AI is everywhere. And that’s both exciting and exhausting.

If there was one thread running through every session, discussion and side conversation across both days, it was artificial intelligence. This was hardly surprising, but what stood out wasn’t the hype, or any superficial push to use AI in any way imaginable as has alarmingly been the case the last few years, but the honesty. There was a shared sense that the AI landscape is overwhelming not because it isn’t relevant, but because it’s so relentlessly fast-moving, has an extremely wide application scale and very high, positive or negative, impact.

There were several ideas and articles shared, highlighting some “uncomfortable” truths. For example, how the Harvard Business Review argues that the expectation that AI will simply lighten the load hasn’t quite matched the reality of integrating it into teams, workflows and systems, especially in complex, devolved organisations like universities, but the result has been quite opposite: to intensify it.

AI doesn’t reduce work – it intensifies it – Harvard Business Review

Keeping with the AI theme, there was very interesting discussion related to its impact related to digital accessibility. I have published my thoughts in a separate blog post, earlier this week.

AI and accessibility: keeping the human in the middle – Blog post by Stratos Filalithis

On of my personal takeaways is the confirmation that taking our time to adopt and adapt with cutting edge technologies, which is often the case with the Higher Education sector, it creates a lot of “Fear of Missing Out” moments, but it has a lot of benefits too. We get to learn from others’ experiments, mistakes and successes, putting us in a lot better position to get the best value of our investment. I am expecting something similar to happen with AI, too.

From SEO to GEO and AEO. The era of the answer engine is here.

There were several sessions covering the theme of discoverability in the age of AI. Matthew McQueeny iterated the reality that web search as we know is changing. AI-generated answers are increasingly replacing the traditional list of blue links, which means website traffic from search engines is declining. Adobe’s winter holiday 2025 metrics, a typical season of web search getting busier due to the search for gifts, showed that AI referrals had increased by 700%, with AI-referred visitors converting 30% better than those from traditional search. Bain & Company have shared more detailed analytics about how the average user behaviour has changed dramatically, for example approximately 80% of web search users rely on AI summaries at least 40% of the time.

Goodbye Clicks, Hello AI: Zero-Click Search Redefines Marketing – Brief from Bain & Company

To optimise content discoverability for AI, the traditional Search Engine Optimization (SEO) is transforming to what practitioners are calling Generative Engine Optimisation (GEO) or AI Engine Optimisation (AEO). William Borg Barthet’s lightning talk summarised that good SEO is good AEO. The familiar techniques of structured data, clear metadata, use of schema.org markup, and accessible writing are not going away, to the contrary: they matter more. One phrase from Tom Cranstoun’s session on Machine Experience paraphrased the title of a UX classic book by Steve Krug: “Don’t Make the Agent Think”, which translates to: if your content isn’t structured for humans to grasp at a glance, it won’t be structured for AI agents either.

With AI taking control of the narrative, we need to rethink our content

The rise of AI has made us ask some very interesting questions. For example, who do we design and build websites for? Surely, our end users and audiences are at the top, but how should we approach the fact that a rising amount of them interact with AI summaries, and choose not to visit our websites?

Kate Kenyon presented on how website content has been designed and built for screens, for human eyes to read and hands to scroll and click. AI is not interacting with the content in the same way. It scans it, looks out for clear content structure, context and relationships, usually hidden from the human eye, and uses them to respond to user questions, setting the narrative using our content.

I was encouraged to think that our central website platform, EdWeb 2, has been designed using structured content in its pages by default by using the Paragraphs module in Drupal.

Digital Sovereignty, it’s closer to home than you think.

Living and operating in a, quite fluid at the moment, geopolitical environment has triggered questions about ownership of digital platforms. Simon Jones delivered a lightning talk on digital sovereignty that sparked a lively conversation. Following political decisions and the shift of alliances, the physical boundaries between countries are starting to extend to the digital world, too, sadly. About 80% of EU governments currently rely on US-owned platforms and tools for their digital infrastructure, and some of them are already considering the associated risks, or actively moving away from them.

France, for example, is actively moving away from Zoom and Teams, building its own government platform, LaSuite. Other EU-based technologies that are considered are Matomo for privacy-first analytics, Proton for encrypted communications and Bunny.net as a content delivery network.

LaSuite – France’s public digital services suite

Matomo – privacy-first analytics

Proton – Email and encrypted communications

Bunny.net – Content Delivery Network

This made me think about our own reliance on non-UK based technologies, and the potential impact of that technology not being available or shifting its direction towards more privacy invasive strategies. It was quite a scary thought to make.

If any of my reflections have sparked a thought, a challenge or a question relevant to your own work, I’d love to hear from you. Use the comments below, or get in touch with me directly at stratos.filalithis@ed.ac.uk. These conversations are a lot more useful and productive when they go out of the room and evolve.