News Discovery Brief
A publisher-controlled record that gives AI discovery services enough to help a reader choose the original story, credit its creators, and send a measurable visit back to the publisher.
Why this proposal exists
A rights-safe preview alone may still give readers little reason to visit the publisher. NDB combines a bounded public brief with the named reporting team, an approved image, the story type, and an optional hook describing what the reader will find in the full article. A canonical link and optional referral parameters make the visit possible to observe.
The record is a discovery aid. It is not the article, a substitute for reading it, a blanket training license, a paywall bypass, or proof that the reporting is correct. The brief, hook, image, and attribution assets remain subject to publisher terms. Nothing here grants rights to the article body or decides fair-use questions.
Candidate record
One JSON record represents one canonical article. This fictional example uses a reserved example domain. The hook describes material available in the full report without disclosing its central finding.
{
"format": "ndb/0.2-proposal",
"canonical_url": "https://publisher.example/news/transit-pilot",
"publisher": {
"name": "Example Publisher",
"url": "https://publisher.example/"
},
"headline": "Inside a city transit pilot",
"bylines": [{
"name": "Alex Rivera",
"role_or_beat": "Civic technology reporter",
"profile_url": "https://publisher.example/staff/alex-rivera"
}],
"section": "Civic technology",
"article_type": "investigation",
"published_at": "2026-11-02T18:00:00Z",
"access": "subscription",
"image": {
"url": "https://publisher.example/media/transit-pilot.jpg",
"alt": "A city bus at a transit stop",
"credit": "Example Publisher",
"discovery_license_url": "https://publisher.example/terms/ndb-image"
},
"public_brief": "The city launched a transit pilot. The full report examines how it was planned and funded.",
"brief_terms_url": "https://publisher.example/terms/ai-discovery",
"hook": "Read the records and interviews behind the rollout.",
"reading_time": { "minutes": 6 },
"media": ["documents", "chart"],
"topics": ["transit", "public procurement"],
"subscribe_url": "https://publisher.example/subscribe",
"attribution_display": {
"name": "Example Publisher",
"link_text": "Read the full report at Example Publisher"
},
"referral_params": {
"utm_source": "ndb",
"utm_medium": "ai_discovery"
},
"updated_at": "2026-11-03T09:00:00Z",
"update_type": "updated",
"corrections_url": "https://publisher.example/corrections"
}Required fields
A publisher participating in the pilot must supply these fields. Required fields describe identity, reader expectations, licensed display material, and the limits of reuse.
| Field | Purpose and rules |
|---|---|
format | The exact value ndb/0.2-proposal identifies this draft version. |
canonical_url | HTTPS URL of the original article. The brief must come from its publisher-controlled origin or a verifiably delegated origin. |
publisher | Publisher name and HTTPS homepage. This identifies the source but does not alone prove rights ownership. |
headline | Publisher-approved public headline. |
bylines | At least one credited person or institutional byline. Each entry has name, role_or_beat, and profile_url. Use null for a profile URL that genuinely does not exist; never invent one. |
section | Publisher's section or desk, such as Corporate & Legal. |
article_type | One of news, news_analysis, investigation, opinion, explainer, or other. Consumers must preserve the distinction in display. |
published_at | Original publication time in ISO 8601 format, distinct from crawl or discovery time. |
access | One of free, registration, subscription, or unknown. It describes access, not permission to reproduce the article. |
image | Publisher-approved thumbnail with url, alt, credit, and discovery_license_url. The field may be null when no image is licensed; consumers then display no image and must not obtain one elsewhere for this brief. |
public_brief | Short publisher-written context for discovery. It should not reveal a paywalled investigation's central finding. |
brief_terms_url | HTTPS terms for use of the brief. It does not license the full article body or override applicable law. |
Recommended fields
These fields make the brief more useful for discovery, updates, subscriptions, and measurement. A publisher should include them where applicable, using only claims and assets it has approved.
| Field | Purpose and rules |
|---|---|
hook | One publisher-approved line explaining what the reader gains by opening the article. It describes material inside, not the central finding, and must be accurate. |
reading_time | Estimated minutes, labeled as an estimate and kept current when the article changes. |
media | Available material such as video, charts, documents, or a data download. List only what is actually in the article. |
topics | Publisher-supplied subjects for matching relevant queries without retrieving the protected body. Named entities may be added separately in a later draft. |
updated_at and update_type | ISO 8601 update time and either updated or corrected. Do not silently treat a correction as an ordinary update. |
corrections_url | Publisher's correction notice or policy so consumers can resolve stale or wrong descriptions. |
subscribe_url | Publisher-controlled path to subscribe or see an offer. The consumer should label it separately from the article link. |
attribution_display | Preferred publisher name, approved logo URL if licensed, and descriptive link text. Accessibility and truthful attribution still apply. |
referral_params | Static publisher-supplied campaign parameters for the canonical article URL. No user identifiers, secrets, or redirect destinations. |
Publisher and consumer behavior
- The publisher hosts the JSON over HTTPS and links it from the canonical article page, for example with an HTML
linkelement usingrel="alternate"andtype="application/json". Theformatvalue identifies the record. This discovery convention is proposed for pilot testing, not established practice. - The consumer verifies the canonical URL and publisher-controlled origin or delegation. A copy on an unrelated site is not evidence of permission.
- The consumer displays the article type, access status, publisher, bylines, and canonical link. It may display the public brief, hook, and licensed thumbnail only within their stated terms. It must not fill a missing image from a mirror, scrape the protected body, or invent a more compelling hook.
- If the publisher returns HTTP 403, the consumer stops requests to that site for the run and holds the story for review. A paywall is handled through the public NDB if available and permitted; the consumer does not use another path, tool, cache, or mirror to obtain the article body.
- If present, referral parameters are appended to the canonical article link without changing its destination. Consumers must not add personal identifiers or secrets. The publisher should be able to distinguish article visits from subscription clicks.
- Consumers recheck changed briefs and correction notices. If a brief is missing, stale, contradictory, or untrusted, they fall back to conservative metadata and a link rather than treating an old copy as current authorization.
Limits on what the brief reveals
Show the value of opening
The hook can mention documents, interviews, charts, or other material the publisher confirms is in the article. Bylines and article type help readers judge what kind of work they are opening.
Keep the reporting in the article
Do not include key quotes, the full article body, or an investigation's central finding in the brief. A public preview should aid a reader's decision without replacing the story.
Use approved visuals
A reachable image URL is not a license. The record names a thumbnail, alt text, credit, and discovery-use terms, or explicitly says there is no approved image.
Measure before claiming benefit
Referral parameters can help count visits. A pilot must still test whether visits, reader understanding, or subscriptions improve; the format itself proves none of those outcomes.
Relationship to existing formats
Schema.org NewsArticle ↗ describes article metadata. Google's paywall guidance ↗ covers access markup, while its preview controls ↗ include snippet limits. IPTC NewsML-G2 ↗ addresses news metadata and rights. NDB proposes a smaller publisher-authored discovery payload and referral behavior alongside those mechanisms; it does not override them.
Pilot questions and version history
- Which origin or delegation checks work for syndicated articles and publisher CDNs?
- What hook and image combinations increase qualified visits without giving away protected reporting?
- How quickly must consumers react to corrections, withdrawals, and changed terms?
- Can publishers measure referrals and subscriptions without tracking individual readers through the brief?
0.1 introduced the canonical link, access status, public brief, and reuse boundary. 0.2 adds bylines, approved imagery, section and story type, click value, corrections, subscription paths, and referral measurement. The short 0.1 excerpt in the accompanying article is an early sketch, not the complete 0.2 record.
