When will the ChatGPT super-app and apps arrive in Europe and the UK?

OpenAI has given no date for the European Economic Area. ChatGPT is far above the European threshold of 45 million monthly users that triggers "very large platform" status under the DSA, which imposes reinforced obligations: the arrival of the commercial and agentic features in Europe depends on that compliance work.

Can I stop ChatGPT using my conversations for advertising?

Yes. OpenAI states that conversations are not shared with advertisers and that it does not sell the data. On the Free and Go plans (United States), ad tracking has been on by default since the update of 30 April 2026; it is turned off in Settings, Data Controls (marketing privacy and personalised ads). Turned off, you still see ads, but tied to the current conversation rather than to your history.

Is buying inside ChatGPT safe? Who handles payment and after-sales?

Payment goes through Stripe using a secure token: ChatGPT never exposes your bank details. The merchant remains the official seller ("merchant of record") — they take the payment, ship the order and handle returns and after-sales support. On the merchant side, a commission is taken per sale (refunded if the item is returned); the buyer pays no extra fee.

Do merchants pay to be recommended inside ChatGPT?

OpenAI states that product recommendations are organic, with no advertising or paid placement, ranked by relevance (price, availability, quality). The nuance: whether a merchant has enabled Instant Checkout is one of the ranking criteria, and OpenAI takes a commission on sales — an indirect commercial incentive. A paid placement format could arrive later.

Will the overhaul cut features on free accounts?

Partly, according to the Financial Times (not officially confirmed): basic chat would stay free, but the most powerful features — multi-step agents and the Codex coding tool — could be reserved for, or prioritised towards, paying subscribers. The stated goal is to convert free users ahead of the stock market listing.

What is the ChatGPT super app? Aria, ads and checkout

ChatGPT wants to become a super-app: useful assistant or shopping mall?

OpenAI wants to turn ChatGPT into a super-app: a single interface to search, create, code, book, buy — and see advertising. That is what the Financial Times revealed on 7 June 2026 (Reuters relayed the information without being able to verify it independently, OpenAI not having responded immediately). The project is said to carry an internal code name: “Aria”. But the real subject is not the name, it is the direction.

In practice, this would no longer be about offering a better chatbot but — according to the FT — the biggest overhaul of ChatGPT since its launch: an assistant that no longer merely answers, but acts on your behalf through AI agents. A senior OpenAI executive reportedly summed the shift up in three words: “Chat is dead”.

From a chatbot to a platform that acts

A chatbot advises; an agent executes. That is the whole shift. This move from advice to action is already visible in our AI agent ranking, where the stake is no longer text quality but the ability to carry out long, concrete tasks — and in the proprietary and cloud AI agents, where ChatGPT Agent, Codex, Claude Code, Copilot and Gemini Enterprise are competing to be the assistant that can act, not the best chatbot. We have already documented what this implies when agents start acting on their own.

What will change, concretely

According to the FT as relayed by Reuters, the overhaul would concern ChatGPT’s roughly 900 million weekly users, on web and mobile, in the coming weeks. OpenAI wants to push directly into the main interface: multi-step agents, its coding tool Codex, image generation, third-party applications and integrated services. Partners are already named: Canva, Booking.com, Expedia, Figma, Spotify, Coursera, Zillow.

Apps inside ChatGPT: the groundwork is already laid

This overhaul does not come out of nowhere. OpenAI has already launched applications inside ChatGPT through its Apps SDK (the former “connectors”, renamed “apps” on 17 December 2025): running an external service inside the conversation, instead of sending you off to a website.

Crucial point for European and British readers: at launch, the apps are available neither in the European Economic Area, nor in Switzerland, nor in the United Kingdom. So Europe is excluded at first. You can read that as regulatory caution; you can also read it as a two-speed AI, where the commercial and agentic features arrive in the United States first — a sovereignty question we have already raised about AI in banking and healthcare.

Why now: the stock market, and an economy under strain

The timing is not neutral. OpenAI filed a confidential S-1 with the SEC on 8 June 2026 (the company announced it itself, “we expect this to leak”). The financial press mentions a valuation of around 850 billion dollars, a listing window between September and November 2026, with Goldman Sachs and Morgan Stanley as lead underwriters — with some analysts eventually imagining more than 1,000 billion dollars, that is a trillion.

But a gigantic audience is not enough to justify those figures. According to data reported by The Information and analysed by journalist Ed Zitron — unofficial estimates, in non-GAAP terms (excluding stock-based compensation), so to be taken with caution — OpenAI would have made 5.7 billion dollars of revenue in the first quarter of 2026 for an operating margin of −122%, that is an estimated quarterly loss of close to 7 billion, and up to around 36 billion of losses over the year if the margins hold.

AI costs a fortune: compute, inference, training, research, salaries, infrastructure. That is precisely where the super-app becomes strategic. A chatbot sells subscriptions; a super-app sells subscriptions, advertising, commissions, transactions and professional tools. OpenAI no longer only needs ChatGPT to be used — it needs it to pay.

Monetisation (1): advertising

OpenAI has been testing advertising inside ChatGPT since 9 February 2026, in the United States, for signed-in adults on the Free and Go plans; the Plus, Pro, Business, Enterprise and Education plans are excluded. The pilot was then extended to Canada, Australia and New Zealand (26 March), then announced for the United Kingdom, Mexico, Brazil, Japan and South Korea. The first technical advertising partner is Criteo.

OpenAI advertises guardrails: ads labelled “sponsored”, kept separate from the answers, with no influence on what ChatGPT says, and advertisers with no access to conversations. For now, targeting stays rudimentary (country level and the context of the conversation, with no audience segmentation). But the stake is not only technical: when an assistant advises you and then displays a sponsored card just below its answer, the experience mixes advice, recommendation and commerce on the same screen. ChatGPT does not need to lie in order to become commercial — it only needs to organise attention.

Monetisation (2): built-in checkout

Second building block, already wired in: Instant Checkout (“Buy it in ChatGPT”), launched on 29 September 2025 with Stripe, on the basis of the Agentic Commerce Protocol. In the United States, Free, Plus and Pro users can buy directly from Etsy sellers without leaving the conversation; more than a million Shopify merchants are due to follow (Glossier, SKIMS, Spanx and Vuori are cited). Purchases are made item by item for now; the user pays no fee, but merchants pay a commission. On the day of the announcement, Etsy shares jumped by around 16% (CNBC): the market immediately understood that ChatGPT was becoming a sales channel, not just an assistant that helps you choose.

Monetisation (3): agents that act

The most sensitive building block is neither the ads nor the buy button: it is the agent. An agent that gets it wrong does not give a bad answer — it causes a bad action: booking the wrong hotel, buying the wrong product, accepting a hidden cost, picking a sponsored offer. Who carries that? The user, OpenAI, the third-party service, the merchant, the app developer, the payment processor? The question is not theoretical: the Grok sexualised deepfakes case showed that responsibility does not always stop at the end user. With commercial agents, it now touches purchases, payments, data and bookings.

Le Recul’s reading

For three years, ChatGPT was an assistant: it answers, explains, drafts, corrects, codes. The super-app changes the nature of the product: it becomes a place where you search, work, book, buy, compare — and receive advertising. That is not necessarily bad: a single interface genuinely simplifies life. But what changes has to be named. When the same tool advises, displays partner apps, shows adverts and lets you buy, the potential conflict of interest does not necessarily sit inside the answer — it sits in the architecture of the experience. The question is no longer “does ChatGPT answer well?” but “who benefits from the action ChatGPT is proposing to me?”.

What to take away

OpenAI is preparing, under the code name “Aria”, the biggest overhaul of ChatGPT: AI agents, Codex, images, third-party apps and integrated services (FT, 7 June 2026; relayed by Reuters, not independently verified).
The commercial building blocks already exist: apps inside ChatGPT (Apps SDK), advertising tested since February 2026 (Free/Go), Instant Checkout with Stripe since September 2025.
The apps are not available in the EEA, in Switzerland or in the United Kingdom at launch: Europe is waiting.
The engine is money: a confidential S-1 filed on 8 June 2026, a valuation mentioned at around 850 billion dollars, a listing targeted for September to November 2026 — against a backdrop of massive estimated losses (figures from The Information / Ed Zitron, unofficial).

The figure to remember: around 900 million weekly users. At that scale, the smallest commercial recommendation becomes a market.

Our verdict. “Aria” is not just an interface overhaul: it is ChatGPT moving from the status of assistant to that of central commercial platform. OpenAI states that advertising does not influence its answers — that will have to be checked in practice. Because the real subject is not the quality of the answer, but the economic interest behind the action being proposed. When the assistant becomes the shopping arcade, the question is no longer “is ChatGPT helping me?”, but “who wins when ChatGPT recommends that I act?”.

ChatGPT super-app: what OpenAI is hiding behind "Aria"