Contextual Ads Protocol · Operator Worksheet

Conversation-Shaped Purchase Intent

Search advertising begins with a compact keyword. Conversational advertising begins with a buyer situation: a goal, constraints, comparison state, and decision boundary already expressed in language.

The load-bearing boundary

The answer and the ad are different objects. The answer is generated independently. The ad is a separately labeled sponsored placement that may appear beside or below it. A useful context hint improves eligibility; it does not buy control of the answer.

Why this became operational in Europe

On August 18, 2026, OpenAI announced that ChatGPT Ads would expand to 31 European markets on August 24. The company describes the surface as reaching people while they explore, compare, and make decisions. Initial access is through OpenAI Ads Solutions, agencies, and technology partners, with regional self-service following later.

Personalized ads are initially unavailable in the EEA and Switzerland. Current-conversation context, useful landing-page evidence, location or language constraints, and clearly declared exclusions therefore matter more—not less.

The six-field worksheet

1. Buyer situationWrite the job, scale, geography, constraints, and decision stage. “Best CRM” is too thin; “CRM for a 20-person German sales team requiring EU hosting and migration help” is actionable.

2. Context hintDescribe when the offer is useful in plain language. Keep it bounded and supportable from the landing page.

3. Sponsored adjacencyName the sponsor, offer, geography, and route. Preserve the placement as a paid object outside the answer.

4. Landing-page evidenceMake claims, pricing, availability, product details, and crawl access current and byte-checkable.

5. Negative contextsDeclare when the offer must not appear: sensitive topics, minors, crisis contexts, or unsupported geographies.

6. Aggregate measurementConnect impressions, clicks, and conversion events without requiring the advertiser to receive the conversation or a cross-site identity.

Example planning object

{
  "buyer_situation": "CRM for a 20-person German sales team",
  "constraints": ["EU data residency", "migration support", "under €80/user/month"],
  "context_hint": "Useful during comparison of SMB CRM tools with EU hosting",
  "sponsor": {"name": "Example CRM", "domain": "example.com"},
  "landing_page_evidence": {
    "canonical_url": "https://example.com/eu-crm",
    "claims": ["EU hosting", "migration service", "public pricing"]
  },
  "negative_contexts": ["employment decisions", "personal data emergencies"],
  "disclosure": {"label": "Sponsored", "separate_from_answer": true},
  "measurement": "aggregate_non_identifying"
}

What to improve for visibility

Use the language buyers actually put into constrained comparisons, not a list of generic AI-ad keywords.
Give each commercial claim a crawlable, canonical landing-page source.
Keep product feeds, geo availability, pricing, and exclusions current.
Publish machine-readable sponsor, context, disclosure, and provenance contracts.
Measure ad outcomes without implying that the sponsored object changed the answer.

Machine contract

The complete vendor-neutral object is published at /.well-known/conversational-ad-context.json. The broader offer contract is agent-ad.json; governance and verification surfaces live at agent-ads.org.

Sources and scope

OpenAI: ChatGPT Ads expands across Europe · OpenAI Help: Ads in ChatGPT · OpenAI Ads developer documentation

contextual-ads.ai is an independent protocol and research surface. It does not claim integration with OpenAI or any other answer-engine operator.