On 3 February 2026, Diplo will host the 11th edition of the Geneva Engage Award. For over a decade, this initiative has celebrated and tracked the digital pulse of International Geneva, working with communities and citizens worldwide. But this year feels different. After years of exploring new social media platforms or fresh content strategies, we are turning our attention to something foundational. Over the years, the web was a library we visited, and the communication playbook for international organisations was clear: we publish comprehensive reports, maintain a rich website as the primary information source, and ensure every official document is perfectly formatted as a PDF. Now, the primary gateway to information is shifting from search bars to conversations with AI assistants. We are witnessing a radical change in the architecture of the internet. With AI integration, the web is moving from an era of information retrieval to an era of intelligent synthesis. This brings us to our theme for 2026: Back to basics. It may seem paradoxical to look backward when facing advanced technology. However, in an age where AI generates content easily, the main function of international organisations, to serve as guardians of verified, contextualised truth, becomes more critical than ever. Diplo’s latest research on AI and web evolution points to 2026 as a turning point and suggests four deep adjustments we must make to our digital posture. For International Geneva, this is both a challenge and a real opportunity.
→ Learn more and register for Geneva Engage, 3 February
In a digital space saturated by AI-generated content, fluency is cheap. Credibility, however, is expensive.
Imagine a user asking an AI chatbot: ‘How should my country regulate AI?‘ The chatbot might provide a confident, neatly phrased answer, perhaps blending elements of the EU AI Act with voluntary US safety commitments. Yet, the risks are clear: hallucinations, missing context and incomplete evidence. When answers arrive in a single, authoritative-sounding paragraph, users may be less likely to cross-check sources than they were in a traditional search journey.
In this environment, authoritative websites become the place people go to verify, i.e., to confirm what is known, what is uncertain, and what remains debated.
A global body (like UNESCO, the ITU, or the like) doesn’t necessarily need to provide a single ‘final answer’ to AI regulation. Instead, its added value is to provide a trusted reference point: agreed guidance where it exists, clearly flagged uncertainty where it doesn’t, and transparent visibility into how different regions or countries (e.g., US vs. European Union) are approaching the problem.
We need to offer reliable reference points for AI systems to use. This means moving from persuasive writing to clear, factual information. Our data should be well-structured, time-stamped, and clearly sourced. In this new environment, an organisation’s influence will depend less on website visits and more on how often and accurately AI models use its data.
PDFs have been useful as digital versions of printed pages and help keep official documents looking the same. But now, as the web is shaped by AI, PDFs are starting to get in the way of understanding.
AI agents, like people looking for quick answers, have trouble finding meaning in long, unstructured files. If a recommendation is hidden in a 200-page report, it is almost impossible to find through a search.
PDFs will still matter for formal distribution, record-keeping, and archiving. Yet they will lose relevance as the primary interface for everyday communication. The future demands content designed for fragmentation and reuse: structured pages, clear HTML, and formats that allow both humans and machines to extract, interpret, and reference key passages accurately.
Returning to our previous example on AI governance, the most helpful resource isn’t, for instance, a 200-page ‘Global AI Governance Report’ PDF. The most helpful resource is the ability to directly access the specific risk assessment checklist or the definition of ‘high-risk systems’ within that report, without having to scroll through fifty pages of introduction.
We must move toward granular, semantic publishing. We need to break our monolithic reports into their constituent parts, recommendations, data sets, executive summaries and publish them as structured HTML. By doing so, we allow the core meaning of our work to be discovered, cited, and utilized. We must liberate the text from the format.
An AI can answer a question like ‘What are AI risks?‘ in seconds. What it often cannot do reliably is provide full context, i.e., the trade-offs between innovation and safety, how these risks impact marginalised communities differently, or how governance models have evolved over time.
This is where trusted institutions retain a clear advantage.
Rather than competing to deliver the fastest single answer, the opportunity is to build knowledge ecologies around complex issues. A well-designed website should offer a curated pathway, e.g., plain-language explanations of technical concepts (e.g., generative vs. predictive AI), comparative guides on policy options (risk-based approach vs. rights-based approach), links to national strategies and international treaties, and access to ongoing multi-stakeholder debates. Multimedia resources, FAQs, and expert commentary can help users move from a quick answer to an informed understanding.
Don’t measure your content by weight, but by how well it weaves together. Create spaces that guide the user from a spark of curiosity to genuine understanding helping them weigh options, navigate uncertainty, and trace the evidence. Your real value is in trusted curation: presenting the full picture, rather than just a single verdict.
In the world of digital policy, we often talk about the difference between a ‘soft norm’, something that is nice to do, and a ‘hard law’, something you must do. For the longest time, making the website accessible was only a soft norm. It was a gesture of corporate goodwill.
The arrival of the European Accessibility Act (EAA) in 2025 was an attempt to change this dynamic.
We are seeing a ripple effect much like what happened with data privacy a few years ago. By requiring banking, e-commerce, and transport services to be accessible to everyone, the EU has set a new global standard. This shows that web accessibility is no longer just a technical issue. It is now seen as a human right. In our digital governance discussions, we argue that if someone cannot use a government website or banking app because of poor design, they are being excluded from the modern economy. The EAA makes sure that digital participation is a standard for everyone, not just a privilege for those without disabilities.
When we consider how AI is now part of this field, we notice a familiar challenge: finding the right balance between efficiency and empathy.
AI has clearly sped up our progress. Now, algorithms can automatically write text descriptions for images or add captions to live video. This automation makes the internet much more accessible for millions of people who use screen readers. It also helps developers find errors in their code before a website launches.
However, we still need to look at these changes with a critical eye.
There is a risk that organizations might see AI as a quick fix, using automated tools to cover up deeper problems. In policy discussions, this is called ‘solutionism.’ An AI might make a website readable for machines, but it often misses important context.
For example, an AI might describe an image on a news site as ‘a group of people standing’, but miss the important detail that it is a protest. This difference is important. While AI gives us strong tools to help connect people, the real design and intent must stay human. We cannot automate inclusion; we can only use tools to help support it.
These shifts in web communication are not speculative; they are already underway.
We invite you to explore what is changing and what comes next at the Geneva Engage event on 3 February 2026. Our discussion will draw on Diplo’s in-depth analysis and insights from its research on AI and the evolution of web communication.
Let’s make sure we aren’t just adjusting to the changes, but actually setting the direction.
→ Learn more and register for Geneva Engage, 3 February.
This post is based on insights from Diplo’s 2025 research initiative on AI and the Evolution of Web Communication.
