Research · Social accounts
Official government social media accounts
As of 2026-11-17. Crawls CC-MAIN-2026-39 and CC-MAIN-2026-43.
Residents follow their city, county and special districts on social media for road closures, meeting notices, emergencies and events. Knowing which accounts are official helps everyone find the real one. We read the links that official government websites publish to their own accounts, and we are building a reviewed record of them, account by account.
What we found
The pages hold 14,496,971 links to the five platforms. After setting aside links that are not accounts, they name 83,533 accounts, counted once per site.
| Platform | Accounts | Share |
|---|---|---|
| 45,631 | 54.6% | |
| 12,783 | 15.3% | |
| X | 11,167 | 13.4% |
| YouTube | 9,444 | 11.3% |
| 4,508 | 5.4% | |
| All five | 83,533 | 100% |
A link to a platform is not always a link to an account. These were set aside and counted by kind:
| Link set aside | Links |
|---|---|
| Share and intent buttons | 1,174,696 |
| Single posts, photos, videos and playlists | 305,368 |
| Links in the body of a news or blog page | 140,805 |
| No valid account name in the link | 140,737 |
| Platform pages: login, help, policies and the like | 103,107 |
| Hashtags and searches | 19,024 |
| Embedded widgets | 14,776 |
Some accounts carry a flag that asks a reviewer to look closer. A flag is a pointer for review, not a judgment, and an account can carry more than one.
| Flag | Accounts |
|---|---|
| Linked from a single page of the site | 38,453 |
| Possibly a neighbouring government's account | 4,018 |
| Shared across 3 or more unrelated sites, often a vendor's | 1,306 |
| Linked from news pages | 529 |
| Linked from a staging or test site | 270 |
The strongest group is the site-wide core: 14,826 accounts on 6,757 sites.
Each is linked from the header, footer, navigation or JSON-LD of at least half of its site's captured pages, and carries no flag.
For example, x.com/portlandgov
is linked from 25,777 pages on portland.gov.
How an account enters Jurisdictional
- One owner: each account belongs to one record. That is a government, an agency, or a board or council. An office account that passes from holder to holder, such as a mayor's, belongs to the seat. An officeholder's own official account belongs to the person.
- Official accounts only: we follow the policy of the congress-legislators project and record accounts run by the government or office, with no campaign or personal accounts.
- Reviewed before it shows: every account starts as a suggestion, hidden from public pages. A reviewer confirms it after checking that the account loads, carries a verified or government mark, says it is official or links back to the site, and has posted recently.
- Provenance: the source of each account is the official page that links it, recorded like any other change.
-
Open formats:
confirmed accounts are published as schema.org
sameAsand Popololinks.
Related efforts
Many people have worked to make official accounts findable and to preserve what governments publish. Our work builds on theirs, and we're grateful for it.
- U.S. Digital Registry (GSA)
-
The registry let anyone confirm that a federal social media account, mobile app or mobile website was official. It originated in OMB Memorandum M-17-06, which M-23-22, "Delivering a Digital-First Public Experience", rescinded in September 2023. GSA's registry page says the public search page was deprecated on September 19, 2024.
FedScoop reported in April 2024 that agencies were no longer required to register their accounts. The documentation for its earlier API is archived on GitHub.
- End of Term Web Archive
-
A collaborative initiative that preserves U.S. government websites at each change of administration: 2008, 2012, 2016, 2020 and 2024. Its 2024 crawl includes federal content in other domains, social media among them, and the Internet Archive makes the captures available through the Wayback Machine.
Its partners for 2024 include the Internet Archive, the University of North Texas Libraries, Stanford University Libraries, the Library Innovation Lab at Harvard Law School, Webrecorder, Archive Team, the Environmental Data & Governance Initiative, and Common Crawl.
- congress-legislators
-
Its
legislators-social-media.yamllists current social media accounts for Members of Congress, official accounts only, with no campaign or personal accounts. Its policy on which accounts count is the one we follow. - Open States people data
-
Curated YAML records of state legislators, governors and some municipal leaders. Each person's
idscan hold the usernames of their official Twitter, YouTube, Instagram and Facebook accounts. - Wikidata
-
Wikidata records accounts as properties of an item: P2002 X (Twitter) username, P2013 Facebook username, P2003 Instagram username, P2397 YouTube channel ID, and P4264 LinkedIn company or organization ID. We keep handles in the same form, so records can be matched in both directions.
- NARA Bulletin 2014-02
-
The National Archives' guidance on managing social media records, issued October 25, 2013. It says content on social media is likely a federal record, and gives agencies questions to decide how to manage it.
- Social Feed Manager
-
Open-source software from George Washington University Libraries that helps researchers and archivists collect social media. Its team says it is no longer in active development, and others may run their own instances.
Method
- Pages: every HTML page in the municipal Common Crawl subset, 4,198,667 from CC-MAIN-2026-43 and 1,144,893 from CC-MAIN-2026-39. The subset holds the crawls' pages for 15,931 official municipal domains, from the same two crawls as our Common Crawl study. The data comes from Common Crawl, and we're grateful for its work.
-
Links:
a scripted HTML parse that reads every link and JSON-LD
sameAson each page and records where on the page it sits. No AI model read the pages, and no platform was contacted. -
One account, one link:
each account link is rewritten to one canonical form, such as
https://x.com/handle, and counted once per site. - Flags: scripted rules that point a reviewer at accounts to check, never rules that drop them.
-
Run:
municipal-common-crawl 0.2.4, run
social-c37a6392on 2026-11-17, with all 120 input files scanned and checked against the run's manifest.
Limits
- Common Crawl visits only part of the web, and many small government sites have few pages in it or none. An account not found here may well exist.
- Every count describes these two crawls, as of their capture dates. Accounts open, close and change names after that.
- A link from an official page is strong evidence, not proof, that an account is official. That is why every account waits for review.
- The flags are heuristics. A flagged account may be official, and an unflagged one may not be.
- We publish counts here, not lists of governments by what they link or don't.
- Data in Jurisdictional is published under the Open Database License (ODbL) 1.0.
What's next
Next, we plan to review the suggested accounts, starting with the site-wide core. Review happens in a browser, account by account, against the checks above.
Later, we plan to read confirmed accounts through each platform's own API, for an intelligence model in GovCenter. That work has not started, and we will describe its method here before we publish anything from it.
Take part
-
Add an account:
know your government's official account?
Add it with your own agent through the Jurisdictional skill, or through the MCP tool
create_entitywithentity_typesocial_account. Cite the official page that links it. - Get the data: see the data license for how to request it.
- Work with us: run a government account, keep a registry, or study official accounts? Write to support@jurisdictional.org.
Suggested citation
Jurisdictional. 2026. "Official Government Social Media Accounts." Data as of November 17, 2026. https://jurisdictional.org/research/official-social-accounts.
83,533
We scanned 5,343,560 pages from 15,931 municipal domains in two Common Crawl crawls. Those pages link 83,533 distinct accounts on X, Facebook, Instagram, YouTube and LinkedIn, from 10,105 sites. The scan is a plain reading of HTML links, with no AI and no calls to any platform.