UK ad watchdog bans AI companion advert over sexualised imagery

The UK’s advertising regulator has upheld a complaint against a TikTok advertisement for the AI companion app SoulTalk, saying the ad used sexualised imagery that objectified women and promoted a harmful gender stereotype.

The Advertising Standards Authority (ASA) said the paid-for ad, seen in January 2026, showed five AI-generated female characters with their mouths open around an object cropped out of view. Other images showed hands on characters’ heads. The ad presented the images as an unsolicited “Airdrop” with “accept” and “decline” buttons, alongside a message promising a 24/7 AI assistant.

What the regulator found

The ASA said the framing and repeated use of the images led viewers to understand them as sexual. The accept-or-decline interface, it said, treated the women depicted as content to be selected and used rather than as people. The regulator also found the imagery unrelated to the product’s advertised function.

The ruling concluded that the ad was irresponsible and likely to cause serious offence. It breached rules in the UK’s CAP Code covering social responsibility, harm and offence, and unreasonable delay in responding to the regulator’s enquiries. The company did not respond to the ASA investigation.

The ASA ordered that the ad must not appear again in its current form and told Optimize Business, which trades as SoulTalk, to ensure future advertising does not objectify or sexualise women or include harmful gender stereotypes.

Why it matters

The ruling illustrates a widening regulatory focus on how AI products are marketed, not only on what their software can generate. The ASA has separately examined advertising for image-generation and companion applications, while Meta has said it has removed large numbers of ads promoting “nudifying” tools from its platforms.

Those developments do not establish that every AI companion or image tool is unlawful. They do show that the presentation of a product can attract enforcement even when the underlying service is described as entertainment or communication. Regulators are assessing the ad’s imagery, context and likely effect on audiences.

The decision also leaves practical questions for platforms and developers. Automated systems may be able to detect explicit material, but the ASA’s reasoning turns partly on implication: cropping, poses, interface design and the way multiple images are combined. That makes review of surrounding context important, particularly where synthetic characters resemble young people or where real people’s images could be manipulated without consent.

The ruling is about one advertisement and does not resolve the broader policy debate over AI-generated sexual content. It does, however, set a clear boundary for UK advertising: marketing that presents women as sexual objects, or encourages viewers to treat their images as selectable content, can be judged harmful and offensive even when the ad is generated or distributed through digital platforms.