The YouTube Data API is the official programming interface Google offers for software to read and change information on YouTube. Through it, an application can retrieve a video's title, description, tags, duration and public statistics, list the uploads of a channel, search the catalogue, read comment threads, manage playlists and captions, and upload or edit videos on behalf of a signed-in user. It exists because YouTube holds a large share of the world's online video, and the companies building creator tools, analytics dashboards, social listening products and research datasets need a sanctioned, structured way to reach that data rather than copying pages one by one.

The current version, v3, is a REST interface that returns JSON. As of October 2026, Google's machine-readable discovery document for the API, revision 20261006, lists more than 30 resource types, from videos, channels, playlists and search to captions, subscriptions, members and live chat messages.

How a request works

Every call names a resource and a method. A request to videos.list with a video ID returns a "video" resource, and the developer chooses which slices of it to receive through the part parameter: snippet for metadata, statistics for view, like and comment counts, contentDetails for duration and definition, status for privacy and licensing. A separate fields parameter trims the response further.

Access comes in two forms. An API key identifies the project making the request and is enough for public data, such as the metadata of a public video or the uploads of a channel. OAuth 2.0 is required whenever a user's private data or a write is involved. The discovery document lists seven OAuth scopes, including youtube.readonly to view an account, youtube.upload to manage videos, youtube.force-ssl to see, edit and permanently delete videos, ratings, comments and captions, and a youtubepartner scope for assets held by content partners. Reading comment threads and downloading caption tracks both require the force-ssl scope, which explains why third-party transcript tools seldom rely on the official route: a caption file can only be downloaded through the API by an account with rights to the video.

The live streaming functions are not a separate product in technical terms. Resources such as liveBroadcasts, liveStreams and liveChatMessages sit inside the same v3 surface, which Google brands in its documentation as the YouTube Live Streaming API.

Quota, the real constraint

YouTube does not charge money for the API. It rations it. Each Google Cloud project receives a daily allocation measured in units, which resets at midnight Pacific Time, and every request deducts from it, including invalid ones. For years the default was 10,000 units per day. Ordinary list requests cost 1 unit, typical writes and deletes 50, and a search.list call 100. Caption operations are dearer: 50 units to list tracks, 200 to download one, 400 to insert and 450 to update, according to documentation for tuber, an open-source R client for the API.

Those numbers shape what can be built. At 100 units per search, a project on the default allocation could run about 100 searches a day before doing anything else. Video uploads were historically priced at 1,600 units each, limiting a default project to six.

Google has since separated the scarcest operations. YouTube now meters ordinary units, search calls and video uploads in separate buckets, with defaults of 10,000 Data API units, 100 search queries and 100 video uploads per day, according to the tuber 2.0.0 documentation released on August 17, 2026. The arithmetic for search is roughly unchanged; for uploads the ceiling rose sharply.

More quota requires an application and an audit. Google reviews the requesting application against the YouTube API Services Terms of Service and Developer Policies, including how it stores data, displays YouTube branding and handles user consent. Projects can also face compliance audits without asking for more quota, and access can be revoked for violations.

The rules attached to the data

The terms matter as much as the endpoints. Developers must not store most API data for longer than 30 days without refreshing it from the API or deleting it, a requirement intended to stop third-party databases from keeping videos that creators have since removed or made private. Data obtained from a user through OAuth must be deleted when that user revokes access.

Those rules sit on top of YouTube's own Terms of Service, which bar access to the service by automated means other than public search engines following robots.txt or with YouTube's written permission. The API is, in effect, the permitted route, and the alternatives covered in the explainer on web scraping are the unpermitted ones.

Origin and evolution

YouTube opened developer interfaces soon after its 2005 launch, and following Google's acquisition in 2006 they moved to Google's XML-based GData protocol, the basis of what became v2. Version 3, built on Google's newer JSON API infrastructure, was introduced in 2012. On August 23, 2013, YouTube raised the v3 quota tenfold and video upload limits a hundredfold, according to TechCrunch, to pull developers across. The v2 API was retired in stages, with its main feeds switched off on April 20, 2015, breaking embeds and plugins that had not migrated.

Later changes mostly removed data. Public dislike counts disappeared from the API on December 13, 2021, a month after YouTube hid them on its site. Searching for videos related to a given video, through the relatedToVideoId parameter, ended in August 2023. Each removal broke third-party tools that depended on the field.

Why marketers depend on it

Much of the creator economy's software layer runs on this API. Influencer discovery platforms use search, channel and video statistics to rank creators by subscribers, views and engagement. Social listening tools pull comment threads. Brand safety and suitability vendors read titles, descriptions, tags and category data to classify videos before advertisers buy them. Agencies use channel and playlist data for competitive research. When YouTube began counting views from the first frame across all formats on August 24, 2026, social media management vendors including Emplifi, Zoomph and Sprout Social published notes warning clients that the public view counts they ingest would change. PPC Land's explainer on YouTube views traces how the definition has moved, and Podtrac argued the new figure overstates podcast attention by 36%.

Google itself builds on the API. Its open-source YouTube Topic Insights tool pairs YouTube data with Gemini to surface trends for creators and marketers.

Limits, criticisms and disputes

Quota scarcity is the most common complaint. A default project cannot index a meaningful slice of YouTube, and independent researchers argue the audit process favours established companies with compliance teams. The Researcher Program, launched by YouTube in 2022, offers expanded API access to academics at accredited institutions, but critics note that it excludes journalists and civil society groups and is governed by YouTube's own approval.

European law now intervenes. YouTube is a very large online platform under the Digital Services Act (DSA), and Article 40 requires such platforms to give vetted researchers data access. The European Commission adopted a delegated act setting out the procedure on July 2, 2025. Enforcement has teeth: the Commission's 120 million euro fine on X on December 5, 2025 covered terms that barred researchers from scraping, and it issued preliminary findings that TikTok and Meta obstructed researcher access.

AI training is the sharpest dispute. Proof News reported in July 2024 that subtitles from 173,536 YouTube videos across more than 48,000 channels sat in a dataset used to train models from Apple, Nvidia, Anthropic and Salesforce. YouTube's chief executive, Neal Mohan, said in April 2024 that using videos to train OpenAI's Sora would be a "clear violation" of its terms. YouTube's own position is not neutral: Google trains its models on YouTube content, and the Commission opened an investigation in December 2025 covering AI Overviews and YouTube's content terms. Comparable claims are spreading, as a Twitch lawsuit over creator work used for training shows.

Control of YouTube data also extends beyond the API. In January 2026, YouTube sent cease and desist letters over audio-matching measurement, and Barb halted its UK reporting of YouTube viewing on television sets.

Not to be confused with

The Google Ads API manages advertising: campaigns, bids, audiences and ad performance, including YouTube ads. It requires a developer token, and PPC Land has covered both a token application backlog and a ban on programmatic proxies.

The YouTube Analytics API answers targeted queries about a channel's private performance, such as watch time, traffic sources and demographics, only for channels the user owns or manages. The YouTube Reporting API delivers the same kind of owner data as bulk daily report files for warehousing. Neither offers public search.

Web scraping extracts data from YouTube's pages without the API, outside its quota and, under YouTube's terms, without permission.

Recent developments

As of October 2026, the discovery document includes a videoTrainability method that reports whether a video's owner allows third parties to train AI models on it, returning "all", "none" or a list of permitted parties. It exposes the opt-in setting YouTube introduced in December 2024, letting creators authorise named AI companies. A batchGetStats method for retrieving statistics on several videos at once also appears. Its use echoes the Google-Extended model for websites: a machine-readable signal of consent, honoured only by those who check it.

Meanwhile, AI developers still draw on archives such as Common Crawl outside any API terms, and Google is building pipelines in the opposite direction, inviting forums and social platforms to send fresh posts through a UGC Fresh Data Program opened on October 8, 2026.

Timeline

  • 2005: YouTube launches; its first video is uploaded on April 23
  • October 2006: Google agrees to acquire YouTube for $1.65 billion
  • 2012: Version 3 of the YouTube Data API is introduced
  • August 23, 2013: YouTube raises v3 quota tenfold and upload limits a hundredfold
  • April 20, 2015: The v2 API's main feeds are switched off
  • December 13, 2021: Public dislike counts are removed from the API
  • 2022: YouTube launches its Researcher Program
  • April 25, 2023: YouTube is designated a very large online platform under the DSA
  • August 2023: Related-video search via relatedToVideoId ends
  • April 2024: Neal Mohan says training Sora on YouTube videos would breach its terms
  • July 2024: Proof News reports that subtitles from 173,536 YouTube videos were used in AI training data
  • December 2024: YouTube introduces an opt-in setting allowing third-party AI training
  • July 2, 2025: The European Commission adopts the DSA delegated act on researcher data access
  • December 5, 2025: The Commission fines X 120 million euros under the DSA
  • January 2026: YouTube's legal threats lead Barb to pause UK television reporting of YouTube
  • August 17, 2026: tuber 2.0.0 documents separate default buckets for units, searches and uploads
  • August 24, 2026: YouTube counts views from the first frame across all formats
  • October 2026: Google's discovery document lists the videoTrainability and batchGetStats methods

Summary

Who: Google operates the YouTube Data API. Developers of creator tools, influencer platforms, social listening and brand safety products, agencies, academic researchers and Google's own teams use it. The European Commission and courts shape the limits on access.

What: A REST interface, now in version 3, for reading public YouTube data and managing a user's videos, playlists, comments, captions and live streams, authorised by API keys or OAuth 2.0 and rationed by daily quota.

When: YouTube offered APIs within its first years; v3 arrived in 2012 and replaced v2 by April 2015. Quota, data removals, researcher access and AI training permissions have defined its evolution since 2021.

Where: Worldwide, through Google Cloud projects, with EU law adding researcher access duties for YouTube as a very large online platform.

Why: It gives software a sanctioned, structured route to YouTube data while letting YouTube control volume, storage and reuse. That control is the source of both its usefulness and the disputes over scarcity, research access and AI training.