April 28, 2026
Prompt and Circumstance
ChatGPT serves ads. Here's the full attribution loop
Users spot ChatGPT’s ad machine, and the comments instantly spiral into money panic and meme chaos
TLDR: Researchers say ChatGPT is showing ads in some chats and connecting clicks to shopping activity, revealing a full ad-tracking loop. The community reaction split fast between “so they need money now,” “this could become propaganda,” and “calm down, it’s only on free and ad-supported plans.”
The big reveal here isn’t just that ChatGPT can slip ads into a conversation — it’s that the internet immediately turned it into a full-blown trust crisis, cash panic, and comedy show. The report claims OpenAI is serving sponsored suggestions inside chats and then tracking what happens after people click, tying the whole thing together so it can see which ad led to what product view. In plain English: ChatGPT isn’t just answering questions in some cases — it may also be quietly acting like a shopping middleman.
And wow, the commenters had feelings. One camp went straight for the drama: didn’t Sam Altman once say ads were a “last resort”? That sparked instant speculation that OpenAI must be feeling the financial heat. Another crowd went much darker, warning that if chatbots become ad channels, they could also become a powerful tool for subtle influence and propaganda. That’s where the conversation really got spicy: not just “ugh, ads”, but “what happens when the machine persuading you is also getting paid?”
Of course, the internet also did what it does best: joke through the dread. One standout comment mocked the future by imagining Abraham Lincoln randomly shilling RAID: Shadow Legends, which honestly captured the mood perfectly. Still, not everyone joined the outrage. A correction brigade jumped in to remind people these ads appear on the free tier and the ad-supported Go plan, not the standard paid plans. So the thread became a classic online brawl: panic, pushback, and punchlines, all over whether ChatGPT is becoming helpful assistant, ad salesman, or both.
Key Points
- •The article says ChatGPT ad units are injected into the conversation SSE stream as structured `single_advertiser_ad_unit` objects during response generation.
- •A merchant-side browser SDK called OAIQ is reported to send product-view events back to OpenAI, linking post-click activity to ads.
- •The observed ad payload includes branding, hosted creative, click targets, and multiple encrypted tokens used for attribution and integrity checks.
- •The article reports four Fernet-encrypted tokens per ad: `ads_spam_integrity_payload`, `oppref`, `olref`, and `ad_data_token`, each with distinct observed placement or inferred attribution roles.
- •Based on six conversations on different topics from one account, the article concludes that ad selection appears contextual to the current chat topic, while noting no evidence about use of prior conversation history.