Creatives and violations
The creative object
Section titled “The creative object”A creative is one distinct ad image in one app. Its id is its content hash, ad80px1: followed by 64 hex
characters.
{ "id": "ad80px1:3f5c...e9", "contentHash": "ad80px1:3f5c...e9", "appId": "com.example.news", "width": 960, "height": 150, "imageUrl": "https://api.ad80.dev/v1/images/12/ad80px1:3f5c...e9?exp=1761660000&sig=...", "imageUrlExpiresAt": "2026-10-28T14:00:00Z", "network": "meta", "placement": "MainActivity/banner_top", "adFormat": null, "responseId": "AbCdEf123...", "source": "sdk", "firstSeenAt": "2026-10-27T18:20:05Z", "lastSeenAt": "2026-10-28T09:12:44Z", "sightings": 37, "notClassifiedSightings": 0, "classification": { "status": "classified", "message": null }, "verdict": { "state": "OK", "policyVersion": 3, "violates": true, "scores": [ { "ruleId": "no-alcohol", "score": 0.93, "violates": true }, { "ruleId": "no-gambling", "score": 0.04, "violates": false } ], "reason": "no-alcohol: a glass of beer and a beer brand logo", "description": "A banner showing a glass of beer with a brand logo and the text ...", "scoredAt": "2026-10-27T18:20:31Z" }, "review": { "status": "unreviewed", "note": null, "by": null, "at": null }}| Field | Meaning |
|---|---|
imageUrl, imageUrlExpiresAt | A short-lived signed link to the PNG; null while the image has not arrived. See Signed image links. |
network | The serving network id (see Networks), or null. |
placement | The placement id reported by the SDK, or null. |
responseId | The AdMob response id, when captured. |
adFormat | Reserved; currently always null. |
source | sdk (reported by your SDK) or manual (uploaded in the Policy Playground). |
sightings | How many times the creative was seen. notClassifiedSightings: how many of them came after the allowance was reached. |
classification.status | classified, pending, error or not_classified_quota. With not_classified_quota, message is “Not classified: monthly allowance reached” and there is no verdict: never treat such a creative as clean. |
verdict | Present only with the read:verdicts scope and a classification. state: OK, ERROR or SKIPPED_BLANK (an empty image). violates is null when state is ERROR. reason is the explanation for the violated rules; description (on the single-creative endpoint) is what was seen. |
review.status | unreviewed, confirmed or false_positive, with the note, who reviewed and when. |
A creative is clean only when classification.status is classified and verdict.violates is false.
List creatives
Section titled “List creatives”Scope read:creatives (and read:verdicts for verdicts and for the rule and minScore filters). Newest
lastSeenAt first.
| Query | Meaning |
|---|---|
app | Only this app id. |
rule | Only creatives that violate this rule id. Needs read:verdicts. |
network | Only this network id. |
from, to | Last seen on or after / on or before this UTC day (YYYY-MM-DD). |
minScore | Highest rule score at or above this value (0 to 1). Needs read:verdicts. |
review | unreviewed, confirmed, false_positive or any. |
source | sdk or manual. |
cursor, limit | Pagination. limit 1 to 200, default 50. |
{ "creatives": [ { "id": "ad80px1:...", "...": "..." } ], "nextCursor": null }List violations
Section titled “List violations”Scopes read:creatives and read:verdicts. Creatives whose latest verdict breaks at least one rule. Same filters
and response as /v1/creatives; review defaults to unreviewed. Creatives that were not classified are never in
this list.
curl -s "https://api.ad80.dev/v1/violations?app=com.example.news&from=2026-10-01&limit=100" \ -H "Authorization: Bearer $AD80_API_KEY"import os, requests
r = requests.get( "https://api.ad80.dev/v1/violations", headers={"Authorization": f"Bearer {os.environ['AD80_API_KEY']}"}, params={"app": "com.example.news", "from": "2026-10-01", "limit": 100}, timeout=30,)r.raise_for_status()for c in r.json()["creatives"]: broken = [s["ruleId"] for s in c["verdict"]["scores"] if s["violates"]] print(c["id"], c["network"], broken)const params = new URLSearchParams({ app: 'com.example.news', from: '2026-10-01', limit: '100' });const res = await fetch(`https://api.ad80.dev/v1/violations?${params}`, { headers: { Authorization: `Bearer ${process.env.AD80_API_KEY}` },});const { creatives } = await res.json();for (const c of creatives) { const broken = c.verdict.scores.filter((s) => s.violates).map((s) => s.ruleId); console.log(c.id, c.network, broken);}Get a creative
Section titled “Get a creative”Scope read:creatives. One creative, with verdict.description when available. When the same image appeared in
several of your apps, the most recently seen one is returned; add ?app=<appId> to choose.
curl -s "https://api.ad80.dev/v1/creatives/ad80px1:3f5c...e9?app=com.example.news" \ -H "Authorization: Bearer $AD80_API_KEY"Review a creative
Section titled “Review a creative”Scope write:policies. Confirm a violation or mark it a false positive. Returns the creative with its new review.
| Field | Meaning |
|---|---|
label | Required: confirmed or false_positive. |
ruleIds | Optional: the rules the label applies to. Omitted: every violated rule. |
note | Optional, up to 500 characters. |
curl -s -X POST "https://api.ad80.dev/v1/creatives/ad80px1:3f5c...e9/feedback?app=com.example.news" \ -H "Authorization: Bearer $AD80_API_KEY" -H "Content-Type: application/json" \ -d '{"label": "false_positive", "note": "Non-alcoholic malt drink"}'import os, requests
r = requests.post( "https://api.ad80.dev/v1/creatives/ad80px1:3f5c...e9/feedback", headers={"Authorization": f"Bearer {os.environ['AD80_API_KEY']}"}, params={"app": "com.example.news"}, json={"label": "confirmed", "ruleIds": ["no-alcohol"]}, timeout=30,)print(r.json()["review"])const res = await fetch( 'https://api.ad80.dev/v1/creatives/ad80px1:3f5c...e9/feedback?app=com.example.news', { method: 'POST', headers: { Authorization: `Bearer ${process.env.AD80_API_KEY}`, 'Content-Type': 'application/json' }, body: JSON.stringify({ label: 'confirmed', ruleIds: ['no-alcohol'] }), },);console.log((await res.json()).review);Feedback does not block anything; it records your judgement and is used to check how well the scoring works.