OpenAI introduces visual ChatGPT ads and expands measurement partners

OpenAI introduced a visual ad format for ChatGPT on 5 October, with a US test due later in the month and new measurement partnerships aimed at advertisers using a service the company says reaches 1.2 billion people each week.

Visual ads enter ChatGPT image generation

OpenAI said the initial test will place visual ads alongside the process of generating an image, rather than within the image ChatGPT creates. The ads will carry labels. The company said, as it has of its other ad formats, that advertising does not affect ChatGPT’s answers.

The format is intended to show products in use or illustrate experiences associated with them, according to OpenAI. Its planned test adds another way for advertisers to present images to ChatGPT users, rather than introducing imagery to the ad business for the first time. The format advertisers have used since OpenAI began selling ChatGPT ads in February already includes an image alongside a name, logo, headline, description and landing page, Digiday reported.

The company has also expanded ChatGPT Ads to seven markets in total across Southeast Asia and Taiwan, as AI Affairs previously reported. OpenAI says advertising allows it to reach a larger market than subscriptions alone. Its ad business recently reached a $1 billion run rate, AdExchanger reported.

ChatGPT Ads connects conversion and attribution partners

OpenAI said its Hightouch, Tealium and LiveRamp integrations will help advertisers transfer conversion data from the platforms they already use into ChatGPT Ads. The company also named AppsFlyer, Triple Whale, Adjust, DV Rockerbox, Northbeam, Branch, Singular, Kochava, Airbridge and Tenjin among partners supporting conversion, reporting and click-attribution tools across web and app campaigns.

Those connections address a problem advertisers have raised about measuring the results of ChatGPT campaigns. Advertisers have complained about limited measurement, too few formats and insufficient inventory for the amounts they want to spend, Digiday reported. OpenAI’s new format and integrations address the first two complaints. The company said it is also working with Fospha, Measured and INCRMNTAL on broader measurement approaches.

OpenAI cited early results supplied by partners. DV Rockerbox put WeightWatchers’ attributed cost per acquisition through ChatGPT Ads 15.3% below its blended paid-search benchmark, according to OpenAI. Triple Whale said 93% of Portland Leather’s visitors from ChatGPT Ads were new. WorkMagic reported that 67% of incremental purchases in a study of wellness brand Dose came from new customers.

OpenAI said its work with Haus, Measured and WorkMagic on geographically based experiments remains at an early stage. Those experiments are intended to measure the purchases or other outcomes caused by ChatGPT ads, rather than crediting an ad because someone clicked it before buying. The company’s wider set of partners gives advertisers several ways to assess campaigns using systems they already employ.

DoubleVerify and Integral Ad Science assess placement

OpenAI is developing brand suitability evaluation pilots with DoubleVerify and Integral Ad Science. The company said the partners would assess its advertising safeguards in a controlled setting without access to private user conversations. Its placement controls are designed to keep advertising away from conversations it judges emotionally vulnerable, sensitive or otherwise unsuitable.

The pilots concern where an ad appears, a separate question from whether it leads to a sale. OpenAI said it uses automated review, human oversight and continuing monitoring to identify violations of its advertising policies. For advertisers with narrower requirements tied to their own policies, it said a feature called Negative Phrases is now available to qualifying advertisers.