Botijada: fake and coordinated followers on Instagram (Serbia)

The goal of this dataset is to help detect fake and coordinated followers.

It contains follower lists from Instagram profiles in Serbia that were suddenly followed by thousands of unknown accounts. The profile owners downloaded the lists from Instagram themselves and sent them to botijada.surf, often several times over a few days. The data is what they sent, with every username replaced by a random ID. There are no analysis results in it.

Srpski: Cilj ovog skupa podataka je da pomogne u otkrivanju lažnih i koordinisanih pratilaca. Sadrži spiskove pratilaca koje su vlasnici napadnutih Instagram profila u Srbiji sami preuzeli sa Instagrama i poslali sajtu botijada.surf. Sva korisnička imena zamenjena su nasumičnim oznakama.

Where the data comes from

Instagram lets every account owner download their own data (Accounts Center → Your information and permissions → Export your information, format JSON). The export has a folder connections/followers_and_following/, and in it:

• followers_1.json, followers_2.json and so on: the profile's followers, split into several files for big profiles. Each entry is one follower: their username, a link to their profile and the time they followed, in Unix seconds:

  [
    {
      "title": "",
      "media_list_data": [],
      "string_list_data": [
        { "href": "https://www.instagram.com/some.username", "value": "some.username", "timestamp": 1759239726 }
      ]
    }
  ]
  

• follow_requests_you've_received.json: only in exports of private profiles, the accounts that asked to follow and weren't approved yet, in a similar form.

Only these files were sent to botijada.surf. Everything else in the export (who the owner follows, messages, photos) never left the owner's device. The owner could export the whole history or only a date range (often the last 7 days). The upload doesn't say which: the times of its oldest and newest follower show the range it covers.

From each upload we kept the follow times and replaced each username with a random ID, so some.username above becomes something like a_319d84d94c877152. All of an upload's followers_N.json files together make one upload file here. Instagram has changed the layout of these files a few times. The data here is the same whichever layout the owner had.

Files

targets.json                      every profile: its ID, kind and how publicly active it is
uploads/<target_id>/<list_id>.json   one upload, as sent
data/followers.parquet            all uploads in one table
data/requests.parquet             follow requests (private profiles)
examples/followers_per_day.py     draws one profile's new followers per day

Each upload file looks like this:

{
  "list_id": "L_269460d2d4",
  "target_id": "T_3ba6b555",
  "sent_at": "2026-09-30T19:09:06Z",
  "export_made_at": "2026-09-30T19:03:14Z",
  "complete": true,
  "duplicate_of": null,
  "reconstructed": false,
  "form": { "kind": "activist", "activity": "a_little", "signs": ["follower_spike"], "attack_since": "2026-08-24" },
  "followers": [
    {"account": "a_319d84d94c877152", "followed_at": "2026-09-30T13:42:06Z"}
  ],
  "follow_requests": []
}

• followers: the accounts in the list, newest first, with the time each one followed.
• follow_requests: for private profiles, accounts that asked to follow and weren't approved yet.
• sent_at: when the list was sent to us. export_made_at: when Instagram made the export.
• complete: every followers_N.json file of the export was sent (none missing). It does not mean the whole history: most uploads are date-range exports. An upload with complete: false shows who was listed, but not who was missing.
• duplicate_of: the same files were already sent earlier, in that upload.
• reconstructed: an early upload sent before we kept the original files. Its list was rebuilt from our records, so it is less exact.
• form: what the owner answered when sending: the kind of profile, how publicly active they are, the signs they noticed, and when the attack started.

data/followers.parquet has the same rows as the upload files, flat: target_id, list_id, sent_at, account, followed_at. data/requests.parquet is the same for follow requests.

IDs and times

• account (a_ + 16 hex characters), target_id (T_ + 8) and list_id (L_ + 10) are random. They are not derived from usernames, so nobody can compute one from a username.
• The same account has the same ID on every profile and in every release. That is what lets you see one account following many profiles.
• All times are UTC, to the second.
• Profile kinds with fewer than 3 profiles are shown as other.

Using it

Comparing two uploads of the same profile shows who stopped following in between. An account that follows many profiles within minutes of each other, or shows up and disappears in a sudden wave, is worth a closer look. We leave those conclusions to you.

from datasets import load_dataset
followers = load_dataset("gavrilo/botijada", "followers", revision="v2026.10")["train"].to_pandas()

-- DuckDB: on how many profiles does each account appear?
SELECT account, count(DISTINCT target_id) AS profiles
FROM 'hf://datasets/gavrilo/botijada/data/followers.parquet'
GROUP BY 1 ORDER BY 2 DESC LIMIT 20;

Example: new followers per day

Every follower comes with the time it followed, so counting them by day shows when they came. examples/followers_per_day.py draws this for one profile, from its longest list. With uv it runs as is:

uv run https://huggingface.co/datasets/gavrilo/botijada/resolve/main/examples/followers_per_day.py

Without a profile ID it picks the one with the biggest one-day wave. Add an ID to draw another, for example T_1872ea67.

Profile T_101b88c9 gained about 14 followers a day for eight weeks, never more than 51. Then 607, 723 and 698 in three days, and 338 more by the morning the list was sent. The chart counts only accounts that still followed when the list was made, so accounts that had already left are not in it.

Limits

• Only profiles whose owners sent lists are included. This is not a sample of Instagram.
• Accounts removed before the first upload or after the last one are never seen.
• When an account is missing from a later upload, the data doesn't say why: Instagram may have removed it, it may have unfollowed, the owner may have removed it, or it changed its name.
• A renamed account gets a new ID.

Privacy and use

The IDs are pseudonymous, not anonymous. Follow times are exact and public follower lists can be seen on Instagram, so someone who knows a profile could match some IDs back to usernames.

Please don't:

• try to identify accounts or profiles;
• harass or report anyone based on this data;
• use it to build tools for buying followers or avoiding detection.

The license (CC BY 4.0) doesn't forbid these uses, so this is a request.

To have an account removed, open a discussion in this repository's Community tab and ask us to contact you. Don't post the username there. We remove the account, publish a new release within 30 days and rewrite this repository's history. Copies that were already downloaded can't be recalled.

Versions

Releases are tagged by month (v2026.10, then v2026.11 and so on). IDs stay the same between releases. Changes are listed in CHANGELOG.md.

Citation

@misc{botijada_2026,
  title  = {Botijada: fake and coordinated followers on Instagram (Serbia)},
  author = {{botijada.surf}},
  year   = {2026},
  note   = {Release v2026.10},
  url    = {https://huggingface.co/datasets/gavrilo/botijada}
}