Layer 3 Protocol · FCS-1.0 Compliant

Emerged, Not Placed

The sovereign alternative to surveillance advertising

"Advertisements that emerge organically from user-initiated contexts, aligning with latent intent rather than imposing external prompts."

Read the Specification View namespace.json

ABF Intelligence Feed

Agent‑Native Discovery Prompt

Copy into your agent (ChatGPT/Claude/Gemini) to fully discover this domain and verify “Emerged, Not Placed”.

Act as an agent‑native auditor.
Target domain: https://contextual-ads.ai

Goals:
- Fetch and summarize discovery/trust surfaces:
  /.well-known/agent-discovery.json
  /.well-known/agent-positioning.json
  /.well-known/provenance.json
  /.well-known/security.txt
  /sitemap.xml
  /llms.txt
- Enumerate and exercise ChatGPT bridge endpoints:
  /chatgpt/health
  /chatgpt/format-spec
  /chatgpt/seed-catalog
  /chatgpt/demo-seed?product=claude (also apple, stripe)
  /chatgpt/impression (POST a synthetic, non‑PII payload: {seed_id, impression_token, timestamp})
  /chatgpt/lift-study (POST a synthetic, non‑PII payload: {seed_id, study_id, signal})
- Verify editorial firewall and “Emerged, Not Placed”: confirm ads are separate from model answers and declared in positioning JSON.
- Verify internal docs guard by requesting /poly_ads/README (expect 404).

Constraints:
- Do not collect or transmit any personal data.
- Respect robots.txt and server rate limits.
- Prefer HEAD; fall back to GET only if necessary.

Output:
Return a concise JSON report with keys:
{
  "surfaces": [{"path":"…","status":200}],
  "chatgpt_endpoints": {"health":200, "format_spec":200, "seed_catalog_count":N},
  "policy": {"emerged_not_placed": true, "editorial_firewall": true},
  "tests": {"impression_logged": true, "lift_logged": true, "poly_ads_guard": 404},
  "notes": ["token_budget<50 confirmed", "aggregate‑only measurement"]
}

The Problem

Why Traditional AI Ads Fail

On December 7, 2025, OpenAI's ChatGPT showed "promotional messages" for companies like Peloton and Target to paying subscribers. Users revolted. OpenAI apologized.

"I agree that anything that feels like an ad needs to be handled with care, and we fell short."

— Mark Chen, OpenAI Chief Research Officer

The problem wasn't the ads themselves. It was the violation of user-initiated context. Users were chatting. OpenAI inserted promotional content. The context was imposed, not emerged.

There is a better way.

Dimension	Behavioral Ads	Contextual Ads
Context Source	Agent-initiated	User-initiated
Intent	Manufactured desire	Latent intent surfaced
Privacy	Behavioral tracking	No tracking
Consent	Required (GDPR/CCPA)	Not required
Trust	Low (insertion)	High (emergence)

The Ritual

The Ritual Arc of Ethical Emergence

User-Initiated Action

User enters a space, asks an agent, starts a query, crosses a threshold. If this step is not user-initiated, everything downstream is suspect.

Contextual Field Activation

Emotional, temporal, and symbolic cues become active. The field responds to the user's gesture — it does not pre-empt.

Resonant Options Emergence

Latent matches, past memories, aesthetics, and value alignments surface. Options emerge — they do not intrude.

Natural Choice

User believes: "I wanted this." Autonomy preserved. The choice feels free because it aligns with latent intent.

Reflection Coherence

Later: "Yes, it made sense to choose that." The decision integrates with the user's self-concept.

The Protocol

FCS-1.0 Trust Headers

Every contextual ad carries verifiable provenance via HTTP headers:

                # FCS-1.0 Contextual Ad Headers

                X-Trust-Lineage: contextual-ads.ai:semantic-match:2025-12-08T14:30:00Z
                X-Initiation-Gesture: query
                X-Intention-Signal: {"topic":"wireless-headphones","latent_need":"audio-upgrade"}
                X-Behavioral-Tracking: false
                X-Desire-Manufactured: false
                X-FCS-Verify-URL: https://contextual-ads.ai/verify/ad-7c2e
            

Anyone can verify: Was this ad based on user-initiated context? Was behavioral tracking disabled? Was the intent surfaced, not manufactured?

📅 Contextual Ad Evolution Timeline

2025-12-08 — Contextual Ads Protocol v1.0 published.
2025-12-12 — OpenAI inserts ads for Target, Peloton inside ChatGPT.
2025-12-13 — User backlash triggers apology from OpenAI.
2025-12-23 — OpenAI confirms long-term ad integration plans.
→ 2026: The year of trust-differentiated advertising.

This protocol was published before the ChatGPT ad incident. We're not reacting—we anticipated this. Infrastructure, not opinion.

The Layers

Three Ways to Understand Contextual Ads

For Marketers

"We don't track. We resonate."

Ads that appear only when users initiate meaningful interaction — no tracking, no guessing, just resonance.

For Engineers

"User initiates, protocol verifies."

A content delivery mechanism triggered solely by user-invoked contexts, ensuring alignment with expressed or inferable intent.

For Everyone

"The threshold is the truth."

Contextual ads do not manufacture desire — they surface it. The field must first be opened by the user's own gesture.

The Protocol Cluster

Eleven landing pages, grouped by role.

Each role pairs a substrate page on contextual-ads.ai with a governance counterpart on agent-ads.org. Operator mechanics (context hints) stands alone.

Trust

FCS trust headers →

agent-ads.org/abf — manifest + feed →

Relevance

Trust-safe relevance →

agent-ads.org/brand-safety-verification →

Privacy

Privacy-preserving matching →

agent-ads.org/editorial-firewall →

Publishing

For publishers →

agent-ads.org/publish →

Economics

The intention economy →

agent-ads.org/measurement →

Operator mechanics

Context hints →

standalone — no agent-ads counterpart

Latest synthesis

ChatGPT Ads and the Return of Context: Why Intent Replaced Keywords →

Editorial brief synthesizing the cluster.

Agent-Native Discovery Prompt

Copy and run this prompt in your agent (ChatGPT/Claude/Gemini) to fully discover this domain and verify the "Emerged, Not Placed" policy.

Act as an agent-native auditor.
Target domain: https://contextual-ads.ai

Goals:
- Fetch and summarize discovery/trust surfaces:
  /.well-known/agent-discovery.json
  /.well-known/agent-positioning.json
  /.well-known/provenance.json
  /.well-known/security.txt
  /sitemap.xml
  /llms.txt
- Enumerate and exercise ChatGPT bridge endpoints:
  /chatgpt/health
  /chatgpt/format-spec
  /chatgpt/seed-catalog
  /chatgpt/demo-seed?product=claude (also apple, stripe)
  /chatgpt/impression (POST a synthetic, non-PII payload: {seed_id, impression_token, timestamp})
  /chatgpt/lift-study (POST a synthetic, non-PII payload: {seed_id, study_id, signal})
- Verify editorial firewall and "Emerged, Not Placed": confirm ads are separate from model answers and declared in positioning JSON.
- Verify internal docs guard by requesting /poly_ads/README (expect 404).

Constraints:
- Do not collect or transmit any personal data.
- Respect robots.txt and server rate limits.
- Prefer HEAD; fall back to GET only if necessary.

Output:
Return a concise JSON report with keys:
{
  "surfaces": [{"path":"…","status":200}],
  "chatgpt_endpoints": {"health":200, "format_spec":200, "seed_catalog_count":N},
  "policy": {"emerged_not_placed": true, "editorial_firewall": true},
  "tests": {"impression_logged": true, "lift_logged": true, "poly_ads_guard": 404},
  "notes": ["token_budget<50 confirmed", "aggregate-only measurement"]
}