AI vape-shop assistants and product claims: the 27 August CAP note retailers should read before adding chatbots

CAP has not banned vape-shop chatbots. It has not created a new vape-specific AI rule either. The practical point is simpler: if an adult UK vape retailer adds a chatbot, product finder, pod-compatibility assistant or automated recommendation widget, the claims that tool makes still sit inside the ordinary advertising and product-information rules.

That matters because these tools can sound casual while doing commercially important work. A chat panel might say a pod is compatible with a device, that a product is in stock, that an e-liquid is 20mg/ml, that an item is UK-notified, or that one kit is the best match for a customer.

Those are not just technical labels. In a retailer-controlled ecommerce journey, they can become objective claims that need evidence under the CAP Code's misleading-advertising rules, and nicotine-product content also has the extra limits in CAP Code Section 22 and Article 20(5) guidance.

This guide is for adult UK vape retailers and ecommerce operators. It is not legal advice, and it is not a sales argument for using AI. It is a pre-launch compliance audit for any tool that gives product information, recommendations or stock advice about vaping products to adult customers.

On 27 August 2026, CAP published a note called Regulating the Illusion of Intelligence in Ads. The note is not about vaping specifically. It explains how CAP would apply the existing CAP Code to ads for AI assistants, including chatbot-style tools that respond to user inputs in real time.

The vape ecommerce link is obvious once the same technology appears on a retailer website. A product finder may ask what device a customer uses, then suggest compatible pods. A stock assistant may answer whether a flavour or nicotine strength is available. A recommendation widget may rank kits by battery capacity, price, refill format or pod family.

For an adult vape retailer, the question is not whether the panel is called AI, assistant, finder or chat. The question is what it says, where it appears, who sees it, and whether the retailer can evidence the claims it controls.

That also means AI is not a workaround for vape advertising restrictions. The Vapour Hut's wider vape marketing audit checklist still applies: map the claim, map the channel, check the audience, and keep evidence before the copy goes live.

CAP's AI note starts from a narrow regulatory point: the CAP Code applies to ads, not to the underlying product being advertised. CAP says the ASA is limited to assessing how AI products are presented in ads, rather than judging every action the software might take in use.

That distinction matters for vape retailers. A compliance review should not drift into vague questions such as: Is AI allowed? The better question is: What are we claiming this tool does, and what marketing content does it output?

CAP's short practical advice for AI ads is to hold evidence for objective claims about what the product can do and to make sure the content is responsible. That sits neatly alongside CAP Code Section 3, which requires marketers to hold documentary evidence before publication for objective claims capable of substantiation, and says marketing communications must not mislead consumers materially.

So, if a chatbot is presented as checking live stock, matching every pod to every device, filtering only UK-compliant products or selecting the best device for a user's needs, those are claims to test before launch. If the data is incomplete, rules-based, delayed or manually maintained, the output should say so in plain English.

A chatbot on a retailer's own website should be treated as part of the retailer's controlled ecommerce presentation when it gives product information or recommendations. It may feel conversational, but the retailer still decides the data feed, prompt rules, product catalogue, placement and guardrails.

For nicotine-containing e-cigarettes and components not licensed as medicines, CAP Code Section 22.12 restricts marketing communications with the direct or indirect effect of promoting those products in online media and some other electronic media, except trade-targeted media. The same CAP rule says factual product claims are permitted on marketers' own websites and, in certain circumstances, other non-paid online space under the marketer's control.

GOV.UK's Article 20(5) guidance draws the same practical line for company websites and retailer sites: no advertising or promotion, but factual product information can be provided. GOV.UK gives examples of consumer-requested factual information such as price, instructions for use, ingredients, flavour, nicotine content and product-component descriptions, provided the presentation is non-promotional.

That is why our earlier guide to factual product information versus promotion is directly relevant to chatbot design. A user-initiated own-site answer such as: This pod is listed as compatible with this device family, is a different risk from a proactive pop-up saying: Limited-time deal: this is the perfect upgrade for you.

Channel also matters. CAP's online vaping advice says paid-for online display ads are prohibited for nicotine-containing vapes, and that social content actively shared to users by an algorithm can also fall within the prohibited online-media space.

That is why a tool embedded only in an age-gated own-site product search is not the same as a chatbot response repackaged into paid ads, commercial email, text messages, social posts or algorithmically shared content. For more background on that channel distinction, see our guide to vape ads on social media.

The highest-risk chatbot claims are often the most ordinary-looking ones. Retailers should audit not only headline claims about AI capability, but also every product statement the assistant can generate.

The key word is before. CAP Code Section 3 expects substantiation to exist before publication, not after a complaint. For a chatbot, that means keeping the evidence behind the data feed and the evidence behind the output rules.

UK product-limit examples should be handled just as carefully. GOV.UK/MHRA guidance says e-cigarette tanks are restricted to no more than 2ml, nicotine-containing refill containers to no more than 10ml, and e-liquids to no more than 20mg/ml nicotine strength; it also says nicotine-containing e-cigarettes and e-liquids must be notified and published by the MHRA before sale.

If an assistant claims to filter for any of those points, the retailer needs records showing how the filter is populated and checked.

The most important guardrail is that the chatbot should not turn factual product information into health, medical or cessation-style advice. CAP Code Section 22 says e-cigarette marketing communications must not contain medicinal claims unless the product is authorised for those purposes by the MHRA.

In ordinary vape ecommerce copy, outputs such as best for quitting, healthy, safe, doctor recommended or treats cravings should be blocked or rewritten before launch.

The same section of the CAP Code adds adult-only responsibility requirements. It says e-cigarette marketing must be socially responsible, must not encourage non-smokers or non-nicotine-users to use e-cigarettes, must not be likely to appeal particularly to under-18s, and must not be directed at people under 18 through media choice or context.

For chatbot design, that means avoiding youth-coded names, character-led personalities, cartoon-style avatars, game mechanics, playful challenge language and copy that talks to people who do not already use nicotine products. If a person is shown in associated marketing, CAP Code Section 22 also says people shown using e-cigarettes or playing a significant role must neither be, nor seem to be, under 25.

Price and urgency language needs its own check. GOV.UK's Article 20(5) guidance treats price as possible factual information, but lists highlighting price promotion as an example of promotion rather than information.

If the assistant pushes scarcity, countdowns, today-only messages, exaggerated savings or unsupported RRP comparisons, run it through the same discipline as any other pricing claim. Our CAP pricing and discount-presentation checklist is the better place to audit those claims in detail.

A chatbot audit does not need to start with the model. Start with the claims and the channels.

For borderline implementations, do not rely on a generic compliance checklist as final advice. CAP's online vaping advice points advertisers towards its Copy Advice service for non-broadcast ads, and specialist legal advice may be needed for complex journeys involving direct marketing, paid media, influencer content, social platform tools or medicinal-claim risk.

AI does not remove the retailer's responsibility for ecommerce content it controls. A chatbot output may be generated in a conversational format, but the product data, recommendation rules, guardrails and placement still need the same compliance thinking as any other website copy.

The practical test is three-part. First, evidence objective claims before launch. Second, keep own-site nicotine-product information factual and non-promotional where Article 20(5) and CAP Section 22 require it. Third, block outputs that drift into health claims, youth appeal, non-smoker targeting, unsupported comparisons or pushed promotional channels.

For adult UK vape retailers, the sensible next step is to run a pre-launch chatbot claims audit before the assistant goes live, then repeat it whenever the product catalogue, prompt rules, recommendation logic or marketing channels change.