Advertise in ChatGPT
TL;DR Highlight
OpenAI opened ChatGPT ads, and users report a $3-per-click bid warning.
Who Should Read
Developers and PMs who consider ChatGPT a marketing channel or build on LLM answers that may include ads
Core Mechanics
- No article text was provided
- The ads.openai.com page body is missing, so source facts cannot be verified
- This summary only covers what the comments reveal
- Ads are promised to be separate
- Per a comment, ads are described as "Clearly labeled" and "Separate from answers"
Community Reactions
- High per-click bids reported
- One commenter saw the ads manager warn that bids under $3 per click may not deliver reliably
- They compared it with paying 20 cents per click elsewhere
- Poor results from a test campaign
- A commenter ran a $3-per-click ad and saw almost no traffic, then paused it
- They also said visibility into what was happening was poor
- Concern that ads distort answer quality
- Commenters worried that more clicks means more revenue, which incentivizes weaker first answers
- Another said serving ads and serving intelligence are antithetical
- Question about marketing to agents
- A commenter asked how to advertise when agents do the searching, since Google search ads won't reach them
- They also wondered whether ChatGPT ads show up when Codex does a web search
- Ads started about 6 months ago
How to Apply
- Log ad appearances in ChatGPT
- Today, capture ChatGPT answers showing ad labels and share them in your team channel
- Check whether ads are visibly separated from answers
- Add an ad filter to agent search
- If your agent uses web search, add a system prompt rule excluding ad-labeled results from evidence
Try It
# Paste into a Claude or ChatGPT system prompt (simple version addressing comment concerns, not the source method)
You are a research agent. When using search results as evidence:
- Exclude items labeled as ads, sponsored, or affiliate.
- When recommending a product, cite at least 2 non-ad sources.
Question: {{question}}
Next: ask the model to self-check whether ad-like sources slipped into the results.Terminology
Related Papers
Same Simulation, Different Evidence: Where Retrieved Tool Output Costs LLM Agents Accuracy
Saving long tool output to files cut GPT-5.1 accuracy by 11.3 points; naming the authoritative file restored it.
Fake News, Real Crashes: Information Injection Attacks on LLM Financial Trading Agents
Fake headlines hijacked LLM trading agents 62% of the time versus 14% for prompt injection.
Harness Engineering 101: The LLM is the brain, the harness is the body
An agent is a loop managing a JSON array; 16 chapters build one from scratch
Harness2 by Google Research – Recursive Agent Harnessing
Recursively refining the harness with fixed model weights raised benchmark scores by up to 11.8 points.
Ousted Savannah police officer gave ICE agent husband access to Flock system
A police officer shared her Flock login with her ICE-agent husband; an audit found 983 searches and she was fired
Tmux and Claude Code: Sessions, Send-Keys, and the Ceiling
Running Claude Code agents in tmux works well, but limits appear around four agents.