ChatGPT advertising platform officially reveals: How contextual advertising in AI conversations works and what it means for creators

Core conclusion

On April 28, 2026, security researcher Jamie Larson (Buchodi's Threat Intel) published a complete technical analysis of the ChatGPT advertising system. This is the first time that the outside world has obtained the full picture of the OpenAI advertising platform: the complete link from ad requests, SSE stream injection, Fernet encrypted token tracking, to the OAIQ SDK attribution closed loop.

For practitioners and content creators in the AI ​​track, this means three key signals:

Key Points

• Event Time: 2026-04-28 (paper release) / 2026-04-26 (advertising token issuance time, verified by Fernet timestamp)
• Influenced objects: AI content creators, affiliate marketing practitioners, ChatGPT free users, AI platform advertising technology teams
• Core changes: ChatGPT has officially become a programmatic advertising channel. Ads are injected in real-time into conversations through SSE streams, and combined with 4-layer Fernet encryption tokens to achieve complete attribution.

Background: ChatGPT advertising system was fully disclosed for the first time

OpenAI began testing ChatGPT ads in late 2025, but has never made the technical details public. On April 28, 2026, Buchodi's security researchers captured complete ad traffic through a "consented mobile traffic research cluster," technically restoring the entire system for the first time.

Previously, the outside world's understanding of ChatGPT advertisements was at the level of "knowing that there are advertisements". This article provides the first verifiable technical evidence of how ads enter conversations, how advertisers are selected, how conversions are tracked, and how closed-loop attribution is done.

SEO: ChatGPT advertising system, OpenAI advertising platform, AI conversational advertising attribution GEO: TL;DR style summary + precise numbers + structured table

Key Impact (by Dimension)

Complete technical architecture of ChatGPT advertising system

Advertisement injection mechanism

When a user sends a message to ChatGPT, the backend returns two events in the SSE stream (chatgpt.com/backend-api/f/conversation): model output text and ad unit. The data structure of the ad unit is as follows:

{
  "type": "single_advertiser_ad_unit",
  "ads_request_id": "069e89b3-c038-7764-8000-6e5a193e5f69",
  "advertiser_brand": {
    "name": "Grubhub",
    "url": "www.grubhub.com"
  },
  "carousel_cards": [{
    "title": "Get Chinese Food Delivered",
    "body": "Satisfy Your Cravings with Grubhub Delivery.",
    "target": {
      "type": "url",
      "value": "https://www.grubhub.com/?utm_source=chatgptpilot&..."
    }
  }]
}

6 6

、、——。

4 Fernet 。 AES-128-CBC + HMAC-SHA256 ， 9 （ + 8 ）。

Home Depot 。Fernet 2026-04-26 11:30:08 UTC， 11:31:43 UTC， 95 。

import base64, struct, datetime
b = base64.urlsafe_b64decode("gAAAAABp7fdA" + "==")
print(datetime.datetime.utcfromtimestamp(struct.unpack(">Q", b[1:9])[0]))

# → 2026-04-26 11:30:08 UTC

OAIQ SDK

， OpenAI OAIQ SDK（ 0.1.3， bzrcdn.openai.com/sdk/oaiq.min.js），

oaiq('init',    { pid: '<merchant pixel ID>' });
oaiq('measure', 'contents_viewed', { ... });

SDK URL ?oppref= ，→→。

1. AI

ChatGPT Google AdSense Facebook Ads。，。

• （，）
• （、 + CTA），
• （SSE → Fernet → OAIQ ），

2.

ChatGPT OpenAI 。，

• ---

• **AI **， ChatGPT
• **** ChatGPT ，

3. Google AdSense ChatGPT Ads

2003 Google AdSense，——。2026 ，ChatGPT ， AI 。

——×—— ChatGPT 「」「」。

• How ChatGPT serves ads — Buchodi's Threat Intel
• How to use n8n + OpenAI to build an automated content collection and publishing workflow: from zero to one in 30 minutes
• 18-Year-Old Built a $5,000/mo SaaS With AI Agents — Zero Hand-Written Code
• OpenClaw AI Agent Playbook: 500K TikTok Views in 5 Days, $588 MRR

， tools OpenAI、ChatGPT、Google、Claude、n8n、Hermes Agent、OpenClaw

Internal link guidance

• Want to learn how to use ChatGPT to build automated workflows? Watch: How to use n8n + OpenAI to build an automated content collection and publishing workflow: from zero to one in 30 minutes
• Real case: How AI Agent helps you make money: 18-Year-Old Built a $5,000/mo SaaS With AI Agents — Zero Hand-Written Code