TikTok data: what you can and can’t get in 2026

TikTok is the hardest of the big platforms to get data from legitimately. Instagram at least has Business Discovery for other professional accounts, and YouTube’s official API reads any public channel with an API key. TikTok offers nothing comparable to a business that wants to look up creators who haven’t connected to it.

So most commercial TikTok data comes from third parties. That works, but you need to know what they can return, what they can’t, and how they fail, before you build on them.

TikTok’s own APIs

TikTok for Developers has three read-side products. None of them is a general lookup for public creators.

Display API

Through Login Kit, a user authorises your app and you can read their basic profile (display name, avatar, bio, profile link) and metadata for their own public videos. Rate limits are generous: 600 requests a minute on each of the user-info, video-list and video-query endpoints, with HTTP 429 when you go over. But it only covers accounts that have connected to you. It’s the right tool for a creator-facing app where creators sign in; it’s no use for discovery.

Research API

The one official API that reads arbitrary public accounts: video search, comments, user info, follower and following lists, and liked, pinned and reposted videos. It’s also the one most people can’t use. More on it below.

Commercial Content API

Ads and advertiser data: when an ad ran, its targeting and how many people saw it, searchable by advertiser or keyword. It’s open to the public after an application, which TikTok says it answers within two working days, but the data covers the EU only.

One more 2026 change to note. Since late January, US users contract with TikTok USDS Joint Venture LLC, and TikTok rewrote its US terms and Developer Terms to match. What that means for US developers’ API access in practice isn’t yet clear, so read the current terms before you build on any official product.

The Research API: who gets in, and the catch

Eligibility is narrow. TikTok accepts academic institutions in the US, EEA, UK, Canada and Switzerland, and not-for-profit or independent research bodies in the EU, a service it’s beta-testing with select researchers in the US, UK, Switzerland, Norway, Iceland and Liechtenstein. Non-academic non-profits only get access through a locked-down Virtual Compute Environment. In Brazil it’s limited to academic and non-profit researchers studying youth safety online. You need a research proposal, disclosed funding, ethical review and independence from commercial interests. TikTok says decisions typically take up to four weeks.

If you get in, the limits shape what you can study:

• 1,000 requests and 100,000 records a day, shared across your whole organisation. Video and comment endpoints return 100 records a request.
• Follower and following lists have a separate allowance of up to 2 million records a day across 20,000 calls.
• Stale data. New videos can take up to 48 hours to reach search, and views and follower counts up to 10 days to update.
• Short-lived tokens. Access tokens expire every two hours, and quotas reset at midnight UTC.

The terms, in force since 22 January 2026, add pipeline work: no commercial use, refresh stored data at least every 30 days and delete anything no longer available, and give TikTok your publication at least seven days before it comes out. Older write-ups quote a 15-day refresh; that’s out of date.

In the EU there’s a second route. The Digital Services Act gives vetted researchers a right to request data from very large platforms, TikTok included, under a delegated act in force since 29 October 2025. It’s a separate, more formal process, and TikTok’s research terms include a schedule for vetted researchers.

What third-party APIs return

Because there’s no commercial official route, third-party APIs fill the whole gap. They work in four ways: live APIs that call TikTok when you ask (ScrapeCreators, EnsembleData, Influship, TikHub, LamaTok), scraper platforms that run jobs or sell datasets (Apify, Bright Data), task-based APIs that queue a collection and return it later (Data365), and TikAPI, which also acts on accounts your users connect through OAuth.

Between them, here’s what you can expect:

• Profiles from everyone: name, bio, and follower, following, like and video counts.
• Videos from everyone: captions, views, likes, shares and music. A profile feed isn’t always a complete history; ScrapeCreators says so in its own docs.
• Comments from everyone, paged, with replies from most.
• Follower lists from most, capped at the 10,000 most recent. Bright Data doesn’t sell them.
• Search by keyword, hashtag, user and sound from most. Results depend on the country the request comes from.
• Transcripts from a few: ScrapeCreators (native captions, with an AI fallback), Apify through add-ons, and Influship, which generates one when a video has no captions.
• Live, Shop and ads from some. ScrapeCreators and TikHub cover all three, and Bright Data has Shop data.

Our TikTok ranking has the full coverage table, provider by provider.

What nobody can get you

Some TikTok data isn’t for sale from anyone, whatever a sales page implies:

• History. Views and followers are point-in-time. Nobody sells last month’s numbers, so if you track growth, fetch on a schedule and keep your own timestamps.
• Exact audience demographics. Only the creator sees them in their own account. ScrapeCreators sells an audience-country estimate at 26 credits a profile; treat any figure like that as an estimate.
• Full follower lists for big accounts. The 10,000 cap applies to everyone.
• Private accounts, without the owner’s permission.
• Shop sales. GMV and sales figures are estimates, not data TikTok publishes.

Where it breaks

TikTok is harder to scrape than it looks and changes often. Keeping up with it is most of what you pay a provider for. These are the failures that come up again and again in public bug trackers and status pages:

• Signed requests. TikTok’s web endpoints need signed parameters and device and session tokens that rotate. Open-source wrappers need a real browser to keep up.
• Empty successes. TikTok can answer HTTP 200 with an empty body. The popular TikTok-Api library logged exactly this for user-level queries in July 2026.
• Overnight changes. yt-dlp logged a TikTok page change that broke its extractor between 31 August and 1 September 2026, a short-lived live-stream protocol change in October, and new URL types (short drama episodes, photo slideshows) that older code didn’t recognise.
• IP blocks. TikTok blocks datacenter IPs. On 14 July 2026 EnsembleData’s status page logged TikTok banning a large batch of its IPs, which hit profile and search lookups for about six and a half hours.
• Login-only data. Some comment threads, age-restricted videos and full follower lists need a logged-in session. Providers that run pools of accounts can reach more, at more risk.

You can’t fix TikTok, but you can stop bad responses reaching your database. Check what comes back, not just the status code:

def looks_empty(video: dict) -> bool:
    """TikTok can answer 200 with nothing in it. Check the payload, not the status.
    Field names vary by provider; use the ones yours returns."""
    stats = video.get("stats") or {}
    return not video.get("id") or stats.get("play_count") is None

resp = fetch_video(video_id)          # your provider's client
if resp.ok and not looks_empty(resp.json()):
    save(resp.json(), fetched_at=now())
else:
    retry_later(video_id)             # don't store blanks as zeros

What it costs

Headline prices don’t compare well, because providers bill per call, per page or per record. Work out the cost of 1,000 of the thing you actually need. At published rates:

• ScrapeCreators: credit packs that never expire, $47 for 25,000 (about $1.88 per 1,000 calls) or $0.99 per 1,000 on the $497 pack.
• EnsembleData: a monthly plan with a daily allowance, from $100 for 1,500 units a day. If you use it all, that’s about $0.24 per 1,000 posts and $0.04 per 1,000 followers. Unused units reset at midnight UTC.
• Influship: $1.80 per 1,000 live calls, billed after use with no minimum, like the rest of your infrastructure.
• Apify: API Dojo’s TikTok scraper is a flat $0.30 per 1,000 results, if a job-based scraper suits you.
• TikHub: from $0.001 a request. Read its terms first: they let it sell aggregated data derived from customers’ requests unless you opt out, all sales are final, and its status page was down when we checked.

Then check what failures cost. Influship doesn’t charge for failed calls, EnsembleData doesn’t charge for its own errors, and ScrapeCreators doesn’t charge for empty comment pages. LamaTok, from the team behind HikerAPI, bills 403 and 404 responses, and Bright Data bills records that fail on a bad input. On a platform that fails as often as TikTok, that difference adds up. Our guide to scraping API costs shows how to work out cost per successful request.

The law

TikTok’s US terms ban extracting data with bots or other automated tools without its written approval, so every third-party provider works against that clause. There’s no published US court ruling on scraping TikTok. The closest cases, Meta v. Bright Data (2024) and hiQ v. LinkedIn (2022), suggest scraping public, logged-out pages isn’t unlawful access, while hiQ’s later loss on breach of contract, partly over fake accounts, shows where the line is.

Personal data is the bigger risk. In the EU and UK, GDPR covers public profiles too, and the Dutch regulator says scraping by private companies is almost always unlawful unless it’s very targeted. TikTok’s large under-18 audience raises the stakes. Collect only what you need, keep it only as long as you need it, avoid minors’ data, and prefer providers that work logged-out. This isn’t legal advice.

Which provider to choose

For most developers it comes down to billing, not data. ScrapeCreators covers the most and its credits never expire. EnsembleData is cheapest if your volume is steady every day. Influship is built for production apps and AI agents: billed after use, failed calls free, an MCP server, and a direct line to its founders and solutions engineers whatever you spend. If your users connect their own accounts, TikAPI is the only option here that acts on them.

Questions