Rethinking the website in the age of AI

A group of digital, marketing and communications leads from across different sectors came together recently for a Kayo Digital roundtable (Kaizen Club) to talk about how their websites and content are holding up in a world increasingly shaped by AI.

The conversation was not about predictions or hype. It was about what these teams are actually seeing in their data, their search consoles and their day-to-day work. How traffic patterns are shifting, how users behave once they arrive, and how organisations are adapting content, research and messaging in response.

Below is a summary of the main themes from the discussion.

Traffic is down, but engagement is up

The clearest thread running through the conversation was the changing relationship between websites and search.

AI overviews, particularly on Google, are now answering many queries directly. Members mentioned this had meant a decline in page visits, with users getting what they need from the search result itself unless their question is more complex.

What is interesting is what happens to the people who still click through. Across the group, and supported by wider data from FTSE 100 companies, the picture was consistent: fewer visitors, but more engaged ones, spending longer with the content once they arrive. That shift is prompting a rethink of user journeys built for a higher-traffic, lower-intent world.

Search engine behaviour is not uniform either. Despite assumptions about audience differences, Google remains the dominant source of traffic even in sectors expected to lean towards Bing. Bing's AI results tend to be accurate but can lag behind, which makes keeping profiles current across platforms more important, not less.

When the answer matters, accuracy matters more

For organisations where safety and accuracy are heightened, AI-generated answers raise real concerns. The group spent time on this directly: how do you safeguard information when an AI system might summarise it incorrectly, out of context, or out of date?

There was no single solution, but a shared sense that the underlying content has to do more work than before. If a page is the thing being scraped, summarised and repackaged by AI, it needs to be unambiguous on its own terms.

Understanding users when half of them are invisible

Several teams are investing heavily in understanding behaviour that traditional analytics increasingly miss. Internal tools are being used to map user journeys and separate audience types, while surveys, user testing and behavioural science are all being layered together rather than relied on individually.

One particularly interesting development was the use of synthetic personas: AI-driven personas built from real user data, used to test and refine journeys before they go live. Tools like Hotjar, Clarity and Glassbox were mentioned as part of a broader, multi-method approach.

The harder problem is the growing number of users who avoid cookies or only ever interact through search, leaving a "black box" that is genuinely difficult to analyse with conventional methods.

Reaching new audiences, consistently

Attracting people who do not already know an organisation came up as a persistent challenge, particularly where a name is not intuitive to search for. This has pushed some teams to rethink keywords and branding choices that once felt fixed.

Social media engagement is also shifting. A noticeable drop on platforms like Facebook has left LinkedIn as the main channel for professional interaction, while other platforms sit largely static.

For organisations managing multiple brands or sites, the balance between central consistency and local voice came up repeatedly. Some groups run a master marketing plan with shared tools, while others give local brands more autonomy to reflect their own markets and culture. Neither approach was presented as universally right, more a question of what a given organisation actually needs.

Where AI is already proving useful

Despite the caution around accuracy, there was real enthusiasm for what AI is already enabling. Twice-yearly internal hackathons were highlighted as a way of encouraging staff to experiment and surface practical efficiency gains, some of which have turned into new internal processes or service offerings.

Looking ahead, the conversation touched on AI APIs that would let AI agents interact directly with website features, such as booking a course or making a payment, without a person navigating the site at all. That shift, if it comes, changes what a website is even for.

The same caution from earlier in the discussion returned here: automation has to sit alongside, not replace, human support. Several attendees stressed that their audiences still value direct contact, and that younger users avoiding phone calls does not mean every audience is ready to go fully automated. The right answer, for now, looks like a mix of options rather than one default channel.

Budget, grants and getting the basics right

Practical tools also had their place in the discussion. Google's PMax advertising and the idea of defensive advertising, protecting interest captured on a site from being picked up by competitors afterwards, were discussed alongside Google Grants, which gives eligible charities up to $10,000 a month in free search and Maps advertising. The application process was described as having become more straightforward in recent years, making it a worthwhile route for charities yet to take advantage of it.

Writing for two audiences at once

A recurring point tied much of the discussion together: content now has two audiences, people and AI systems, and it needs to work for both at the same time.

That means writing with longer, clearer key terms and enough context that an AI system can summarise a page accurately, while keeping the language accessible enough that a human reader does not need to work to understand it. Ambiguity that a person might read past is exactly the kind of thing an AI system gets wrong. Balancing simplicity and detail is a must. 

What this means in practice

The discussion reflected a shift that is already well underway. Websites are no longer just where people go to find information, they are also where AI systems go to learn it, and where competitors, agents and search engines now sit between organisations and the people they are trying to reach.

That brings real opportunity, but it raises the stakes on getting the fundamentals right. Content needs to be accurate and unambiguous. Research needs to account for users who can no longer be tracked the old way. And automation needs to be balanced with the human contact that audiences still rely on.

None of this is about chasing AI for its own sake. It is about making sure organisations stay clear, trusted and easy to find, however people, or their AI tools are doing the looking.

With thanks to our guests for sharing their time and experience: Matthew, Electrical Safety First, Beckie, Spirax Group,  Richard, Kayo Digital, Joy, SureServe Steve, Patoss Dyslexia Charity, Lesley, Patoss Dyslexia group, Laura, Camellia PLC.

Plans are underway for the next session, with the group keen to keep this conversation going. To join contact: harry@kayodigital.co.uk