Advertising in Conversational AI: What businesses need to know
Introduction
Last month, OpenAI started showing advertisements on ChatGPT in the UK, letting retailers advertise directly to users of the world’s most popular AI chatbot. ChatGPT is not the first AI platform to carry ads, but the move is a significant milestone and a good reason for businesses to take a closer look at conversational AI as a new and distinctive way to reach consumers.
The commercial opportunity is clear. Rather than targeting someone simply because they searched for “kitchen renovation”, an advertiser could reach a user in the middle of a conversation about how to plan a full home remodel on a tight budget, catching them at a moment of genuine interest and in a setting that is directly relevant to what they need. A DIY retailer or home-improvement franchise appearing at exactly that point in the conversation offers a level of relevance that traditional search or display advertising finds hard to match.
This article looks at what retailers and franchisors should know before advertising on this channel, covering how the ads work, the rules that apply, and the practical issues to watch out for.
How Do ChatGPT Ads Work?
Ads appear below ChatGPT’s responses for users on the Free and Go subscription plans; the more expensive Plus and Pro plans do not include ads. Each ad is labelled “Sponsored” and is visually set apart from the chatbot’s answer.
OpenAI says that ads do not affect the content of ChatGPT’s responses and that individual conversations are not shared with advertisers. To decide which ads to show, OpenAI looks at the topic and intent of the current conversation and matches those against descriptions provided by advertisers about the kinds of conversations where their products or services would be relevant. If a user has opted in to personalised advertising, their earlier conversations may also be taken into account.
What Advertising Rules Apply?
The existing UK advertising rules apply in full. The UK Code of Non-broadcast Advertising and Direct & Promotional Marketing (the CAP Code) says that the commercial purpose behind any marketing message must be made clear to the person receiving it. This is backed up by the Digital Markets, Competition and Consumers Act 2024 (the DMCCA), which makes it unlawful to leave out important information, including failing to make clear that a message is commercial and also bans the use of editorial content to promote a product without disclosing that it has been paid for.
Under the CAP Code, marketing messages must be clearly recognisable as ads, and their commercial purpose must be obvious even when the surrounding context does not make it self-evident. Paid-for content must be labelled as such.
The key question for regulators is whether an ordinary consumer would understand that what they are seeing is a paid ad. This is especially important in conversational AI, where users tend to treat the chatbot’s responses as independent, neutral information – making it much easier for people to mistake an ad for a genuine recommendation than it would be on a traditional website or social media feed.
OpenAI currently separates paid content from the chatbot’s own responses and labels ads as ‘Sponsored’. However, that might not be enough. The Advertising Standards Authority (the ASA) has previously taken action against poorly labelled social media ads, especially those posted by influencers. The Committee of Advertising Practice (the CAP) has specifically warned against using the word ‘Sponsored’ on its own, because it considers the term too vague and open to different interpretations.
The usual rules on misleading advertising also apply: any claims made in ads must be accurate and backed up by evidence, and existing rules covering pricing, promotions, environmental claims, health and beauty claims, and advertising aimed at children all remain in place.
This is a critical point: while OpenAI controls how and where ads appear and has its own responsibility to keep advertising separate from the chatbot’s answers, retailers and franchisors are themselves responsible for making sure their ads comply with the DMCCA and the CAP Code. OpenAI’s own Advertising Policy reinforces this – it requires advertisers to make sure their ads are clearly distinguishable from ChatGPT’s responses, and it gives OpenAI the right to reject ad content that mimics ChatGPT in a way that could make users think the ad is part of the chatbot’s answer.
What About Data Protection and Privacy?
Because information from conversations, including personal data, is used to decide which ads to show and how, OpenAI must comply with UK data protection law.
OpenAI says that ads will not appear alongside sensitive topics, including personal health, mental health, and politics. Its Advertising Policy lists a range of “sensitive user contexts” that are excluded from ad targeting. This is an important safeguard, but it raises a difficult question: does the process of working out whether a conversation is sensitive itself involve handling or drawing conclusions about particularly protected types of personal data (known in law as “special category” data)?
The Information Commissioner’s Office (the ICO) has made clear that tracking technologies used for online advertising, including tools that measure whether an ad led to a sale, require user consent under the Privacy and Electronic Communications Regulations (known as PECR), and that any related handling of personal data must comply with UK data protection law. OpenAI’s Privacy Policy says it relies on ‘legitimate interests’ (essentially, a reasonable business need) as the legal basis for showing general ads, and on user consent for personalised ads. However, the Privacy Policy does not separately explain how it justifies handling particularly sensitive personal data in connection with advertising.
OpenAI has confirmed that it will not share conversations with advertisers, which limits retailers’ direct involvement in this data handling. The bigger privacy concern for retailers is likely to come after a user clicks on an ad. OpenAI offers tracking tools (a “Pixel” and a “Conversions API”) that let advertisers measure how many users went on to make a purchase. Using these tools triggers the requirement to get user consent under PECR. Retailers should check their cookie consent set-ups, privacy notices, and advertising technology arrangements before using OpenAI’s measurement tools.
Sector-Specific Rules
Industry-specific advertising rules continue to apply no matter which channel is used to deliver the ads. Financial promotions remain governed by the Financial Services and Markets Act 2000 and FCA rules, although, as mentioned, OpenAI currently does not allow financial services advertising outside the US. Retailers in regulated sectors such as food, health, and gambling should treat the ChatGPT channel as subject to the same industry rules they would apply to any other advertising platform.
Conclusion
For businesses, advertising within conversational AI is a genuinely new way to reach customers; one that offers relevant, well-timed exposure close to the point of purchase in a way that traditional search advertising cannot match. The channel is new; the rules governing it are not.
Businesses that rush to take advantage of this opportunity without first making sure their advertising is clearly labelled, accurate, and responsibly targeted risk enforcement action from the ASA, the ICO, and the CMA under the DMCCA, as well as the reputational damage that comes with it.
Moving into this channel should be treated as a trigger for a thorough review of advertising compliance not simply as an add-on to existing digital marketing activity.
