OpenAI will start testing visual ads inside ChatGPT image generation later this month, in the US, with a first group of advertisers. The moment someone waits for a picture they asked for is now ad space. Your product image will be judged next to an image made to that person’s own prompt, and that is a harder room than a feed.
ChatGPT’s new ad slot: next to the picture being made
On 5 October OpenAI introduced a visual ad format: images showing “product inspiration, product usage, or the experiences they make possible”. It will be tested first during image generation. Ads will be clearly labeled and kept separate from the image being created. OpenAI says ChatGPT reaches 1.2 billion people a week and that ads do not influence its answers. Brand-suitability pilots with DoubleVerify and IAS, and Negative Phrases for qualifying advertisers, came in the same post (OpenAI, 5 Oct).
My read: Read OpenAI’s three words as a shot list: inspiration, usage, experience. That is lifestyle imagery, the cheapest thing to generate and the easiest place to get the pack wrong. On the Shirin Asal range reel every product image was generated with AI, and the real work was making eight categories read as one system: one grid, one timing, one tile movement. A single hero shot will look thin next to a picture the user just designed. OpenAI gives no sizes, prices or specs yet, so build the three-shot set now and wait for the spec sheet before you crop.
TikTok puts AI creative and creator content in one approval queue
At Advertising Week New York on 5 October, TikTok announced Smart+ Creative Selection. It brings a brand’s own assets, AI-generated creative from Symphony and creator content into one approval workflow, and the advertiser controls what goes live. TikTok also announced Buy Direct, one-click checkout from the For You feed, and Shopping Assistant, a conversational shopping agent (TikTok Newsroom, 5 Oct).
My read: One queue means the difference between a creator’s video and an AI film that only looks like one becomes a row in a dashboard. The Tiny Challenge borrows UGC grammar on purpose: a phone on a tripod, a classroom, a front camera. That was a creative choice, made in the brief and shipped as an ad. In an approval queue the same choice gets made by whoever clicks fastest. Write the rule before the queue fills: which looks you allow, and how each one is labeled.
OpenAI will watermark ChatGPT text in the EU
Also on 5 October, OpenAI said that over the coming weeks it will add an invisible watermark to eligible ChatGPT and Codex text in the European Union, in response to the EU AI Act. API customers worldwide can opt in. Detector access goes first to approved researchers. Its own tests show the limit: replacing 25% of words cut detection from about 92% to 17%. Supported images already carry Content Credentials (OpenAI, 5 Oct).
My read: OpenAI says it plainly: a watermark “does not measure human contribution”, and no watermark does not prove a human wrote it. So the watermark will not settle any argument about who made your script. A record will. On the nine films I added to my portfolio in September, the credit box lists idea, script, character design, location sheets and music. That list is the provenance a detector cannot give you. Keep one for every job.
What I’d do today
Take your top three products and brief one image of each kind OpenAI named: inspiration, usage, experience, with the pack exact in every frame. Then write one page on what your team will approve in an AI-and-creator queue, and who signs it. If you want the production side of that, this is how I work.
Sources
- OpenAI — Building advertising for the way people use AI, 5 Oct 2026 — openai.com
- TikTok Newsroom — TikTok Unveils AI-Powered Updates for Advertisers, 5 Oct 2026 — newsroom.tiktok.com
- OpenAI — Our approach to EU text provenance rules, 5 Oct 2026 — openai.com