Vet a creator

Check a profile against safety criteria before you work with them.

15 min 6 steps

Vetting reads a creator profile and the posts behind it, then reports against the safety criteria you asked for. You hand it a profile link, not an id, so it works for a creator Zelf has never seen.

The shape is the same as an AI analysis: create, poll, download. The difference is the input, a profile, and the criteria you choose.

Every step shows the same call in all six Zelf clients and in cURL. Pick a language and the install line above the snippet switches with it. Each client reads your key from ZELF_API_KEY.

01Choose the safety checks
02Start the vetting
03Wait for it to finish
04Download the report
05Read the disclosure findings
06See the posts it read

Choose the safety checks

Choose predefined safety criteria, describe your own, or add DISCLOSURE to detect stated commercial relationships. Disclosures are reported separately, not as violations. Only type is required on each entry.

`ADULT_CONTENT`predefined	Sexual or adult material.
`ARMS_AND_AMMUNITION`predefined	Weapons and ammunition.
`MISINFORMATION`predefined	Claims presented as fact that are not.
`PROFANITY`predefined	Strong or offensive language.
`POLITICS`predefined	Political positions and campaigning.
`DISCLOSURE`optional detection	Look for stated sponsorships, affiliate commissions, gifts, offer terms and risk warnings. Findings go in disclosures, not violations.
`USER_DEFINED_SAFETY_CRITERION`yours	Put the criterion in extra, written as you would explain it to a person. extra is ignored on the predefined types.

Start the vetting

POST/api/v1/creator/vetting/creator_vetting.create

content_scope decides which of the creator posts feed the analysis. Omitted, it means their own posts only. Set resync_profile when you want a fresh crawl of a creator Zelf already holds.

`profile_url`string, required	Direct link to the profile on instagram or tiktok.
`content_scope.owned`boolean	Posts the creator published. Defaults to true.
`content_scope.repost`boolean	Posts they reposted. Defaults to false.
`content_scope.liked`boolean	Posts they liked. Defaults to false.
`resync_profile`boolean	Re-index the creator even if Zelf already holds them. Defaults to false. When it triggers a crawl the response carries sync_job_id.

pip install zelf-sdk

import os

from zelf import ZelfAPIClient

client = ZelfAPIClient(api_key=os.environ["ZELF_API_KEY"])

result = client.creator_vetting.create(
    profile_url="https://www.instagram.com/milesandmornings",
    resync_profile=False,
    content_scope={
        "owned": True,
        "repost": True,
        "liked": False,
    },
    safety_checks=[{
        "type": "ADULT_CONTENT",
    }, {
        "type": "PROFANITY",
    }, {
        "type": "POLITICS",
    }, {
        "type": "DISCLOSURE",
    }, {
        "type": "USER_DEFINED_SAFETY_CRITERION",
        "extra": "Promotes competing running shoe brands",
    }],
)
print(result)

Response

{
  "results": [
    {
      "id": "bod_5Rn1Wf8d",
      "creator_id": "cr_4Tz9Hs1k",
      "type": "CREATOR_ANALYSIS",
      "status": "IN_PROGRESS",
      "created_at": "2026-10-07T09:31:44Z",
      "updated_at": "2026-10-07T09:31:44Z",
      "creator": {
        "id": "cr_4Tz9Hs1k",
        "username": "milesandmornings",
        "name": "Miles and Mornings",
        "platform": "instagram",
        "platform_url": "https://www.instagram.com/milesandmornings",
        "follower_count": 128400,
        "active_content_count": 612
      },
      "sync_job_id": null,
      "analysis": [
        { "id": "ban_9Ld3Kp2m", "scope": "CONTENT", "status": "IN_PROGRESS" }
      ],
      "safety_checks": [
        { "type": "ADULT_CONTENT" },
        { "type": "PROFANITY" },
        { "type": "POLITICS" },
        { "type": "DISCLOSURE" },
        {
          "type": "USER_DEFINED_SAFETY_CRITERION",
          "extra": "Promotes competing running shoe brands"
        }
      ],
      "content_scope": { "owned": true, "repost": true, "liked": false }
    }
  ],
  "item_count": 1,
  "page": 1,
  "next": null,
  "previous": null
}

Vetting analyzes up to 250 posts in total under the current content policy. The stored posts can include historical content no longer visible on the profile. A run can take an hour or more while waiting for preprocessing.

Wait for it to finish

GET/api/v1/creator/vetting/?id={vetting_id}creator_vetting.list

Read the vetting back by id. The statuses are the same four as an AI analysis: IN_PROGRESS, then COMPLETED, EMPTY_RESULT or FAILED. Poll every minute or two, or use the completion webhook to avoid polling.

pip install zelf-sdk

import os

from zelf import ZelfAPIClient

client = ZelfAPIClient(api_key=os.environ["ZELF_API_KEY"])

result = client.creator_vetting.list(
    id="bod_5Rn1Wf8d",
)
print(result)

The creator_vetting.completed webhook replaces this loop. It is a notification, so you still fetch the result in the next step.

Download the report

GET/api/v1/creator/vetting/results/?vetting_id={vetting_id}creator_vetting.results

The findings against each criterion live in the downloaded payload, not in the API response. The response carries a signed link that expires, so fetch it rather than storing it.

`vetting_id`string	Every run under the vetting. Send this or id, not both.
`id`string	One run.
`result`object or null	download_url, expires_at in unix seconds, and format.

pip install zelf-sdk

import os

from zelf import ZelfAPIClient

client = ZelfAPIClient(api_key=os.environ["ZELF_API_KEY"])

result = client.creator_vetting.results(
    vetting_id="bod_5Rn1Wf8d",
)
print(result)

Read the disclosure findings

When safety_checks includes DISCLOSURE, the downloaded report includes disclosures alongside the normal summary, themes and violations. Each list entry is one detected disclosure, with its post id, original URL and source. This illustrative excerpt shows the format, not a live verdict.

`content_id`string	Existing con_... post identifier for matching against the contents endpoint.
`content_url`string	Original post URL, or an empty string if unavailable.
`source`string	frame, audio, title, caption or metadata. frame means on-screen text; audio means speech transcription; caption means the published post text.
`kind`string	sponsorship, affiliate, gifted, relationship, terms or risk. A disclosure is not a violation.
`text`string	The detected disclosure quote.
`time_seconds`number or null	Approximate location in the video, when known. Published post text normally has no video time.

Response

{
  "disclosures": [
    {
      "content_id": "con_7Km3Qx9pLt",
      "content_url": "https://www.instagram.com/p/EXAMPLE/",
      "source": "frame",
      "kind": "sponsorship",
      "text": "Paid partnership with Example",
      "time_seconds": 5
    }
  ]
}

If no disclosure is detected, disclosures is []. Without the DISCLOSURE check, the field is omitted. There is no result wrapper, channels or coverage in the new list. Videos analyzed previously are not re-analyzed for disclosure, and resync_profile does not force a new analysis. Video checks are limited to 10 minutes and sampled frames can miss brief text. An empty list is not proof that every source was checked or that a post has no disclosure. Existing saved reports keep their earlier format. Findings are not a legal compliance assessment.

See the posts it read

GET/api/v1/creator/vetting/contents/?vetting_id={vetting_id}creator_vetting.contents

Same pair as an AI analysis, keyed by vetting_id. Use the statistics call when you only need the count and the totals.

pip install zelf-sdk

import os

from zelf import ZelfAPIClient

client = ZelfAPIClient(api_key=os.environ["ZELF_API_KEY"])

result = client.creator_vetting.contents(
    vetting_id="bod_5Rn1Wf8d",
    page=1,
)
print(result)

Summary

Step	Request	Input	Output
1	`POST /api/v1/creator/vetting/`	profile_url, content_scope, safety_checks	vetting id (bod_...)
2	`GET /api/v1/creator/vetting/?id=`	vetting id	status
3	`GET /api/v1/creator/vetting/results/`	vetting id	signed download_url
4	`GET /api/v1/creator/vetting/contents/`	vetting id	the posts it read