Report30th September 2026
About the participants
- 74 total participants:
- 63 survey respondents
- 8 one-to-one interview participants
- 3 diary study participants
- Participants were recruited via my email newsletter, posts on my social media accounts, LinkedIn ads to reach outside my network, and emails and social posts from William Joseph and other contacts
- All participants work for nonprofits, except for a small number (7 respondents) working for agencies/as consultants in the sector
How I used (and did not use) AI in this project
It feels important to be transparent and disclose when and why I chose to use, or not use, AI in on this project:
- I used AI (Claude) to help me write the full survey questionnaire. I did this by entering a very detailed prompt, containing a list of the questions I wanted to ask and the topics I wanted to explore. AI’s main role was to help me refine the wording, and generate the options for multiple choice answers. I reviewed the AI output word-by-word.
- I did not use AI to build the survey. I have tried this feature in Tally, my survey tool, in the past but have found that it makes mistakes and also that it’s hard to understand if the output is reliable and the logic is correct until it’s too late.
- I used AI (Google Gemini) to produce interview transcripts. I did a quick scan to review each transcript for overall accuracy. I did not use Gemini’s interview summaries at any point, as in my experience they lack nuance.
- I did not use AI to do the initial analysis of the survey results. It felt like an important step in understanding to turn the raw data into stats and to read through all the free text responses. I did use Claude to do some more detailed analysis and data comparisons later in the process to save time. I checked each stat that Claude produced against the original data.
- I used AI (Claude) to speed up the coding of qualitative data. By coding, I mean assigning descriptive tags to text responses and interview transcripts so that I could get a picture of the key themes. I ran a spot check on c.10% of tags to check that they matched my understanding. I found Claude to be accurate based on this sample. I only felt comfortable with taking this step because I had carried out the interviews myself and read all the survey responses, which meant that I already understood what it contained.
- I used AI (Claude) to augment my own draft notes on what the trends in the data set were. I asked Claude to review the qualitative and quantitative data and identify the key trends. I then compared these to my own conclusions. This step added value, as Clause spotted a couple of patterns that I had not noticed. However, it also came up with a number of suggestions that I disagreed with and felt did not hold up to scrutiny.
- I did not use AI to conduct research, identify extra source materials, or do any of the background reading for me. Everything you see in the bibliography was something I found through my own research or something that someone else shared with me. I have not found AI useful as a research tool, and tend to find it easier to use a search engine and communities to find material. I do not tend to use AI to summarise reading material, as I think the summaries can be inaccurate and I need to read the material myself to understand it.
- I wrote the report myself. I did not use AI to draft, write, reword, or edit it. AI-generated wording may be present in small snippets:
- Quotes – the transcripts were generated by AI
- Survey question wording – AI was used to generate this wording
- Findings – I copied some wording from Claude’s trend analysis into my first draft, but this has largely been replaced and reworded
- I got feedback on my drafts from human beings. I made the edits based on this feedback myself.
- I used AI as a sounding board when I wrote the introduction. I yapped at Claude for about 10 minutes to get all my thoughts out, and asked it to turn that into a bullet point list. I used this as a guide when I was writing the introduction. I would rather have done this with a human collaborator.
- A human editor – Ruth Oliver – proofread and copyedited the report.
Reading list and resources
AI and Accessibility: abdicating engagement?, UNESCO
AI and British Sign Language, RNID
AI and user behaviour: 8 months of meetups, Stephanie Coulshed, Scope
AI Cheerleading, AI Abstention and AI Redirection, Vanessa Andreotti
AI in Action: Insights from our AI Essentials charity bootcamps, Media Trust
AI Is Your Bridge to Smarter Content Operations, Content Science
AI Skills for Life and Work: General Public Survey Findings, GOV.UK
AI as a Healthcare Ally, OpenAI
Australian influencer Lily Jay’s tangled web of AI manipulation, ABC News
Can AI replace your content team? Hear from someone who tried (and failed), Dan Onken
CAST’s AI survey 2026: All the results — and the support available, CAST
Charity Digital Skills Report, Zoe Amar, Nissa Ramsay
Claude Skills for content design: inside Intuit’s AI build, Sarah Moes, UX Content Collective
EasyMaker: Easy Read Software to Make Documents, Photosymbols
Enabled emissions: How AI helps to supercharge oil and gas production, Global Witness
Exploring the role of AI in our design process at Co-op, Matt Tyas
Getting started is not getting it right, Nicole Alexandra Michaelis
How AI-ready content can boost policy effectiveness, Department for Business and Trade
How NN/G Uses AI in Its Editorial Process, Raluca Budiu, NN/G
How we stopped outsourcing decision-making to AI, Eva Ratcliffe, UX Content Collective
How we’re preventing AI misinformation at DBT, Giorgio Di Tunno, Neil Starr, GOV.UK
Library of AI experiments, CAST
Plausible but Not Valid: A Psychometric Audit of LLMs as Synthetic Survey Respondents, Mantas Lukauskas,Viktorija Šarkauskaitė
Practical actions for the AI search era – how to drive traffic to your website, Madeleine Sugden
Supporting your organisation with AI: A free self-serve course, CAST
The 2026 AI Index Report, Stanford HAI
The Invisible Charity – how the top 15 charities in the UK are doing in AI search, Platypus Digital
The Web Is Being Made Accessible for AI, Not People, TechPolicy.Press
Top Ten: 10 ways the sector is experimenting with AI, Ellen Smyth, CAST
Trauma-informed content and AI: Amplifying empathy at scale, Adrie van der Luijt, Button
Unlawful by design: Exposing the human rights costs of generative AI, Amnesty International
UN Scientists: AI Is Threatening Natural Resources for Billions, UNRIC
Update: AI Overviews Reduce Clicks by 58%, Ahrefs
Using AI ethically and sustainably – AI Knowledge Hub, GOV.UK
Walton Family Foundation Gen Z Research Hub, Gallup
What Makes Content Operations Successful? Full Report 202%, Content Science
Why your charity website needs to be a museum, not just a library, James Gadsby Peet, William Joseph
Read the full report
Introduction
Part 1: AI, organic search traffic and content discovery
Part 2: How charity content practitioners use AI
Part 3: The ethics of AI for charities
Part 4: Organisation-wide policy, leadership, and support on AI
About the study
This report draws on primary research carried out in summer 2026. It combines a sector-wide survey of 63 people who produce content, mostly in UK charities and non-profits; 8 in-depth interviews with content leads; and 3 diary studies, in which practitioners logged their use of AI over a week. Survey responses were analysed quantitatively, and free-text answers, interview transcripts and diary entries were coded into themes to find patterns across all three methods.
It was produced with support from William Joseph.