What marketing teams need to know to build websites that actually work

By Matt Lemmon, Sr. Marketing Strategist

A website can no longer just be built for humans. The narrative about your brand is being decided and communicated to your potential customers by our very opinionated friends – the AI powered search engines.

The paradigm is shifting for high performing websites, and those who can extend the impact of their website into the conversations that happen before a user arrives will be the winners.

AI-powered search is changing the context: how you are discovered and how your brand is framed, this, in turn, influences what they need to see when they get there. As AI search behaviour evolves, it will be increasingly important to identify which parts of the purchase journey are influenced beyond just discovery, with trusted AI chat providing advice and recommendations. Understanding this shift — and acting on it — is one of the most important things a marketing team can do right now.

The good news is that the fundamentals have not changed as much as the headlines suggest. Getting them right has simply become more important.

The message problem nobody talks about

Here is something that surprises most teams when they go through a website project: they do not have their company message as figured out as they thought. It is often implicitly understood internally, but never quite put into words in a way that a stranger would immediately grasp.

A website forces you to make it explicit. And that process is uncomfortable, which is exactly why it is valuable.

The most common mistake is opening with a big, bold brand statement. The kind of line that sounds great in a boardroom but leaves a website visitor cold. Your visitors are not as invested in your brand as you are. They came with a purpose: to find out whether you can solve their problem. They are not there to learn your mission.

A strong opening answers three questions instantly:

•       What do you actually do?

•       What can you do for me?

•       Why you, not someone else? 

These seem like basics. Most companies assume they have them covered. A disciplined web project will tell you whether that is really true. Whether you can really explain why you are different, and crucially why a potential customer should care. A well run web project means there is nowhere to hide from these questions. It can be a great catalyst to push your team make internal positioning discussions a customer-facing reality.

“Anyone claiming they have GEO fully figured out is not being straight with you.”

Your website needs to please robots as well as humans

The visitors who land on your site today are different from the ones who arrived three years ago. A growing share of them have already had a conversation with an AI like ChatGPT, Perplexity, or increasingly without thinking about it with Google’s AI Overview, before they ever clicked on your link. They arrive with an opinion already forming. They have been pre-briefed by a source they trust as impartial (at least for now).

This changes what your website needs to do. It now serves two distinct audiences:

The first is the human visitor. For them, the website’s job is to make a lasting impression. A visual and emotional experience that validates and deepens what the AI told them. If the AI said you are a credible player in your space, your website needs to back that up with clarity, confidence, and proof.

The second audience is the AI itself. This is what Generative Engine Optimization (GEO) is about. Your website needs to arm AI systems with the right information so that when someone asks an AI about your category, your company, or your problem space, the AI can represent you accurately and favourably.

These two jobs are connected. A website that speaks clearly to humans tends to speak clearly to AI too. But they require slightly different thinking. 

How to speak to AI search (the GEO reality)

Anyone claiming they have GEO fully figured out is not being straight with you. The algorithms are evolving, and the research is still catching up. But there are patterns that are becoming clear.

Authority matters. Just like in traditional SEO, AI systems favour sources that demonstrate genuine expertise in a domain. This is not something you can fake with keyword stuffing or manufactured content. It requires real depth.

Structure matters. Content written in natural language, organised around questions and answers, is easier for large language models to parse and cite. FAQ pages and how-to guides are not just good for human readers; they are among the formats AI systems are most likely to extract and reference.

Originality matters. AI models are trained to recognise regurgitated content. Original insights, proprietary data, and genuine first-hand expertise are the materials most likely to be cited. Reworking what already exists on the web adds little value for either AI or human audiences. Limited studies have shown a short term bump in traffic that lasts around 3 months for these unoriginal pieces (find more on the SE ranking study here). 

Depth matters. Evidence suggests that once a large language model establishes that a domain has authority, it explores deeper into that site’s content. Where a good rule of thumb for traditional search was to be two clicks from the home page, AI powered search destroys this. A third of citations in LLMs come from pages 3 clicks away from the homepage, and 26.2% from even deeper (explore more in this similar web report). A shallow website with five well-written pages will be outperformed by one with real depth. Detailed guides, case studies, and technical resources provide utility and depth. The foundation you build for humans becomes the foundation AI uses to form its understanding of you.

Third-party credibility matters. AI systems do not form their understanding of your company from your website alone. They create a contextual picture, drawing from news coverage, industry publications, analyst commentary, and increasingly, LinkedIn. Being cited in credible external sources is one of the strongest signals an LLM can receive that your expertise is real and your reputation is established. This is where PR earns a new kind of return: a placement in a relevant trade publication or a mention in a credible news story does not just reach that outlet’s readers, it becomes part of the evidence base AI uses to represent you.

LinkedIn matters. It is rapidly becoming one of the primary sources large language models draw from when building a picture of individuals and organisations. Thought leadership published on LinkedIn reinforce your authority in ways that complement what is on your website. The strongest approach is to publish original thinking in both places: long-form on your site for depth and citability, and on LinkedIn for reach and the signals it sends to AI about who your people are and what they know.

Going back to basics to win the future

Here is the paradox of the AI era: the more sophisticated search becomes, the more it rewards companies that understand their users at a fundamental level. Conversational AI search is making keyword volume less reliable as a proxy for relevance and google has made it clear in their most recent announcement that more and more AI enhancements will replace the traditional search experience. 

As more searches become unique, specific, and conversational, the companies that will win are those who have genuinely thought through what their customers are asking, and why.

The best way to get that insight has not changed. Talk to your users directly. If that is not possible, talk to sales and customer support. They hear the real questions every day. They can unearth the hesitations, the objections, the moments of confusion that can make a difference to your top line. That intelligence is more valuable than any keyword tool.

Traditional keyword research still has a role, but it is shifting. Think of it as a guide to broad topic interest and to understanding your “entities”. Your entities are how an LLM understands relationships, context and determines relevance. These are your company, your competitors, your products, the industry concepts you want to own. Volume matters, but it matters less than relevance.

One emerging tool worth paying attention to: embedded AI chat on your own website. Early adopters are already using it not just to serve visitors, but to listen to them. Using visitors own inputs to their chats to understand what questions people are actually asking in real time. That data feeds directly back into content strategy, FAQ development, and messaging. It creates a more qualitative feedback loop between what you publish and what your audience needs.

Where to Start

If you are a marketing team that wants to act on this, the starting point is not a new website. It is not a GEO audit. It is a conversation about who you serve and what you uniquely offer them.

Be strict with yourselves. Who is your ideal customer, specifically? What is the offer, precisely? How does what you do make you different, and do your customers actually care about that difference? 

These questions typically require a dedicated leadership session to work through properly. If you can, bring customer voices into it directly through interviews.

From there, map the purchase journey. Particularly in B2B, understanding how decisions are made is the foundation for everything else. Your website, your content, your SEO and GEO efforts: all of it should be organised around being helpful at the right moments in that journey.

Once you have that clarity, build your content plan from the ground up based on the questions your customers actually have. Use keyword research and analytics to refine and prioritise. Then think about how every piece of content serves both your human visitors and the AI systems that are increasingly acting as the introduction between your company and your future customers.

But how do you know if it works?

Measuring GEO is still a work in progress. Dedicated monitoring tools now exist that simulate queries in AI chat platforms to track whether and how your brand appears in responses. They are worth exploring, but treat them as directional rather than definitive.

Simulated searches introduce bias and will never fully replicate the variety of real-world queries your potential customers are actually making.

For now, the most reliable signals are closer to home. Track branded search volume in Google Search Console, but not just clicks, look at impressions. Growing awareness of your brand, driven partly by AI referrals, will show up here before it shows up anywhere else. Alongside this, implement AI source tracking in Google Analytics to monitor referral traffic arriving from AI tools directly. Together, these two measures give you a grounded view of whether your GEO efforts are translating into real visibility and real visits, and a baseline to build from as measurement in this space matures.

To sum it up.

The website is not just a marketing asset anymore. In the AI era, it is the source of record — the place AI systems come to understand who you are. Make sure it tells the right story, to the right audiences, clearly.

Matt, spotted in the wild in an IRL post-it habitat

Want to know more? Reach out!