On March 24, OpenAI made three moves. It hired Dave Dugan, a former top advertising executive at Meta, to lead ad sales. It revamped ChatGPT's shopping experience to let users upload images and search for products visually. And it petitioned the UK Competition and Markets Authority to include AI chatbots on Google's mandated search engine choice screens — the screens Google was forced to offer as a condition of antitrust enforcement. Three announcements. One business model. The company founded to ensure AI benefits humanity would like to be on the screen where users pick their search engine.

Nineteen Days

The shopping revamp makes sense only in the context of what happened nineteen days earlier.

On March 5, The Information reported that OpenAI was scaling back Instant Checkout — the feature that let users complete purchases inside ChatGPT. Conversion rates were three times lower than simply clicking out to the retailer's site. On March 19, Walmart brought its own chatbot into ChatGPT rather than use OpenAI's commerce infrastructure.

The commerce model failed. Users would discover products through ChatGPT but wouldn't trust it with their credit cards. Discovery without transaction. Attention without conversion. Foot traffic without a register.

If that description sounds familiar, it should. It's the advertising business model. You don't own the checkout. You sell the aisle. And on March 24 — nineteen days after scaling back Instant Checkout — OpenAI hired the person to build it.

Fifteen Months

Dugan's hire isn't the beginning of OpenAI's advertising business. It's the end of the question of whether there would be one.

Fifteen months from "weighing" to "hiring a sales team." The progression is legible in retrospect: explore, test, price, staff. Every phase shorter than the last. The CFO floated the idea for nine months before anyone acted. Once the ads launched in January, the infrastructure followed in weeks. Dugan's hire means the experimental phase is over. You don't recruit Meta's ad leadership to run a test.

The Filing

March 2026
Filing: OpenAI petitions the UK CMA to include AI chatbots with a search function in Google's mandated Chrome and Android default search engine choice screens
Telegraph

The CMA petition is the piece that reveals the strategy beneath the tactics.

Google has been required to show search engine choice screens on Android since 2019, following EU antitrust enforcement. Users setting up a new phone or browser see a list of alternative search engines they can select as their default. Google fought it, then auctioned the slots, then was forced to make them free. These screens are the distribution mechanism that regulators designed to break Google's self-reinforcing default advantage.

OpenAI wants ChatGPT on those screens. Not as a chat assistant. As a search engine — with a search function, product results, and now, ads. The petition is OpenAI formally telling regulators: we compete with Google. Treat us accordingly. Put us where users choose.

Why would a company voluntarily invite regulatory scrutiny as a search engine? Because the upside is distribution. Every Android phone and Chrome browser showing a choice screen becomes a potential ChatGPT user. And every ChatGPT user sees ads at $60 per thousand impressions. The choice screen is the funnel. The ads are the revenue. The CMA petition is the growth strategy.

The Model

Google's origin story: Stanford research project builds a better way to find information → adds advertising → becomes the most profitable attention business in history. The ads weren't the product. The search was the product. The ads monetized the attention that search attracted.

OpenAI's arc: nonprofit research lab builds a better way to interact with information → adds subscriptions → tries commerce (fails) → adds advertising → reaches $500 billion in projected valuation. The subscriptions funded the research. The ads will fund the scale.

And the shopping revamp completes the picture. On the same day as the ad hire and the CMA petition, OpenAI rebuilt ChatGPT shopping to let users upload images and search for products visually — a feature designed not for checkout (that failed) but for browsing. Browsing is what advertising monetizes. You don't need users to buy inside ChatGPT. You need them to look.

The company that was the mall is now selling the window displays.