Web Is Not Dead: Why it still anchors Digital Impact in the AI Age

Authors Note:

Lisa Adams is the Founder of Citizen Code, an African–based technology collective building inclusive digital products for social impact. She has led the design of youth-focused platforms, AI-enabled chatbots, and web ecosystems that reach millions of young people across Africa and East Asia, with a strong focus on gender justice and digital equity.

She frequently speaks and writes on tech-facilitated gender-based violence, feminist approaches to technology leadership, and the intersection of heritage, culture, and innovation. Lisa represented Civil Society at the W20 pre-panel on the Future of STEM ahead of the G20, bringing her perspective as a technologist and advocate for accessible, community-rooted digital futures.

In many technology consults I sit in, social impact organisations feel torn between two paths: invest in the web or pivot to chat.

It is a fair dilemma, considering that budgets are often constrained and paralleled by Chat platforms starting to dominate daily life. Emerging market tools like WhatsApp and Moya, alongside AI services such as ChatGPT, have rapidly started becoming the internet for millions. It is tempting for social impact organisations to ask: why invest in websites anymore?

At this point, I urge us to take a pause.

AI does not invent knowledge. It retrieves, predicts, and recombines from what already exists. And where do most large language models retrieve from? The web.

OpenAI has confirmed that public web crawls remain a core input for today’s LLMs, with the Common Crawl dataset (3.1 billion pages, ~390 TB uncompressed) forming part of the training pipeline [1]. At the same time, research shows that only 10–40% of web-derived content actually survives filtering and curation before making it into training sets [2]. The implication is clear: if your site is outdated or poor quality, it risks being excluded. If it is invisible, someone else’s words will stand in for you.

And in sectors such as health, GBV, or education, that is not just inconvenient. It can cause harm.

A website is no longer only a public-facing product. It has become a source of truth, not just for people, but for the AI systems that translate your work into chat. Ignore it, and you surrender narrative control when vulnerable audiences are looking for critical information.

The New Imperative: Maintenance as Prevention

One of the most overlooked shifts is the need to budget for website maintenance. This is now a frontline defence against misinformation.

Too often, impact organisations invest heavily in building a site and then let it stagnate. In the AI age, outdated content will be treated as current. Broken pages undermine credibility. Gaps are filled by disinformation.

We already know how fragile this is. A recent study showed that injecting as little as 0.001% of malicious or misinformative data into an LLM’s training set can degrade its performance and increase harmful outputs [3].

Regular updates and active content governance are now essential. A well-maintained website prevents AI hallucinations and misinformation by ensuring accurate, timely content is always available to be retrieved.

If your work involves critical information on HIV treatment, menstrual health, or GBV, this is non-negotiable. Just as you would never distribute an outdated health brochure, you cannot afford a neglected website.

The State of Web in the Global South

In the Global North, it is common to say that web is finished and I have been watching this trend or shift of focus. In the Global South, that is not the case. The web is still one of the most important entry points for trust, reach, and reliable information, especially in health and humanitarian contexts.

One organisation I have worked with closely is Girl Effect, a global non-profit focused on empowering adolescent girls through media, technology, and culture. They have consistently invested in web platforms as part of their strategy. Not just chatbots or campaigns, but websites that are built to be resilient, accessible, and designed for the people they serve.

With them, our team at Citizen Code has developed a set of platforms that show how the web can still be powerful when it is context-driven:

• Teeka (India): A site that helps parents understand HPV vaccination. It simplifies complex health information and speaks directly to the concerns and stigma parents face.
• Wazzii (Kenya): A platform built in the voice of Kenyan youth. It makes SRHR content feel relevant and trustworthy by matching the culture and language of young people.
• Tukisonga (Kenya): Co-created with women’s groups. It supports women to take ownership of SRHR information and decisions, making the platform a tool for empowerment, not just information.
• Big Sis (South Africa): A digital companion for girls. The web presence anchors the project, while the chatbot offers a safe one-to-one space for questions that might not be asked anywhere else.
• Jikizinto (South Africa): A youth culture hub that combines SRHR and mental health. It is optimised for data-free access through the Moya App, which is where many young people already spend their time.

These platforms are not just websites. They are ecosystems of trust. Looking at them shows why continued investment into the web is still essential for organisations working in the Global South.

Workshops, Frameworks, and Purpose-Built Approaches

At Citizen Code, we do more than build sites. We facilitate design workshops that press the hard questions:

• How will people find and use this platform?
• Is the content worth the data it consumes?
• How do we protect sensitive audiences, such as survivors of GBV?

From this work, we’ve developed, tested, and bulletproofed frameworks for web structures in low-access contexts, humanitarian settings, and “audiences of care.” One example is our support for UNICEF Lahaa’s GBV Web Platform, where we helped in their research phase to shape a survivor-centred approach to web safety, accessibility, and care in collaboration with representatives in their target markets, starting in Afghanistan. This is what distinguishes our practice: not “one-size-fits-all” web design, but purpose-built approaches.

Design choices matter an exceptional amount in this industry of critical information. Research shows that well-built Progressive Web Apps can reduce bounce rates by over 40% compared to mobile websites, while driving big gains in engagement and page views [4][5]. In other words: context-aware design is not just ethically sound, it also performs better.

A Call to Social Impact Organisations

So yeah, Chat and AI platforms dominate engagement, and for many, apps like WhatsApp or Moya are the internet. But I am certain that this is not a rivalry. Not just yet.

Chat remains the distribution and engagement layer, however, Web remains an unconditional anchor for credibility, and responsibility. Without the anchor, chat has nothing stable to distribute.

So ask yourself:

• Is your website actively serving as a trusted source of truth?
• Is it strong enough that AI systems will retrieve your words, not someone else’s?
• Or is it idle and quietly weakening your impact?

Your Social Impact website is not just an artefact & we probably shouldn’t jump too soon on writing web obituaries. It’s evolving in its relevance as an anchor and barrier against misinformation while still being a foundational component of your digital presence and our AI future.

Someone told me recently “Just because you cannot always see the foundation does not mean the walls do not need it.”

The web is not dead (for now). It is alive, the relevance is evolving, and it is still essential for digital impact in the Global South.

If you are interested in working with us on reliable tested web frameworks for your organisation we are happy to share our open-source tooling and github repositories or assist you in building your digital ecosystems through web development, strategy & research or content.

References

1. OpenAI Blog — How much data from the public internet is used for training LLMs? link
2. Gao, L. et al. (2022). The Pile: An 800GB Dataset of Diverse Text for Language Modeling. arXiv:2211.04325. link
3. Futurism — Researchers Warn That Compromised Training Data Could Poison AI Models. link
4. Piwik PRO Blog — Measuring Progressive Web App Performance: Analytics Essentials. link
5. Straits Research — Progressive Web Apps Market Size, Growth, Trends (2023–2030). link