ChatGPT Ads are here – OpenAI and Anthropic are talking all about it so here’s my run down and view on what’s going on.
Coming off the back of some high value Super Bowl spots from the organisation behind the Claude LLM – OpenAI’s move to include advertising in its ‘Free’ and ‘Go’ user tiers (Go is also a new addition, potentially to facilitate a larger and slightly higher value audience pool for ads) is a significant shift in the digital World. Generative AI apps built on Large Language Models have had huge impact on search – organic and paid – and now, with the introduction of ads within the chat interface, we have turned a significant corner on the road to the AI dominated future of the internet.
The Math isn’t Mathing:
OpenAI’s transition to an ad-supported model stems from an unavoidable financial tension. The compute and energy costs of training and running AI models are eye watering, with projections suggesting the costs could exceed $1 trillion over the next decade. Even with its revenue passing the $20 billion mark in 2025, the operating expenses in Sam Altman’s (CEO of OpenAI) organisation continue to outpace earnings – a gap that needs to be closed and will need advertising revenues to help do that.
To reach its goal of $125 billion in annual revenue by 2029 – OpenAI’s own internal projections – advertising and its associated revenues are essential if the books are going to balance. While this shift isn’t surprising – all the major walled gardens have done it – what surprised me is how quickly AI has reached this turning point. For advertisers and marketers, we have never been able to get access to advertising on a channel that is so nascent and, maybe more importantly, so dynamic.
Meta, Amazon, Google – their platform proposition and feature set were much more mature when ads were introduced and, yes, there have been feed and feature changes along the way but ultimately, we are still in a world where a banner is served to a user based on what they searched or what their known demographic parameters are. ChatGPT ads represent a genuinely novel opportunity for brands – to appear in a channel that is highly trusted by the user and often seen as a system that has superior knowledge and understanding of digital information than the user themselves.
The Who, What, Where’s
The current advertising ecosystem is a restricted pilot involving major agencies like Omnicom, WPP and Adobe. Requirements include:
- Minimum spend commitments of $200k–$1m for early testing
- Live commerce feeds via platforms such as Shopify or Google Shopping
- Eligibility limited to logged-in adults in the U.S.
- Exclusion of minors and sensitive categories like health and politics
This is classic platform behaviour. High barrier beta programs create controlled environments for testing and produce clean case studies. But it also means early movers – especially major retailers – are likely to gain a disproportionate advantage in understanding the new ad mechanics.
The New Ad Formats: Conversational, Contextual, and Native
ChatGPT’s ads are designed to feel “earned” rather than interruptive with four experimental formats known at this stage:
- Sponsored Recommendations (appear at the end of a response)
- Search Ad Carousels (product strips in high-intent queries)
- Sponsored Actions (buttons like “Book” or “Try”)
- Bazaar Commerce Wrappers (technical layer for commerce execution)
This is an interesting mix of native, banner and shopping ads – all of which (at least currently) don’t interrupt your conversation but assist it. The Sponsored Action format is particularly interesting and, in my opinion, could become the most powerful direct response unit we’ve ever seen, combining relevance, timing and function all at once. The Bazaar Commerce Wrappers are also super interesting – a clear nod/step towards the agentic commerce future.
How Ads Are Ranked: Relevance Over Bids
Unlike search, ChatGPT doesn’t auction keywords. Ranking depends on:
- Conversation topic
- User intent: research vs. purchase mode
- Past interactions (hides/reports)
- Optional “Memory” settings for users
OpenAI also maintains strict Answer Independence, ensuring ads are appended after the organic response, not influencing it. It will be very interesting to see how strict this separation remains as the financial pressures and incentives start to increase. OpenAI clearly want to avoid some of the backlash that Google received as their ad placements increased – but this is wise because the AI / LLM landscape looks very different to the world of online search that Google was taking over back in the early 2000’s.
The Measurement Model: Impression Based and Privacy Preserving
Advertisers currently receive:
- Impressions (primary metric)
- Aggregate CTR
- Ad hides/reports
- Emerging engagement metric: “Ask ChatGPT” follow-ups
There is no access to conversation text, personal data, or query level logs. The “Ask ChatGPT” engagement metric could become a goldmine. A follow-up question is far more predictive of purchase intent than a click. It’s a brand new behavioural signal we’ve never had access to before and it will be very interesting to see how advertisers can leverage this to retarget or apply bid modifiers to those users who are following up.
What are the users and the competitors saying?
Early user sentiment has been negative, with many fearing ChatGPT will become “Google 2.0” with these ‘private’ conversations with the LLMs clearly holding a place of higher sentiment amongst users than simply entering a search term. But ultimately, this move was/is inevitable and every single time that platforms introduce ads, there are complaints. But there haven’t been any platforms where the introduction of ads has been terminal and often the contrary is true – just ask Netflix.
Competitors like Anthropic positioning themselves as the “adfree trusted advisor” alternative is a really interesting move – and corporate rivalries played out through media are always great to watch – Apple and Windows I’m looking at you – but Anthropic are in a very different position to OpenAI thanks to their business user penetration. Plus, with the latest reports suggesting that Claude and other AI agents are ready to lie, or worse, to hit targets and preserve themselves, I doubt the sound of glass house panes shattering will affect user behaviour too much.
The Future: Agentic Commerce and AI Driven Transactions
The long-term vision is far bigger than ads. OpenAI is building toward agentic commerce AIs negotiating, purchasing and managing transactions on behalf of users.
Protocols like:
- Universal Commerce Protocol (UCP)
- Agent Payments Protocol (AP2)
…will enable AI agents to browse inventory, check availability and complete purchases autonomously.
This is the real disruption. When AIs make decisions instead of humans, the optimisation target changes from “clickthrough rate” to “agent recommendability.” Brands will need machine-readable product data and flawless feed hygiene – or they simply won’t appear in AI mediated buying journeys.
ChatGPT’s move into advertising marks the beginning of a hybrid AI economy – one that blends relevance, conversation and commercial value. For digital marketers, this is the next frontier and is very exciting for the World of Digital Media. We will be watching the information to come out of the trials and pilots closely and are always ready to provide insight and strategies that help business navigate the latest developments in digital advertising and Paid Media.
