UK Advertising Association's New AI Guide: What SMEs Need to Know

The guide was developed collaboratively under the Government and industry-led Online Advertising Taskforce (specifically its AI Working Group sub-group), bringing together advertisers, agencies (including VCCP), media owners, tech companies (including Google), and the Advertising Standards Authority (ASA). It builds on the earlier ISBA/IPA 12 Guiding Principles for GenAI in Advertising published in 2023, translating those high-level principles into day-to-day actionable steps. It carries endorsement from Rt Hon Ian Murray MP, Minister for the Creative Industries, who described it as supporting “the Government’s ambitions to ensure advertising remains trusted and makes the most of the opportunities AI can offer.”

Eight principles that form the backbone of responsible AI advertising

The guide establishes eight core principles that provide a flexible governance framework. These are not new regulations but rather translate existing legal obligations and ethical standards into practical, implementable guidance. Each principle comes with clear, actionable steps:

1. Ensuring transparency. Practitioners should determine disclosure of AI-generated or AI-altered content using a risk-based approach that prioritises prevention of consumer harm. This is not a blanket disclosure requirement — the decision to label AI-generated content depends on context and whether consumers would be misled without that disclosure. This aligns with CAP’s existing two-question test: Would the audience be misled without disclosure? Does disclosure clarify or contradict the ad’s message?
2. Ensuring responsible use of data. When using personal data for GenAI applications – including model training, algorithmic targeting, and personalisation – practitioners must ensure UK GDPR compliance while respecting individuals’ privacy rights. For SMEs, this means checking AI tool terms of service, understanding where customer data flows, and updating privacy notices.
3. Preventing bias and ensuring fairness. Practitioners should design, deploy, and monitor GenAI systems to ensure fair treatment of all individuals and groups. AI tools can perpetuate harmful stereotypes around gender, race, and body image. The ASA has actively enforced rules on harmful gender stereotypes since 2019 and expanded this to racial and ethnic stereotypes in 2023 – these rules apply equally to AI-generated content.
4. Ensuring human oversight and accountability. Appropriate human oversight must be implemented before publishing AI-generated advertising content, with oversight levels proportionate to potential consumer harm. The ASA’s 2023 Stripe & Stare ruling established a critical precedent: automated distribution through tools like Google Performance Max does not absolve the advertiser of responsibility.
5. Promoting societal wellbeing. Avoid using GenAI to create, distribute, or amplify harmful, misleading, or exploitative advertising content. Where possible, leverage AI to enhance consumer protection and advertising standards.
6. Driving brand safety and suitability. Assess and mitigate brand reputation risks from AI-generated content and AI-driven ad placement. Ensure GenAI systems align with brand values and safety standards.
7. Promoting environmental stewardship. Consider the environmental implications of GenAI tools alongside business objectives, favouring energy-efficient options where practical. This principle reflects growing industry awareness that large language models and image generators carry a significant energy footprint.
8. Ensuring continuous monitoring and evaluation. Implement ongoing monitoring of deployed GenAI systems to detect performance issues, bias drift, compliance gaps, or other concerns requiring intervention.

What the SME edition changes — and why it matters for small businesses

The SME edition takes a “more proportionate approach” to the same framework, focusing on the principles most relevant to small businesses and removing unnecessary compliance burden.

The SME guide:

• Focuses on the highest-priority principles for smaller organisations – likely transparency, data use, and human oversight as the three most immediately actionable areas
• Provides simplified, practical implementation steps rather than comprehensive governance documentation
• Reduces the compliance documentation load, recognising that SMEs lack dedicated legal and compliance teams
• Uses accessible, jargon-free language aimed at business owners and marketing managers rather than legal professionals
• Maintains the same core ethical and legal standards as the full guide while scaling expectations to business size

This proportionate approach is significant. Many SMEs are already using tools like ChatGPT, Midjourney, DALL-E, and Canva’s AI features for ad copy, social media content, and visual assets. The SME guide gives them a clear, manageable framework for doing so responsibly – without the compliance overhead that might otherwise discourage adoption entirely.

The legal and regulatory landscape surrounding the guide

The AA guide does not exist in isolation. It sits within a layered regulatory ecosystem that SMEs need to understand. Existing advertising rules already apply to AI-generated content – the ASA has been unequivocal on this point. All ads must be legal, decent, honest, and truthful regardless of how they were produced.

UK GDPR and data protection present the most immediate legal risk for SMEs using AI. When customer data enters an AI tool – whether for personalisation, targeting, or content generation – data protection obligations apply. The ICO requires Data Protection Impact Assessments before using AI tools that process personal data, and businesses must have a lawful basis for any personalised targeting. SMEs feeding customer lists into ChatGPT or similar tools may be creating compliance exposure without realising it.

Copyright remains unresolved and poses real risks. UK law (CDPA s.9(3)) provides unique protection for “computer-generated works,” but its applicability to modern generative AI is untested. The landmark Getty Images v Stability AI case (November 2025) – the UK’s first AI copyright ruling – left key questions around primary infringement and ownership of AI outputs largely unresolved. SMEs should be aware that AI-generated images may inadvertently reproduce copyrighted material, and that ownership of AI outputs remains legally uncertain.

The EU AI Act creates cross-border obligations that UK SMEs advertising to European consumers cannot ignore. From August 2026, Article 50 transparency obligations will require disclosure when content has been artificially generated or manipulated – including deepfakes – with penalties reaching €35 million or 7% of global turnover. Any UK SME running digital ads that reach EU audiences needs to factor this into their planning.

The ASA’s own monitoring capabilities have expanded dramatically. Its AI-powered Active Ad Monitoring System now scans up to 50 million ads per year, up from approximately 30,000 complaint-based reviews in earlier years. This makes non-compliant AI-generated advertising far more likely to be detected and challenged, even without a consumer complaint.

Practical steps SMEs should take right now

For SME clients of a digital marketing agency, the guide translates into several immediate actions. First, establish a basic AI usage policy that clarifies where and how generative AI is permitted in advertising workflows. This doesn’t need to be a lengthy document — even a one-page internal policy specifying that all AI-generated content undergoes human review before publication satisfies the oversight principle.

Second, audit current AI tool terms and conditions. Many SMEs use free or low-cost AI tools without checking intellectual property ownership clauses, data processing agreements, or indemnity provisions. Some tools retain rights to outputs; others are silent on ownership. Understanding these terms is essential before relying on AI-generated content in paid advertising.

Third, implement a claim-verification step for all AI-generated copy. Generative AI can “hallucinate” – producing plausible but false claims about products, services, competitors, or market position. Every factual claim, superlative (“best-selling,” “most popular”), or efficacy statement in AI-generated ad copy must be verified before publication under the CAP Code.

Fourth, review AI-generated imagery for bias and stereotyping. AI image generators trained on internet data can reproduce and amplify gender, racial, and body image stereotypes. A quick human sense-check of all visual outputs against ASA guidance on harmful stereotypes is a straightforward safeguard.

Fifth, keep records of AI involvement in campaign creation. While there is currently no blanket UK disclosure requirement, maintaining clear records of where AI was used in the creative process protects businesses if challenged by the ASA or clients.