Read a Reddit ad group's targeting and schedule
Shows a Reddit ad group's live targeting and schedule in the same shape PUT /v1/ads/ad-sets/{adSetId} takes, so you can prefill an editor.
Read live from Reddit on every call. Location and interest names are resolved with
Reddit's targeting taxonomy; when Reddit cannot resolve them the ids stand in as names and
a location that is not a two-letter country is listed under regions. Excluded DMAs are
listed under excludedLocations.regions, since excluded locations have no metro list.
Languages come back as lowercase ISO 639-1 codes. Targeting Zernio cannot edit (devices,
platforms, carriers, automated expansion) is left out and kept on every edit. Reddit only;
every other platform returns 501.
Authorization
bearerAuth API key authentication: send your Zernio API key in the Authorization header, prefixed with Bearer.
In: header
Path Parameters
Reddit ad group ID
Query Parameters
Disambiguates when the same adSetId string exists on more than one connected platform.
Value in
- "reddit"
Response Body
application/json
application/json
application/json
import Zernio from '@zernio/node';const zernio = new Zernio({ apiKey: process.env.ZERNIO_API_KEY });const { data } = await zernio.adcampaigns.getAdSetTargeting({ path: { adSetId: 'adset_abc123', },});console.log(data);{ "adSetId": "string", "targeting": { "userOs": [ "string" ], "userDevice": [ "string" ], "countries": [ "string" ], "regions": [ "string" ], "cities": [ "string" ], "zips": [ { "key": "string", "name": "string" } ], "metros": [ { "key": "string", "name": "string" } ], "countryGroups": [ "africa" ], "customLocations": [ { "latitude": -90, "longitude": -180, "radius": 0, "distanceUnit": "mile", "name": "string", "address": "string" } ], "excludedLocations": { "countries": [ "string" ], "countryGroups": [ "africa" ], "regions": [ { "key": "string", "name": "string" } ], "cities": [ { "key": "string", "radius": 0, "distanceUnit": "mile" } ], "zips": [ { "key": "string", "name": "string" } ], "places": [ { "key": "string" } ], "neighborhoods": [ { "key": "string" } ], "customLocations": [ { "latitude": -90, "longitude": -180, "radius": 0, "distanceUnit": "mile", "name": "string", "address": "string" } ] }, "ageMin": 13, "ageMax": 13, "gender": "all", "incomeTier": "top_5", "languages": [ "string" ], "interests": [ { "id": "string", "name": "string" } ], "behaviors": [ { "id": "string", "name": "string" } ], "workPositions": [ { "id": "string", "name": "string" } ], "workEmployers": [ { "id": "string", "name": "string" } ], "workIndustries": [ { "id": "string", "name": "string" } ], "industries": [ "string" ], "companySizes": [ "string" ], "seniorities": [ "string" ], "jobFunctions": [ "string" ], "skills": [ "string" ], "audienceInclude": [ "string" ], "audienceExclude": [ "string" ], "communities": [ "string" ], "excludedCommunities": [ "string" ], "conversationKeywords": [ "string" ], "excludedConversationKeywords": [ "string" ] }, "startDate": "string", "endDate": "string"}Read the platform's review verdict for an ad
Shows whether the platform approved an ad, where it may not deliver, and every rejection reason, without re-enabling a paused ad. Reads the ad's review verdict from the platform now: whether it was approved, where it may not deliver, and every rejection reason with TikTok's suggestion and the piece of content it refers to. Read-only, so it works on a paused ad without re-enabling it. **Google**: reads `ad_group_ad.policy_summary` live. `approvalStatus` and `reviewStatus` are Google's verbatim (`approvalStatus`: APPROVED, APPROVED_LIMITED, AREA_OF_INTEREST_ONLY, DISAPPROVED, UNKNOWN; `reviewStatus`: REVIEW_IN_PROGRESS, REVIEWED, UNDER_APPEAL, ELIGIBLE_MAY_SERVE); `approved` is true for the three approved statuses, false for DISAPPROVED, null otherwise. `policyTopics` carries every policy topic entry with its `type` (PROHIBITED, LIMITED, ...) and Google's `evidences` and `constraints` verbatim; each PROHIBITED topic is also listed in `rejections` (reason = the topic). The `forbidden*` arrays are TikTok-only and always empty on Google. **Meta** (Facebook and Instagram ads): reads the Ad's `ad_review_feedback`, `effective_status` and `issues_info` live. `reviewStatus` is Meta's `effective_status` verbatim (ACTIVE, PAUSED, PENDING_REVIEW, DISAPPROVED, PREAPPROVED, WITH_ISSUES, IN_PROCESS, ...); `approved` is true for ACTIVE and PREAPPROVED, false for DISAPPROVED, null otherwise. `reviewFeedback` carries Meta's feedback verbatim: `global` (reason -> explanation) and `placementSpecific` (placement -> reason -> explanation). Each global reason is listed in `rejections`, and so is each placement-specific reason, with that placement in its `forbiddenPlacements`; the review's `forbiddenPlacements` lists every placement with feedback. When Meta returns no feedback (usually an ad in review or approved everywhere), the status comes back with empty `rejections` and empty `reviewFeedback` maps; an ACTIVE ad rejected on one placement still carries that placement's feedback. The `effective_status` read is also stored on the ad, so `GET /v1/ads/{adId}` returns it next. A deleted ad answers 404. `issues` is Meta's `issues_info` verbatim. `forbiddenAges`, `forbiddenLocations` and `forbiddenOperatingSystems` are always empty on Meta. TikTok uses `/ad/review_info/`. Platforms other than TikTok, Google and Meta return 501. On TikTok, use it alongside the ad's `platformStatus`: TikTok reports `AD_STATUS_AUDIT` while the ad is in review and `AD_STATUS_AD_PRE_ONLINE` once it passed and is about to deliver (both map to `status: pending_review`); `AD_STATUS_AUDIT_DENY` maps to `rejected`.
Review status of each creative in a TikTok Smart+ ad
Shows which Spark post, video or carousel of an Upgraded Smart+ ad TikTok approved or rejected, with every rejection reason, keyed by the TikTok post id for a Spark post. Reads TikTok's `/smart_plus/ad/review_info/` live (read-only, works on a paused ad) for the Smart+ ad that holds `adId`, and returns the ad-level verdict plus one entry per creative. `adId` may be the Smart+ ad itself or any creative of it synced as its own ad: Zernio finds the Smart+ ad in the ad's ad group through `/smart_plus/ad/get/`, and uses the ad group's only Smart+ ad when no creative list names the ad. `materialType`, `tiktokItemId`, `videoId`, `name` and `operationStatus` come from the Smart+ ad's `creative_list`. TikTok leaves out of that list the creatives it added on its own (creator content, authorized TikTok posts, Creative Library content) and lists each image of a carousel under its own material id, so those entries come back with the review verdict and null descriptive fields. TikTok returns no review timestamp; `readAt` is when Zernio read it. TikTok's review endpoint has a strict quota of its own (TikTok documents neither the quota nor its scope). One TikTok call reads the review of the live Smart+ ads of the ad group up to 50 creatives in total (TikTok refuses more), and each review is stored per connection, advertiser and Smart+ ad. This read serves a review read in the last 10 minutes from the store, so sibling ads are answered without a TikTok call and `readAt` can be up to 10 minutes old; an older one is read again. Read reviews when you need them instead of sweeping every ad on a loop. When TikTok answers over quota, Zernio returns 429 and answers 429 for that advertiser and connection without calling TikTok for about 60 seconds, then 120 and up to 240 while TikTok keeps refusing, and about 60 again after a successful call. Each pause has up to 10% of random jitter, with the time left as `Retry-After`; reviews read in the last 10 minutes are still served.