourly execution limits; serious volume means the pricier tiers.
3. Bright Data: best for enterprise volume and ready-made datasets
Bright Data sells the infrastructure layer: proxy networks, unblocker tech, per-platform scraper APIs, and fixed-price datasets (an entire Instagram profiles dataset, delivered). It is the vendor other vendors quietly resell.
Meta sued Bright Data in 2023 and lost the logged-out contract claim, which we cover below; the litigation tells you the volume tier this company operates at. Pricing is custom, which in practice means four figures a month and a compliance conversation. If you need ten million records with a delivery SLA, this is the shortlist. If you need a thousand, it’s the wrong tool.
Pros: survives anti-bot at scale; datasets with schema guarantees; serious legal/compliance team on their side of the fence.
Cons: opaque pricing; overkill below enterprise volume; the data’s legal posture in your use case stays your problem.
The free social media scraper shortlist
4. Instaloader: Instagram public profiles without a login
Instaloader is the last major open-source Instagram tool that works logged out. Version 4.15.3 shipped July 26, 2026, the repo has about 13,600 GitHub stars and an MIT license, and it pulls profiles, posts, captions, comments, and geotags from public accounts. Login is only required for private profiles and stories.
Its limit is the one most monitoring jobs actually need: no keyword search across Instagram. You can point it at profiles; you can’t ask “who posted about X this week.” For that, you’re back to the managed platforms. The logged-in alternative, instagrapi, reaches far more and puts your account on the line; its own docs call the private-API approach fragile in production.
Pros: no account needed for public profiles; media + metadata in one pass; actively maintained (pushed October 2026).
Cons: profile-by-profile only; Instagram rotates internals without notice; downloading en masse at speed will get your IP throttled.
5. yt-dlp: YouTube metadata is basically solved
yt-dlp is a downloader, but -J hands you the full metadata JSON: title, channel, view counts, description, chapters, subtitles. It’s public-domain code, it updates within days whenever YouTube shifts internals, and it handles the official API’s biggest weakness: the Data API caps you at 10,000 quota units a day, roughly 100 detailed video lookups.
Run yt-dlp -J "https://www.youtube.com/watch?v=dQw4w9WgXcQ" and you get structured JSON back in a couple of seconds. Comments and transcripts are the gaps, and both have their own deep dives: YouTube scraping, comment exports, and transcripts in bulk.
Pros: free, public domain, huge install base; metadata + media in one tool; fastest fixes when YouTube breaks scrapers.
Cons: rate limits on bulk channel walks; comments need other routes; it’s a downloader first, so pipeline code around it is on you.
6. twscrape: the working X scraper, with the account risk stated plainly
After X closed its guest API in 2023, every logged-out route died. snscrape’s X modules stopped working that year, and Nitter’s mirrors got a cease-and-desist in August 2026. What still works: twscrape, v0.20.1 released August 25, 2026, which logs in with a pool of X accounts (their auth_token and ct0 cookies), rotates when one hits a rate limit, and reads timelines and search through the same endpoints your browser uses.
The honest framing: this violates X’s terms around automation and multiple accounts, and accounts get banned. For a throwaway research account pool, that’s your call. For anything customer-facing, the official API is the compliant path: pay-per-use since February 2026, about $0.005 per post read, free tier gone. Full-archive search stays behind enterprise contracts.
Pros: free at the software level; async, pool-aware; the only maintained open-source X reader.
Cons: burns accounts; ToS-violating by design; X has shown (Nitter) it will enforce.
7. Arctic Shift: bulk Reddit history after the .json lockdown
With Reddit’s unauthenticated .json gone, its public Data API closing to new apps, and Pushshift restricted to moderators since 2023, bulk Reddit research moved to Arctic Shift: monthly dumps (the July 2026 dump shipped August 9), an API, and a web tool that downloads one subreddit or user at a time. It runs logged out and it’s free.
Scope is the trade: dumps are batch jobs you schedule and process yourself, and full-text search runs per subreddit or per user, not across all of Reddit. The old Academic Torrents archive of everything through 2025 is being taken down at Reddit’s request, with partial mirrors on Hugging Face. For live monitoring instead of bulk history, see our Reddit scraping guide and subreddit monitoring how-to.
Pros: free; monthly cadence you can build a pipeline on; no Reddit account involved.
Cons: nothing live; per-community search only; the archive landscape is shrinking, not growing.
8. Magpie: local, free, and honest about covering three platforms
Magpie is ours, so weigh this entry accordingly. It’s a local Go binary with zero-LLM extractors for 27 site types, three of them social: Reddit, Hacker News, and YouTube. No cloud service, no account, and the data never leaves your machine. No LinkedIn, no Instagram, no Facebook, no X, no TikTok. It doesn’t bypass logins or CAPTCHAs by design. If your job is the big five, this is not your tool, and the rest of this list exists for that reason.
On its three platforms it’s fast and boring in the good way. A Hacker News item through the Algolia-backed extractor:
$ magpie vertical "https://news.ycombinator.com/item?id=8863" --name hackernews
{
"record": {
"author": "dhouston",
"comments": 32,
"kind": "story",
"points": 104,
"title": "My YC app: Dropbox - Throw away your USB drive"
}
}(Abridged. The story is from 2010; the API returns sixteen-year-old threads as cleanly as today’s.) A YouTube video returns title, channel, duration, description, and view count: 1.83 billion views on the one we tested this week.
The Reddit vertical is where we’ll be straight about failure modes. Run it tokenless from a home connection right now and Reddit serves a login-wall page; the extractor faithfully returned it, with the title “Bienvenue sur Reddit” inside otherwise-valid JSON. Nothing crashed. You just got the wall. Check the title field before trusting a tokenless Reddit run, or set MAGPIE_REDDIT_CLIENT_ID/SECRET/UA to switch it to Reddit’s app-only OAuth. robots.txt is respected, requests are throttled per host, and there is no “stealth mode,” because a tool that defeats platforms’ blocks is a different legal posture than one that reads what’s public.
Pros: CLI free (AGPL-3.0-or-later or commercial license); desktop app has a 5-run free trial, then $49/month; everything runs locally; same store shared with the desktop app’s history.
Cons: three social platforms; tokenless Reddit currently walled on many IPs; no hosted API, it’s a binary you run.
Social scraping tools that died (and what killed them)
The graveyard is the most useful part of this research, because the causes of death repeat:
- Nitter and XCancel: offline August 25, 2026, one day after X Corp’s cease-and-desist cited the Texas Harmful Access by Computer Act and the Lanham Act. Public mirrors of login-walled content are enforceable targets.
- snscrape (X modules): untouched since November 2023; the endpoints it relied on closed behind X’s login wall. Still installed by tutorials. Still doesn’t work for X.
- Twint: archived March 2023, read-only long before that.
- Pushshift: moderator-only since May 2023, after Reddit’s API pricing change.
- Academic Torrents Reddit dumps: 3.5 TB of history, being removed at Reddit’s request as of July 30, 2026.
The pattern: anything that republishes or mirrors a platform’s data at scale attracts legal pressure; anything depending on undocumented endpoints breaks within months of a platform update. Tools that read public pages politely and tools backed by official APIs survive. That asymmetry should shape what you build on.
Is social media data scraping legal?
Plain-English summary, not legal advice: jurisdiction facts vary and this shifts yearly. The deep versions live in is web scraping legal, the CFAA explainer, and the terms-of-service guide.
Reading public pages is not a federal computer-crime issue in the US. Van Buren (2021) narrowed the Computer Fraud and Abuse Act, and the Ninth Circuit’s hiQ v. LinkedIn decision (2022) held that scraping public data isn’t “unauthorized access” under it. That’s why logged-out scraping is the defensible default.
Contract law still bites. In Meta Platforms v. Bright Data (N.D. Cal., January 23, 2024), Judge Edward Chen granted Bright Data summary judgment on Meta’s breach-of-contract claim: Meta’s terms govern people using the site, and a logged-out scraper “stands in the same shoes” as any visitor the terms can’t bind. Meta’s tortious-interference claim continued. Read together with hiQ, the practical line is the login wall. Data visible logged out is the safe zone; scraping through an account means you accepted terms that ban automation.
Personal data carries its own rules. GDPR doesn’t carve out public data: a scraped profile full of names and photos is personal data, with purpose limits and retention duties attached. If EU residents are in your dataset, read scraping and GDPR before designing storage, and the copyright explainer before republishing any of it. Posts belong to their authors.
Platforms are getting more aggressive, not less. X’s C&D against Nitter used a state “harmful access” statute, not just contract. Reddit is closing free routes on a schedule. Build on the routes with a legal footing: official APIs where they exist, logged-out public data where they don’t.
FAQ
Which social media scrapers still work without an account?
Instaloader (public Instagram profiles), yt-dlp (YouTube), Arctic Shift (Reddit history), and any Hacker News reader. For X, no account-free route survived 2026; twscrape and Twikit both need your accounts.
What are the best scraping tools for social media in 2026?
By platform: Instaloader for Instagram, yt-dlp for YouTube, twscrape for X, Arctic Shift for Reddit bulk history, and Apify if you want one vendor for everything and don’t mind per-run pricing. Magpie if you want local and free on Reddit, HN, and YouTube.
Does social media data extraction violate GDPR?
Not automatically, but the data’s “public” status is irrelevant to GDPR. Names, photos, and posts of EU residents are personal data; you need a lawful basis, a defined purpose, and retention limits. Selling or repurposing scraped profiles is where companies get fined.
Can you get banned for social media scraping?
Your accounts can: PhantomBuster, twscrape, and instagrapi all work by driving logged-in sessions, which platforms explicitly forbid. Your IP can be throttled or blocked on most platforms. Lawsuits target commercial-scale republishers far more often than researchers.
Is it legal to scrape LinkedIn or Facebook?
Logged-out public data is the defensible zone per hiQ and Meta v. Bright Data, but both platforms litigate aggressively and most valuable data sits behind login. Treat LinkedIn scraping as a legal-review decision, not a tooling decision.
Where to start
Match the route to the platform and stay logged out where you can. YouTube: yt-dlp, today, free. Instagram public profiles: Instaloader. Reddit history: Arctic Shift dumps. X: pay the API’s half-cent per read or accept real account risk with twscrape; there’s no third option anymore. Whole-market coverage without infrastructure: Apify’s $19 Starter tier. Not sure which shape you need? The best web scraping tools comparison sorts the wider field. Reddit, HN, and YouTube locally with zero LLM cost: grab the binary from the repo, and read the Hacker News walkthrough for a five-minute first run.
Whatever you pick, re-verify it next quarter. Half the tools in the 2025 editions of this list are gone.
