UK marketers fail to check AI outputs, survey finds

Reading Room found that only 27% of UK marketing leaders always review AI-generated outputs before using them. The survey also found that 12% rarely or almost never review them.

The figures come from a survey of 75 senior marketing leaders at UK organisations with revenue above £30 million, conducted as part of wider research into AI use among medium-to-large businesses.

Most respondents fall between those two groups, suggesting checks on AI-generated material are often applied inconsistently rather than built into workflow. For marketing teams, that matters because much of their work appears in customer-facing channels, where errors can quickly become public.

The research also pointed to a broader gap between AI adoption and results. A third of marketers said AI's impact had been lower or significantly lower than expected, while only 4% said all of their AI programmes had moved beyond the pilot stage.

Governance gap

Security and compliance concerns were cited as the biggest barrier to AI programmes delivering greater value. Yet the survey suggests those concerns are not always matched by systematic review of outputs before publication or use.

The tension comes as businesses face growing scrutiny over inaccurate or misleading material produced with generative AI. Public examples have already emerged in the UK, including withdrawn reports containing fabricated or inaccurate information.

Reading Room argued that pressure on marketing teams to produce more content with fewer resources may be contributing to the inconsistency. Teams are being pushed for scale and speed while also being expected to manage risks around accuracy, bias and compliance.

Amanda Falshaw, AI Enablement Lead at Reading Room, framed the issue in brand terms.

"Marketing is usually the shop window to an organisation - often where outputs are more likely to be seen by the public, customers, stakeholders and journalists. This means the reviewing gap becomes less of an internal process problem and more of a brand and reputational issue. As a minimum, we should all be asking of AI outputs: what's missing? What assumptions have been made? How would we verify this? Would I be comfortable putting my name to this?" Falshaw said.

Content risks

The findings add to a wider debate over how companies govern the use of generative AI in functions that publish large volumes of external material. Marketing has become one of the earliest and most visible corporate use cases for AI tools, spanning copywriting, campaign development, search content and personalisation.

That visibility also leaves the function particularly exposed when mistakes slip through. Inaccuracies, invented claims, and repetitive or generic language can weaken trust in brand communications, especially when AI output is published with limited human checking.

Falshaw said the issue extends beyond factual mistakes to the quality and distinctiveness of brand communications.

"From a marketing perspective specifically, much of today's content is suffering from pattern saturation and when nobody with the right expertise is stopping to critically review the output, this sameness starts to creep into how brands communicate. This then undermines distinctiveness during a time when having a recognisable human voice is becoming a critical marker of brand trust," she said.

Scaling problem

The survey was part of a broader study of 150 digital transformation leaders, split evenly between senior digital and marketing roles. The low proportion of respondents who said every AI programme had scaled beyond the pilot stage suggests many organisations are still struggling to move from experimentation to routine use.

That may help explain the mismatch between interest in AI and reported outcomes. If governance processes are weak or uneven, companies may be less willing to expand deployment into higher-risk or more visible areas. Equally, if teams are adopting tools faster than review practices can be formalised, pilot projects may remain stuck before wider rollout.

For marketing leaders, the data points to a basic operational problem rather than a lack of awareness. Respondents identified security and compliance as major concerns, yet only a minority said they always reviewed outputs. That gap suggests some organisations have yet to embed consistent human oversight into day-to-day use.

Reading Room said marketers should build a routine habit of scrutiny around AI output, checking for errors, bias and compliance issues before material is used externally. The survey found that just 4% of marketing leaders said all of their AI programmes had successfully scaled beyond the pilot stage.