AI in the cultural sector: what all the surveys tell us

Chris Unitt

If you work in arts and culture, someone’s probably asked you to complete a survey about AI recently. Another will probably be on the way in a few moments.

Over the past couple of years, we've seen studies from the National Lottery Heritage Fund, Capacity Interactive, King's College London, the American Alliance of Museums and plenty of others. UNESCO and ICOM have recently published the results of a global survey involving more than 400 museums in 90 countries. It’s all been rather a lot to keep up with.

We made our own contribution to the pile last year. As part of our Cultural Content Report 2025, we asked people working in arts and cultural organisations how they were using AI, what they were using it for, whether their organisations were paying for it, and whether they had policies in place.

With so many surveys doing the rounds, we thought it would be worth looking across them. What do they agree on? Where do they differ? And, after all this, what do we still not know yet?

One last thing before getting into it. The surveys aren’t all directly comparable. Some focus on museums, others on marketers, artists, charity workers or audiences. The questions, countries and sample sizes vary. But still, there are some pretty consistent patterns.

1. AI adoption has happened. AI maturity hasn't

It’s still early days. Looking at the numbers:

  • Cultural Content Report 2025: 77% of respondents said they used AI for at least one part of their content work.
  • HDK's 2025 research found that 67% of 92 UK arts, culture and heritage organisations had tried AI in their marketing.
  • UNESCO and ICOM's 2026 global survey found that 57% of responding museums were using AI.
  • Lauren Pope's September 2026 study of UK charity content practitioners found that 90% were using AI, with 64% using it at least weekly and 29% daily.

The studies cover different groups and define AI use differently, but the overall direction is clear. AI is no longer something that only a handful of enthusiasts are experimenting with.

But what happens when we dig a layer deeper?

  • In our research, although 77% were using AI, only 20% used it regularly for any purpose.
  • Lauren Pope's study said that, despite high levels of weekly use, 52% described their AI use as mostly individual and informal. Only 20% said it was part of a defined team or organisational approach.

So it’s fair to say that individual adoption is moving at a faster rate than organisation-wide adoption.

2. The boring uses are the most common

Go to a conference session about AI and the discussions and showcases are all about AI-generated event posters, museum guides and ways of ‘speaking with’ long-deceased historical figures who don’t get a say in things.

Whereas, most people on the ground are just using AI to help them get through their working day.

In our Cultural Content research, the most common applications were:

  • 49%: Creating written content
  • 48%: Generating ideas
  • 42%: Team productivity
  • 18%: Automating content processes
  • 12%: Creating images or video
  • 8%: Supporting online collections
  • 7%: Customer service

Remember, these figures include occasional use. Regular use was considerably lower.

HDK found a similar pattern among arts marketers. Of those using AI, 80% used it for copywriting and 54% for social content. Image and video generation came some way behind.

Lauren Pope's charity-sector research showed that the most popular applications were editing and improving copy (50%), proofreading and adjusting tone (48%), preparing initial drafts and outlines (42%), summarising information (40%), and internal communications or meeting notes (39%).

UNESCO and ICOM's museum research went beyond content and marketing. Administration was the most commonly reported application, followed by communications and research or documentation.

In other words, the AI revolution in cultural organisations is mostly about people getting past the tyranny of the blank page, summarising meetings or working through piles of documents.

There's also an interesting nuance in Lauren Pope's research: 70% of respondents deliberately avoided using AI for particular tasks. These included long-form writing, imagery, research and evidence, and sensitive health or advice content.

Which is a sign of developing maturity, if anything.

3. Organisations aren't spending much money on AI

For the Cultural Content Report, we asked how many organisations were putting their hands in their pockets for AI tools. We asked because allocating budget is a sign of commitment.

As it turned out, only 31% of organisations were paying separately for at least one AI tool. That means on top of anything bundled into Microsoft Teams, Google Workspace or similar products.

Of the respondents who were using paid tools:

  • 70% paid for a general-purpose LLM, such as ChatGPT or Claude
  • 25% paid for image or video tools
  • 16% paid for audio tools
  • 4% paid for coding tools
  • 2% paid for specialist writing tools

It suggests that much of the adoption we're seeing is happening without significant organisational investment.

Although, if two-thirds of organisations aren't paying separately for AI, it's worth asking how much use is happening through free versions, tools bundled into existing software, or staff members' personal accounts.

Which brings us to another of the pervasive findings across the research…

4. AI use is growing faster than the governance around it

This was one of the most striking findings in our Cultural Content Report.

Only 10% of respondents said their organisation had a finished AI policy.

Another 46% were in the process of writing one, while 43% didn't have one.

What’s more, 67% of respondents whose organisations had no AI policy were already using AI in some capacity.

So an absence of policy certainly didn't mean the absence of AI.

The new UNESCO/ICOM research finds much the same thing at an international scale. Although 57% of responding museums were using AI, 55% had no internal AI policy, strategy or guidelines. Only 17% provided staff training in AI.

Lauren Pope's charity-sector study adds another useful data point: 87% of respondents hadn't received formal AI training.

Which raises some awkward questions:

  • What types of information are people putting into these tools?
  • Which uses are acceptable and which absolutely aren’t?
  • Who is responsible when something goes wrong?
  • What happens if a subcontractor or event promoter uses AI in a way that clashes with an organisation’s values?
  • Are staff being given the skills to use AI effectively, or being left to figure it out themselves?

Governance doesn’t necessarily mean an enormous policy document or a committee approving every prompt. A few principles, some practical guidance and a bit of training could make a useful difference and instil some confidence.

To borrow a quote from Louise Cohen, our Senior Content Strategist, in the Cultural Content Report:

Busy teams may well have a better sense of how AI can benefit their work than senior leadership, which, particularly in larger organisations, can be slow to mobilise on policy.
But content teams would be safer and more productive if they could operate transparently and collaboratively within a set of agreed uses. They may need to instigate these conversations, and should certainly have a formative voice in them.

5. Attitudes are shifting and varied

One of the more useful aspects of the National Lottery Heritage Fund's Heritage Pulse research is that it gives us a comparison over time.

In 2023, only 24% of respondents were using AI within their heritage organisations. Around 39% saw it as presenting more opportunities than risks.

By October 2025, more than 60% thought AI offered their organisation more opportunities than risks.

Which is a pretty substantial shift in just two years. And to be fair, in that time there were leaps in adoption and the capability of the various models.

But there's a problem with talking about the cultural sector's attitude to AI as though it were a single thing. The sector isn’t a monolith. For instance:

  • Ask a marketer whether AI can help them write a first draft and you'll get one answer.
  • Ask an artist whether their work should be used to train a commercial AI model without permission and you'll get another.
  • Ask a museum visitor whether they want AI writing exhibition labels and you may get something else entirely.

Research from DACS involving 1,000 UK artists found widespread concern about consent and copyright. 74% were concerned about their work being used to train AI, while 95% thought they should be asked for permission first.

Meanwhile, the American Alliance of Museums' research gives us the public’s perspective. In its representative US adult sample:

  • 43% expected all museum content to be created by humans.
  • 29% were comfortable with AI being used for routine communications but not exhibitions.
  • 19% accepted AI for brainstorming exhibition text, but not producing the final wording.
  • 9% were comfortable with museums using AI freely, including for exhibition content.

The surveys suggest we need a more sophisticated conversation than simply asking whether people are for or against AI. The purpose, context and consequences of its use matter enormously.

And cultural organisations, whose relationships with audiences often depend on trust, authenticity and expertise, need to take those distinctions seriously.

6. It’s not yet clear where value is coming from

We have lots of statistics about who is using AI, how often they're using it and what they're using it for. But we don’t have much evidence about whether it's producing better results.

In Lauren Pope's research, 35% described AI's impact as mixed, while 41% described it as broadly or very positive.

But despite widespread use and this generally positive perception, respondents didn't provide concrete, measurable evidence of AI applications saving time or money or improving outcomes.

Interestingly, those reporting the strongest benefits were more likely to have received training and organisational support.

And the most significant opportunities may not come from speeding up existing tasks at all. AI-assisted development, analysis and automation could make entirely new kinds of work practical for smaller cultural organisations.

It would help to distinguish between what's possible, what's being attempted and what's demonstrably valuable, and then share that more widely.

What we know by now

Taken together, these surveys tell us quite a lot about where the sector has got to.

  1. Adoption is widespread, even if the headline percentages vary according to who's being surveyed.
  2. Most use is still relatively straightforward, concentrating on writing, ideas, and administrative tasks rather than deeply integrated workflows.
  3. Organisational support is lagging behind, with policies, training and agreed ways of working often absent, even where staff are already using AI.
  4. Attitudes are becoming more positive, but there are substantial differences between the perspectives of staff, organisations, artists and audiences.
  5. We’re still short of good evidence about impact.

We probably don't need too many more surveys asking exactly the same questions.

What we should ask next

We need to start understanding what happens when organisations move beyond individual experimentation and start to operationalise new approaches.

  • Which applications have become dependable parts of everyday work?
  • Which have failed to deliver?
  • Where have organisations achieved measurable improvements?
  • Where has training has made a difference?
  • What have organisations decided not to automate?
  • What new things have become possible that weren't practical before?

Those are the sorts of questions we’re interested in and we’ll be keeping an eye out for answers.

The surveys and research

Here’s the list of everything we’ve found on the topic. It’s not meant to be exhaustive and it’s a snapshot in time, which is to say we won’t be keeping it updated.

Arts, culture and heritage organisations

Charity and nonprofit content teams

Artists and creative workers

Audiences and the public

We used AI to help research this piece, find many of the reports, identify some of the themes, and pull out some numbers. Commentary and checking was provided by a real person.

Join the newsletter

We’ll let you know about useful resources for arts marketers.

We won’t send you spam. Unsubscribe at any time.