When the Algorithm Gets It Wrong: AI Deepfakes, Pride, and the Trust Deficit

The ASA’s new guidance on AI-generated advertising reveals a deeper crisis of authenticity.

The Advertising Standards Authority published two guidance notes on the same day this week. One was about celebrating Pride Month responsibly. The other was about AI-generated deepfakes in advertising. They might look like separate bulletins aimed at different problems. They are not. They are the same problem viewed from different angles: how do consumers decide what to trust?

And the answer to that question is more psychological than most marketers realise.

The Two Bulletins, The One Problem

The Pride Month guidance, published 11 June 2026, warns advertisers against “rainbow-washing” — the practice of slapping a rainbow on your logo for June while doing nothing substantive for LGBTQ+ communities the rest of the year. The ASA’s message is blunt: genuine representation matters, tokenism causes offence, and brands should engage with communities rather than merely performing solidarity.

The deepfake guidance, published the same day, addresses a different kind of performance. AI can now generate convincing celebrity endorsements, fabricated testimonials, and synthetic imagery that looks real enough to deceive. The ASA’s position is that the CAP Code is media-neutral: if an ad misleads, it breaches the rules, regardless of whether a human or an algorithm created it. The advertiser is always responsible. No exceptions, no escape keys.

Different topics. Same underlying question: when a consumer encounters a brand message, how do they determine whether it is authentic?

Dual Process Theory and the Authenticity Problem

This is where psychology earns its keep.

Dual Process Theory, one of the seven pillars of the STAR Framework, distinguishes between two modes of human cognition. System 1 is fast, automatic, and emotional. It processes information intuitively, makes snap judgments, and moves on. System 2 is slow, deliberate, and analytical. It scrutinises, questions, and weighs evidence.

Here is the problem: deepfake content is designed to exploit System 1. You see a familiar face endorsing a product. You feel a flicker of recognition and trust. You scroll on. By the time System 2 kicks in to ask “wait, did that person actually say that?” the emotional impression has already formed. The cognitive damage is done.

This is not a hypothetical concern. The ASA’s own ruling on Polyverse Inc demonstrates that AI-generated images can cause real harm before anyone has time to verify them. The Cosmos Oyun Yazilim case showed AI-produced advertising that objectified women, breaching the Code. In both instances, the content operated at System 1 speed: it was seen, felt, and internalised before rational analysis could intervene.

Rainbow-washing operates on the same principle, albeit at a lower stakes level. A consumer sees a rainbow logo. System 1 registers “this brand supports my community.” The emotional connection forms instantly. System 2, which might ask “but what has this brand actually done?”, is not invited to the party. The impression is cached. The brand benefits from an association it may not have earned.

In both cases, the gap between appearance and reality is the gap where trust is built or broken.

Why Socialisers Feel This Most

The STAR Framework identifies four core consumer types, each driven by a fundamental psychological need from Self-Determination Theory. The Socialiser is driven by relatedness: the need to belong, to connect, to be part of something meaningful. Community identity is not peripheral to the Socialiser. It is central.

This makes the Socialiser archetype uniquely sensitive to inauthentic Pride messaging. Not because they are cynical, but because social identity violations feel personal. When a brand performs solidarity without substance, the Socialiser does not just notice — they experience it as a betrayal of the community they belong to. Social Identity Theory tells us that group membership is a source of self-esteem and emotional security. A brand that exploits that membership for commercial gain without honouring it is not just being dishonest. It is taking something that matters and cheapening it.

The deepfake problem hits differently but arrives at the same destination. When a consumer discovers that a testimonial was fabricated, or a celebrity endorsement was synthesised by an algorithm, the reaction is not merely “I was misled.” It is “I was manipulated.” For the Socialiser, whose trust is relational rather than transactional, that distinction matters enormously.

The Verification Gap

The Advertising Association published a Best Practice Guide for the Responsible Use of Generative AI in Advertising in February 2026, developed with input from the ASA. It emphasises transparency, fairness, human oversight, and harm prevention. These are sensible principles. But they address the supply side of the problem — what advertisers should do — without fully addressing the demand side: what consumers need.

The consumer’s problem is not that they lack information. It is that they lack tools to verify authenticity in real time. The ASA’s Active Ad Monitoring System can identify non-compliance at scale, but consumers cannot. They are left with System 1 processing and a vague sense of unease.

This creates a genuine segmentation challenge for marketers. Different consumer types respond to authenticity signals differently:

Socialisers look for community endorsement and relational sincerity. They want to know that a brand’s Pride commitment extends beyond June.
Thinkers, driven by competence, want evidence. They are more likely to research a brand’s claims and less likely to be moved by emotional appeals alone.
Adventurers, driven by autonomy, may be less concerned with brand authenticity and more interested in whether the product delivers on its promise.
Realists, driven by security, are the most cautious. They default to distrust until a brand has demonstrated consistency over time.

A single authenticity strategy will not land equally across these types. The brands that get this right will segment their trust-building efforts just as carefully as they segment their product messaging.

What This Means for Marketers

The ASA’s twin guidance notes point toward an uncomfortable truth: in an era where AI can fabricate sincerity at scale, the brands that win will be the ones whose authenticity is verifiable, not just performative.

This is not a compliance problem. It is a strategic one.

Compliance says: do not use deepfakes to mislead. Strategy says: build the kind of brand that does not need to mislead. Compliance says: do not rainbow-wash. Strategy says: do the work that makes the rainbow meaningful.

For marketers, the practical implications are clear:

Audit your AI outputs before they audit you. The ASA has made clear that advertisers are responsible for every piece of content, regardless of how it was produced. Human oversight is not optional.
Segment your authenticity signals. Different consumer types trust different things. Data for Thinkers, community proof for Socialisers, product integrity for Adventurers, consistency for Realists.
Extend your commitment beyond the calendar. If your Pride engagement starts and ends in June, you are not building trust. You are spending it.
Make verification easy. In a world where System 1 processing dominates, the brands that give System 2 something to work with — transparent sourcing, verifiable claims, genuine community partnerships — will earn the trust that others can only simulate.

The ASA’s two bulletins are a warning. Not just about compliance, but about the direction of consumer trust in an AI-mediated world. The algorithm can now generate authenticity on demand. The question is whether your brand’s authenticity can survive being tested.

David Chadderton spent his twenties and thirties teaching people how to make life-or-death decisions at forty thousand feet. He now applies the same principles to consumer psychology, which, depending on the brief, can feel equally high-stakes. He’s the creator of the STAR Framework and the author of The STAR Framework: Rewriting the Rules of Consumer Engagement (NYC Big Book Award 2025), The STAR Operating System: Decode Mindset, Understand Motivation, Transform Human Behaviour, and Dear Algorithm, It’s Not Me, It’s You. By day, a Chief Marketing Officer. By night, a behavioural science obsessive who writes The Unoptimised Human because he can’t stop thinking about why people do what they do.