The EU AI Act: what it means for nonprofit and NGO websites
The EU just introduced new transparency rules around the use of artificial intelligence, and they are relevant to how organisations use AI on their websites.
For nonprofits and NGOs, that could mean AI assisted content generation, chatbots, AI powered website search, content libraries or translation tools. Many of these services are provided by third parties, so AI can become part of a website without an organisation necessarily developing or managing the underlying technology itself.
The rules form part of Article 50 of the EU AI Act and apply from 2 August 2026. Their aim is to make it clearer when people are interacting with AI or being exposed to certain AI generated content.
Where are organisations most likely to be using AI?
There are a few common examples.
Content generation
Using AI to draft website copy, summarise reports, create articles or produce other written content.
AI chatbots
Using an AI assistant to answer questions from website visitors or help them find information.
AI powered website search
Using AI to interpret search queries and return answers or recommendations rather than simply matching keywords.
Content and document libraries
Using AI to search large collections of reports, publications, research or other resources and provide answers based on them.
Translation
Using AI to translate website content into different languages.
Many of these tools are often provided by third parties. This is important because your organisation may be responsible for how a tool is deployed on your website even though you did not build the underlying AI system.
What do you need to do?
For most organisations, there are four practical areas to review.
1. Make AI chatbots clear to visitors
If a chatbot or virtual assistant directly interacts with visitors, the AI Act requires people to be informed that they are interacting with an AI system, unless this is already obvious from the context. The information should be provided clearly at the latest when the interaction begins.
For an NGO, this could be a chatbot answering questions about your programmes, helping someone navigate your website or providing information from a collection of reports.
What you should do:
- Review every chatbot or conversational tool on your website.
- Make it clear from the start that the visitor is interacting with AI. Do not present an AI assistant as though they are communicating with a member of your team.
- It is also worth checking the settings and documentation provided by the third party service. The fact that a supplier provides the chatbot does not remove the need to consider how it is presented on your website.
2. Put human review at the centre of AI generated content
This is particularly relevant to nonprofits and NGOs because many organisations publish information about areas such as climate, human rights, development, research, public health and other subjects that can contribute to public debate.
The AI Act requires disclosure of AI generated or manipulated text published to inform the public about matters of public interest unless the content has undergone human review or editorial control and a person or organisation has editorial responsibility for the publication. This distinction is important.
Using AI to create a first draft does not automatically mean that the final webpage needs an AI label. What matters is whether the published content has gone through meaningful human review and whether someone has responsibility for what is ultimately published.The European Commission's guidance makes it clear that simply checking spelling or making superficial changes does not constitute meaningful human review. The reviewer should examine the substance of the content and have the knowledge and judgement needed to assess it.
What you should do:
- Put a clear human review process in place for AI assisted content and make sure someone within your organisation is responsible for approving what is published.
3. Check how your AI search and content tools work
AI powered search can be particularly useful for organisations with large websites and extensive libraries of reports, research and publications.
For example, a visitor might ask:
“What research has the organisation published on sustainable agriculture?”
Rather than simply matching keywords, an AI powered search tool can interpret the question and generate an answer using information from your website and content library. The search may operate in the background rather than directly presenting itself as an AI system, so it does not necessarily create the same disclosure requirements as a chatbot. However, you still need to understand how the tool works and what information it is using.
These are the questions you must ask:
1. Where is the AI getting its information from?
2. Can it accurately identify your organisation's content?
3. Does it link back to the original source?
4. What happens when it gets something wrong?
5. Can you review or update the information it uses?
For organisations with large collections of research and publications, these questions are particularly important. An inaccurate AI generated answer about your work can be much more problematic than a search tool simply returning the wrong page.
My colleague's guide, AI Search for Non-Developers, provides a broader introduction to how AI search works and what organisations should consider.
What you should do:
- Before introducing an AI search or content discovery tool, understand what information it has access to and how it generates its answers.
- For NGOs with large collections of research and publications, this is particularly important. An incorrect AI generated answer can be more damaging when it relates to specialist research or public policy than when it is simply helping someone find a page.
- The AI Act may not require a visible AI disclosure in every search implementation, but accuracy, source control and appropriate governance should still form part of the implementation.
4. Understand how third party AI tools are being used
This is perhaps the easiest area to overlook. Your website may contain AI functionality without your organisation ever having developed an AI system itself. A search platform, translation service, content management system, personalisation platform or document library may use AI behind the scenes. That means your AI review should not stop at asking your internal team whether they use ChatGPT.
You need to understand what your website suppliers and third party services are doing as you are ultimately responsible for the content on your website.
What you should do:
- Create a simple inventory of AI powered tools connected to your website.
- For each one, establish:
- What the tool does
- What information it has access to
- Whether it directly interacts with visitors
- Whether it generates or changes content
- Who provides the underlying AI system
- What transparency measures the supplier has implemented
- Who within your organisation is responsible for the output
This should become part of your normal website governance process whenever a new tool is introduced.
What about AI translation?
Translation is likely to be particularly important for European organisations operating across multiple countries and languages.
The Commission's final guidance treats AI generated translations as an example of an assistive function that can fall within the exception to the machine readable marking requirement, where the translation does not substantially alter the original content or meaning. However, that does not mean every AI translated webpage can be treated as automatically compliant.
The separate rules around text published on matters of public interest still need to be considered. The Commission gives human reviewed AI translation as an example of content that can benefit from the human review and editorial control exemption.
What you should do:
- If you use AI translation, have a suitable person review important public facing content before publication, particularly where accuracy, terminology and context are important. This is especially relevant to NGOs where the same content may need to communicate complex research, policy or programme information across several languages.
A practical AI checklist for your organisation
Rather than trying to understand every part of the AI Act, start with your website and the tools connected to it.
1. Audit your website
Identify where AI is being used, including functionality provided by third parties.
2. Review your chatbots
Make sure visitors know they are interacting with AI from the beginning of the conversation.
3. Review your content workflow
Ensure AI generated content is subject to meaningful human review, particularly for information relating to matters of public interest.
4. Check your AI search and content libraries
Understand what information the system uses, how it generates answers and whether users can trace those answers back to your original sources.
5. Review AI translation
Make sure important content is appropriately reviewed before publication.
6. Check with your suppliers
Ask the companies providing your website search, content management, translation and other digital services how AI is being used and what measures they have in place.
7. Keep a record
Document the AI tools you use and review this whenever new functionality is introduced.
Looking ahead
For nonprofit and NGO websites, the biggest change is unlikely to be a single new piece of functionality or a new label across the website.
It is about making AI part of the way you manage your digital presence. As organisations use AI to publish more content, search larger collections of information and provide new ways for visitors to interact with their websites, it becomes increasingly important to understand what sits behind those experiences.
The European Commission has now published guidance specifically addressing the transparency requirements in Article 50, alongside a Code of Practice covering the marking and labelling of AI generated content. For organisations using AI on their websites, now is a good time to review what is already in place rather than waiting until the next website project.
My take
The move makes sense. We are already seeing a huge amount of AI generated content being produced every day, and it is becoming increasingly difficult for users to know what is genuine, what has been AI generated and, perhaps more importantly, what is actually worth their attention.
With so much low quality AI content (slop) being produced, I think there is a strong argument for content creators to take greater accountability for what they publish. AI can be a useful tool, but it should not become an excuse to fill the internet with more content for the sake of it. Organisations still need to focus on creating information that is accurate, useful and genuinely valuable to the people it is intended for.
I think this is a step in the right direction, although how well it works in practice remains to be seen. It is difficult not to think of the EU cookie rules here. The intention was right, but the way they have been implemented has often resulted in a frustrating experience for users.
The same could happen with AI transparency. If complying with the rules means adding more interruptions, warnings and friction to the user experience, people may quickly become frustrated. The challenge will be finding the right balance between transparency and a seamless digital experience.
We can help
At Liquid Light, we can review how your organisation is currently using AI across its website and the third party tools connected to it. We can assess areas such as AI generated content, chatbots, AI powered search, content libraries and translation tools, and identify where you may need to make changes to the website, your content processes or how these tools are configured.
The aim is to give you a clear picture of how AI is currently being used across your digital presence and what you need to be mindful of going forward.
-
Siddhant Patni
Account Manager
