UK ad regulator bans an AI girlfriend advert for selling female compliance

The Advertising Standards Authority banned a Facebook advertisement for an artificial intelligence companion app on 12 August 2026, ruling that its promise of a customisable and controllable girlfriend reinforced the stereotype that women should be submissive and compliant to men's needs.

Ruling A26-1340009 concerned EverAI Ltd, trading as AI Realities, and an advertisement for the Candy AI app seen on 13 May 2026. It invited viewers to create their perfect AI girlfriend for free, told them they could choose how she would look, and displayed menus for ethnicity, age, hair colour and eye colour alongside selectable personality traits. It carried the phrase "No limits" and stressed that the character would be constantly available and constantly compliant. One complainant said the advertisement objectified women and promoted the idea that women should be submissive. The ASA agreed, finding breaches of CAP Code rule 1.3 on social responsibility and rules 4.1 and 4.9 on harm and offence. The advertisement must not appear again in the form complained of.

EverAI argued that the advertisement was about technical capability rather than behaviour, that the instruction to control her referred to customisation settings rather than domination, that a wide range of personalities was on offer, and that the product was marketed to people of all sexual orientations. The regulator did not accept the distinction. Although the figures shown were generated rather than photographed, it said, consumers would understand an AI girlfriend as an idealised romantic partner, and the combination of controllability, permanent availability and unconditional compliance reduced women to objects valued for their appearance, their availability and their obedience. In the wording reported by UK Tech News on the day of publication, the ASA considered that the emphasis on creating a perfect girlfriend and controlling her behaviour presented the female partner as something that could be tailored to meet the user's preferences and used for their gratification.

The ruling is the third this year in which the ASA has applied rule 4.9 to a generative AI product. On 18 March 2026 it upheld eight complaints against Saeta Tech Ltd over a YouTube advertisement for the video app PixVideo which showed a woman with her clothing digitally removed under the lines "Erase everything" and "No creative boundaries". Saeta Tech told the ASA that its app did not support and was not designed to enable the removal of clothing, and said it had withdrawn the advertisement voluntarily. The ASA still found the advertisement irresponsible and likely to cause serious offence because it suggested a capability for non consensual digital alteration of women's bodies. On 17 June 2026 it upheld a complaint against Optimize Business Ltd over a TikTok advertisement for the SoulTalk chatbot, which used cropped AI generated images of women presented as unsolicited photographs a user could accept or decline. The ASA said this treated women as content to be selected and used rather than as people. That company also breached rule 1.7 by failing to respond, and the case was referred to CAP's compliance team.

Rule 4.9 is not reserved for sexual imagery. On 27 May 2026 the ASA upheld a complaint about an Amazon product listing for a radiator shelf which said "Girls can also install it easily", finding that the wording implied installing a shelf was ordinarily a man's job and that it was notable when a woman managed it. The seller replaced the line with "Easy to install, even for beginners".

The rulings sit inside a system that is now largely machine driven. Analysis by the law firm Lewis Silkin of the ASA and CAP annual report for 2025 records more than 40,000 complaints resolved, 25,397 advertisements reviewed and more than 22,000 amended or withdrawn, with the regulator's Active Ad Monitoring system scanning close to 60 million online advertisements. Proactive work has grown from 5 per cent of regulatory resource in 2012 to 45 per cent in 2025 across 36 projects, and the report flags in app advertising that risks sexual objectification, particularly where children can reach it, as an area to watch. Against that volume, the EverAI case turned on a single complaint.

Australia has no direct equivalent of rule 4.9. The AANA Code of Ethics, in force since 1 February 2021, prohibits advertising that is exploitative or degrading, defining exploitative as taking advantage of the sexual appeal of a person by depicting them as objects or commodities, or focusing on body parts where this bears no relevance to the product. Section 2.1 separately bars discrimination against or vilification of people on grounds including sexual orientation and gender. Neither clause names the harmful stereotype as a standalone category.

What the ruling does not settle is whether it changes anything. The advertisement was banned; the app was not. The ASA has no power to fine, and both EverAI and Saeta Tech are advertisers reachable mainly through the platforms that carried them. It is also unclear how the same reasoning would apply to companion products sold to women, or to same sex users, since the ASA judged this advertisement rather than the market. No regulator has published figures on how many such advertisements are running.