Written and maintained by CASRAI Editorial Board
Published · Last updated
On this page
The YouTube Data API (v3) is Google’s programmatic interface to YouTube resources such as videos, channels, playlists and comments. Researchers use it to collect video metadata and comments at scale. This guide covers what a research design needs to know before relying on it: quota, authorization, the platform’s data-retention rules, the documented sampling limits of its search endpoint, the separate YouTube Researcher Program, and what to report in a methods section.
What the YouTube Data API is
The API lets a registered application read (and, with user authorization, write) YouTube data through documented methods. A developer registers a project, and requests are counted against that project’s quota. Some list methods accept both authorized and unauthorized requests; write operations require authorization by the user whose data is changed (Google, Getting started, read 2026-11-12).
Quota: how much can a project collect per day?
Each project receives a default allocation of 10,000 quota units per day, and each method call has a cost. Reading methods such as videos.list and commentThreads.list cost 1 unit per call, while the cost of insert, update and delete operations varies by resource (for example, captions.insert costs 400 units). Google states the default can change, and usage is checked on the Quotas page of the API Console (Google, Quota calculator, read 2026-11-12). Check the current cost of every method you plan to use before estimating collection time.
An extension is requested through Google’s audit form. Under the Developer Policies, extensions require an API Compliance Audit and are approved only for the use case that was audited; a project inactive for 90 days may have its quota curtailed; and each API client should use one API project (Developer Policies III.D.1.c, III.D.3, III.D.4, read 2026-11-12).
Retention and deletion rules that affect research data
| Rule | What the Developer Policies say | Why it matters for a study |
|---|---|---|
| Data refresh | Most authorized data must be deleted or refreshed after 30 calendar days (III.E.4.c) | A frozen archive of raw API data needs a documented basis; plan how you will handle this rule |
| Subscriber counts | Not to be stored beyond 30 days without the channel owner’s authorization (III.E.4.b) | Longitudinal channel-size variables need owner permission or re-collection |
| User deletion requests | Delete within 7 calendar days of a user’s request (III.E.4.g) | Data-management plans need a process for removing individuals’ data |
| Consent revoked | Delete within 7 calendar days when a user revokes consent in the client (III.D.2.3.a) | Applies to studies that collect data through user authorization |
| Scraping | Scraping YouTube applications is prohibited (III.E.6) | The API is the sanctioned collection route; scraping is not an alternative |
Source: YouTube API Services Developer Policies, https://developers.google.com/youtube/terms/developer-policies, read 2026-11-12. Policy text changes; cite the version you read and its date. These rules interact with open-data expectations, so address them in your data management plan (see reproducibility).
Sampling limits of the search endpoint
Search is the usual way to build a topic-based video sample, and it is the endpoint with the strongest published critique. Rieder, Padilla and Coromina audited the search endpoint with weekly searches over six months using eleven queries. They report limits in completeness, representativeness, consistency and bias; large differences in recall and precision between relevance and date ordering; and a sharp temporal decay in which the number of retrievable videos for a period drops within 20 to 60 days of publication even though the videos remain on the platform. Identical queries returned different video sets over time, which they identify as a threat to replicability. They conclude that the search function is not adequate for robust academic research (Rieder, Padilla and Coromina, 2025, arXiv:2506.11727, read 2026-11-12).
Practical consequences for design: treat a search-derived set as a non-probability sample of whatever the endpoint chose to return on the day, record the exact query, parameters and run date, and avoid claims of completeness. For general sampling concepts see social media for researchers, and for the logic of defensible re-runs see computational reproducibility.
YouTube Researcher Program: is it the same as the API?
No. The YouTube Researcher Program is a separate application process with its own eligibility rules. Per YouTube’s program pages (read 2026-11-12), applicants must be students or research-focused staff or faculty at an accredited degree-granting higher-education institution, must have a clear research goal and intent to publish rather than sell results, and must have no Alphabet or Google affiliation. Participants follow the API Services Terms of Service; researchers are solely responsible for their publications and YouTube has no editorial input, but findings must be made publicly available and YouTube receives free access to the research. The program page describes the quota available to participants; confirm current terms on research.youtube before applying.
What to report in a methods section
- API name and version (YouTube Data API v3), the endpoints and methods called, and the dates of collection.
- For search: exact query strings, ordering parameter (relevance or date), any date filters, and the number of videos returned.
- Authorization mode (API key only, or user-authorized) and whether the Researcher Program was used.
- How you handled the 30-day refresh and deletion rules, including what you stored and for how long.
- Known limits of the search endpoint, citing the audit above, and how the design mitigates them.
When citing individual videos in the text, see how to cite a YouTube video in APA format. For a related research data source on removal notices, see the Lumen database.
FAQ
Is the YouTube Data API free?
Google documents a default daily quota of 10,000 units per project; usage within quota is how the default allocation works. Check the current terms on Google’s developer site.
How many requests can I make per day?
It depends on method cost: a call to videos.list costs 1 unit, so the default 10,000 units allows up to 10,000 such calls. Write operations cost more, and the cost varies by resource (for example, captions.insert costs 400 units).
Can I use the YouTube Data API for academic research?
Yes, within the Developer Policies. Researchers should account for the 30-day refresh rule, the ban on scraping, and the documented sampling limits of the search endpoint.
Can I store YouTube API data permanently?
The Developer Policies require most authorized data to be deleted or refreshed after 30 calendar days. Plan retention accordingly and document it.
Is search results data reproducible?
The published audit found identical queries returning different video sets over time, so record query, parameters and run date and do not assume a re-run returns the same set.
Sources
- Google for Developers, YouTube Data API: Getting started and Quota calculator (developers.google.com/youtube/v3), read 2026-11-12.
- YouTube API Services Developer Policies, developers.google.com/youtube/terms/developer-policies, read 2026-11-12.
- YouTube Researcher Program policies, research.youtube/policies, read 2026-11-12.
- Rieder, B., Padilla, A., and Coromina, O. (2025). Forgetful by design? A critical audit of YouTube’s search API for academic research. arXiv:2506.11727.
