Videos — Page 3

Lightly edited for clarity.

When you use AI in a chat interface, you’ll see that it sometimes searches the web. When it does that it creates a query, retrieves text, and then adds that text into the context window, so that it has richer context to produce a response. Large language models predict language, based on the context they’re given, so connecting a model to the web gives it additional context.

If the retrieved content is clear and factual and complete, then the output from the LLM is usually pretty solid. But if the content has gaps, or it’s outdated or it’s conflicting, then it’ll fill that in with statistically probable tokens, but not necessarily correct ones.

So even when you see fewer people visiting your website, they are still likely consuming that information. But through AI, your content is being extracted and summarized, and recombined and presented to the user. So this means that your website isn’t just a place for people to read, and transact although it will likely continue to be that for years, but your website becomes a source of truth for other systems, that are presenting information to consumers based on your content.

So websites can’t just be brochureware, they become more like an encyclopedia, for the niche that your website covers. If your website’s vague or incomplete, then models are going to guess, and if your site is structured, explicit and authoritative, then models can reference it properly.

And this is also why, a lot of your traditional analytics is starting to feel wrong. Influence in the digital space doesn’t map cleanly to visits, time on site or marketing funnels anymore. AI can shape decisions without a consumer ever visiting your website. And even when they do visit your website, they’re often dropped into the middle or the end of the funnel, and then attribution becomes very difficult.

Lightly edited for clarity.

When you notice fewer visitors coming to your website, your instinctive reaction is that your website matters less. But, part of the problem is that we’re measuring visitors based on clicks, and people are now getting information through AI, and while they may not be visiting your website, your content may still be reaching them, but the process and the way to measure it has changed.

So SEO or search engine optimization, was never really about clicks. It was about people finding your website, and SEO determines whether content is indexed and retrievable, and surfaced when someone goes looking for information. And that interface is changing, but the need has not.

When an AI chat searches the web, it relies on existing search infrastructure to retrieve content, and if your content’s not discoverable, it never enters the context window. So SEO now determines whether your content shows up in AI at all, and all of that happens upstream, ahead of any visits to your site.

So, you see declining traffic and assume your content is losing relevance. But what’s happening potentially, is that instead of a person visiting your site, and reading the content on a page, there’s an agent doing that, extracting parts of that content, combining it with other content, and producing a response. So your content can influence people, without producing visits to your website.

And when people do land on your site, they’re dropped into the middle or the end of the process, and funnels, and attributions start to look really strange in your analytics.

And now this is where GEO, or generative engine optimization starts to come in. It does not replace SEO it extends it. SEO makes content retrievable, and GEO influences how that content is summarized and weighted, and represented once it’s inside a model’s context window.

If SEO fails there’s nothing to optimize for generation. But if generation isn’t considered, the content that is retrieved may have gaps or conflicting information, and the user can receive an answer that is coherent but incorrect.

Ignoring SEO means your content isn’t seen, ignoring GEO means your content may be seen, but not accurately reflected. And this is why the website doesn’t become less important, even though traffic declines. It becomes harder to measure and harder to manage, but also more important.

Lightly edited for clarity.

As AI tools increasingly become the way people ask questions, and make decisions, websites start to play a different role. They’re no longer just destinations for people to visit, they’re reference material for systems that answer on your behalf.

When AI generates a response about a product or your industry, or your expertise it can assemble that response, based on content that it retrieves from the web. And that means that your website is increasingly, being read by machines before it’s being read by people.

And machines don’t infer intent or tone, or credibility the way that people do. They work with what’s explicit. If information is missing or vague or outdated, the system will fill in gaps, using whatever else it can find or make up.

And if definitions conflict across pages, the system will average those. And if details are buried in marketing language, they may not be picked up at all.

So content quality stops being just a branding concern, and websites have to behave more like encyclopedias for their domain. Clear definitions explicit claims, consistent language and up to date facts.

You’re no longer just managing how your brand looks, you’re managing how your brand is represented when someone else, or something else is doing the explaining. And that’s a harder problem than page design, or conversion optimization, because once AI systems become the interface, your website becomes the source, and whatever gaps you leave there won’t stay empty. They will be filled in by the model.

Lightly edited for clarity.

Websites have been primarily for people, so content has been structured around pages and funnels, and search terms. You create pages to rank on specific terms, and to guide people through a journey. And that is still part of the picture, because people will continue to use websites.

But an AI world changes how that content needs to be organized. AI systems don’t read your whole website the way a person might. They pick up snippets and headings and definitions, and question answer pairs, that they can insert into a generated response. Content that’s not structured in a way that makes those pieces obvious, may be overlooked by agents.

And this doesn’t mean that you need to abandon topics and sections, and page hierarchies. It means, you need to rethink how you surface the information, that’s most likely to be used. Pages need clear, question focused headings and direct answer blocks up front, modular sections and bullet lists that have facts or steps, and structured data where appropriate, Schema.org and meta tags and llms.txt.

So machines can interpret the meaning of your content, not just the words. And in practical terms that means that your site’s content, architecture, and on page structure are now part of your visibility strategy.

Traditional SEO encourages you to match a search term. GEO encourages you to match the kinds of questions, AI models are likely to answer. In other words the job’s not just about having good content, it’s about organizing that content, in ways that AI can reliably extract and interpret.

Well organized content makes your expertise machine readable, and increases the chances the AI system will use your material, accurately and frequently. And the content itself still has to be valuable to people. Machines may read first, but humans still decide whether to engage or convert, once they land on your site.

So the organization you adopt has to serve both sides, structured clarity for AI and intuitive navigation for people. And this is the same content that you already have, but it is organized in a way that reflects how it’s being consumed. And that means you need to control the clarity, and the accuracy of your content, at the same time, that you need to be producing much more personalized content, to rank well in GEO.

Lightly edited for clarity.

As people increasingly use AI, traditional website analytics becomes more and more incomplete. The process has changed and the measurement has to change as well. Traditional analytics assumes a human visitor, but when AI tools sit between your content and your audience, that model kind of falls apart.

People can research and compare options and form opinions using AI, without ever visiting your website, and your content can shape that decision, even though you never see the session. And when they do visit your website, it’s often in the middle or the end of the process. From an analytics perspective, that looks strange and attribution feels unreliable, because funnels assume linear journeys, and attribution models assume visibility.

But AI mediated interactions hide large parts of that process, and you start seeing symptoms. Traffic declining but maybe your impact’s not, or conversions that don’t line up with campaigns, or content that doesn’t perform but keeps showing up in AI answers. And nothing’s necessarily wrong with the content, but the measurement model is wrong. It is a different way of thinking, when your content is being used outside your site.

And this is one of the things we do for our clients. We help them understand how their content shows up in AI responses. We help to connect SEO, GEO, and analytics into a coherent picture and help them measure success, in this new environment.

You need to know where your content’s being used, how it’s being represented, and what signals still matter when clicks aren’t the whole story, because you are still trying to be visible and accurate, and influence decisions but the interface and the process, and the ways to measure and optimize have all changed.

Lightly edited for clarity.

Last week we hosted an event called Reply Day. We invited a group of executives in energy, manufacturing, financial services, retail to the Porsche Driving Experience, and yes, we drove cars around a track, which is obviously very fun, but that wasn’t really the point. The real reason they came was to talk about what’s changing inside their organizations and how fast it’s changing.

So we spent the day discussing how AI is reshaping customer discovery, how evaluation of products and services is shifting, how staffing models are compressing, how automation and robotics are expanding, and what leaders actually need to do to respond responsibly.

So I taught a joint session with Optimizely on how AI is changing the way customers discover and evaluate digital experiences, and if your content isn’t structured for AI-mediated discovery and not just human browsing, you’re really gonna feel that impact.

We’ve been fortunate to do this work at scale at Sagepath Reply. We were recently named Optimizely’s North American Solution Partner of the Year, and we also recently won Kentico’s Site of the Year for one of our clients, Oppenheimer. And while I’m very proud of those awards and the work my team put into them, the real story is that executives are not ignoring these changes in technology and customer behavior. They’re trying to understand them clearly so they can adapt their organizations without breaking them.

So over the next few videos, I’ll break down some of the themes we covered, what’s changing in customer behavior, team structure, and digital architecture, because the conversation that’s happening in those rooms is very different than what we see in the headlines.

Lightly edited for clarity.

Search is no longer just ranking links, it’s composing answers. Google and Bing are increasingly synthesizing responses directly into the search result.

And the real question isn’t do we rank. It’s are we included in that composed answer.

Because if AI cites you in the answer, you’re visible, even if you’re not in position one on that list of links. If you’re not included, you don’t exist in that interaction.

So optimization is no longer just about ranking, it’s about being retrievable and citeable. And that changes content strategy entirely.

Lightly edited for clarity.

AI now sits between brands and customers. Research is becoming conversational, and evaluation is becoming synthesized.

And that means your content might influence a buying decision without ever generating a website visit, and traditional attribution models aren’t built for that.

If AI summarizes your content in an answer, you shape the decision even if no one clicked. And that creates an invisible influence layer, and most companies are not measuring it.

Lightly edited for clarity.

For a long time broad positioning worked. If you have a product, you could say things like, “We have enterprise-grade solutions for modern businesses.” And that was enough to get into the conversation.

But that’s starting to break because the way people research is changing. They’re not typing two-word queries anymore. They’re prompting with constraints. They’re saying things instead like, “I need a cloud ERP for a midsize healthcare system that’s under HIPAA constraints and has to integrate with legacy finance tools.”

And so when AI mediates discovery, it tries to answer the exact question being asked. If your content is vague, it doesn’t get retrieved.

So the change isn’t just that AI is replacing search. It’s that specificity now determines visibility. And most companies are still writing their content like it’s 2015.

Lightly edited for clarity.

When people talk about AI search, they often focus on the language model, but that’s not really where visibility is decided.

Modern AI systems retrieve content first. They pull relevant passages into the context window and then generate an answer grounded in those sources, which means the real competition happens before the answer is written.

If your content is not retrievable, technically accessible, clearly structured, semantically precise, it never enters the conversation.

So traditional seo still matters, indexing still matters, but there’s now an additional layer. You’re not just optimizing to rank, you’re optimizing to be retrieved and trusted and cited.

And that’s not just an seo tweak, that’s a content architecture shift.

Lightly edited for clarity.

It’s pretty well established that personalization drives revenue, and increases it.

But there’s a tension that leaders wrestle with, is that the more precisely you tailor experiences, the more customers wonder how you know what you know.

And AI increases the ability to infer intent, but it also increases that opacity.

Customers are expecting relevance, but they also expect ethical boundaries. And if personalization feels invasive, performance drops. If governance slows innovation, then growth stalls.

So personalization is not just a marketing lever, it is a governance problem. And the companies that figure that balance out early are going to outperform the companies who don’t.

Lightly edited for clarity.

AI is making everyone louder, not just faster. Writing emails, doing research, writing code, those things all took time and effort. And the production of that content is what’s been the bottleneck in knowledge work for a very long time.

But now that first draft is basically free. So all the ideas that used to get filtered out by effort don’t get filtered out anymore. Every team can generate five proposals instead of one, and every person can produce a strategy in an afternoon.

And Herbert Simon described this problem decades ago. When information becomes abundant, attention is what becomes scarce. Because more output doesn’t automatically lead to better decisions. You’re not just choosing between two options anymore. You’re choosing between 20 very convincing options.

So the companies that win in an AI world probably won’t be the ones generating the most content. They’ll be the ones that get very good at deciding what actually deserves attention.

Lightly edited for clarity.

Creating convincing information used to be expensive. If you wanted something to look legitimate, a report or research summary, an article, it took time and expertise to put it together.

But now a single person can generate hundreds of possible articles with fake analysis, synthetic images, and even very realistic voices and video. And they can test different versions of that same story until one spreads.

And we already know how information behaves in networks. Studies of social media have shown that false news travels further and faster than true news. And it’s not because of bots. It’s because people are more likely to share things that are surprising or emotionally charged.

So when you lower the cost of producing convincing content, that dynamic scales. And the problem isn’t just misinformation.

It’s that when everyone can produce convincing looking information at scale, people stop judging information based on the content alone. Instead they ask, or they should be asking where did this come from. Who produced it, and is the source reliable. Because when credibility becomes easy to imitate, trust is the thing that actually matters.

Lightly edited for clarity.

Even though AI makes it easier for people to produce good ideas, think about how LLMs work. They’re trained on massive amounts of human writing, books, articles, research papers, code.

So when a lot of people ask the same type of question, they often get variations of the same answer, and that makes the average quality of ideas go up. But the range of ideas can shrink.

And there is some early research showing this. When people use generative AI in creative tasks, their individual output often improves, but the ideas across the group become similar. Everyone performs better, but everyone starts clustering around the same patterns.

And you can already see it. If you ask 10 startups how to launch a product, or ask 10 marketing teams how to run a campaign, a lot of the answers start to sound familiar. So if everyone’s drawing from the same playbook, being different can be your competitive advantage.

Lightly edited for clarity.

How do you know something’s real? Not does it sound convincing, or does it look professional, right? A report can look legitimate, and a voice can sound authentic, and a photo can look like evidence, but none of that tells you where it came from.

Economists have studied a version of this problem for decades. George Akerlof called it the market for lemons. The idea is you’re selling a car, you generally know if it’s a good car or a lemon.

But if you’re buying a car, it’s hard to tell. So you’re not willing to pay as much, because you assume it’s probably a lemon. And that forces prices down, which prices good cars out of the market.

And that same dynamic can happen with information. When convincing content becomes cheap to produce, people stop trusting the content itself, and they start relying on signals like brand and reputation, and verification.

Who published this? Where did it originate? Has this source been reliable before? And organizations that build that real credibility end up with something that’s very difficult to fake.

Lightly edited for clarity.

Technology moves fast, and the institutions around it don’t. AI is an example of that, but it’s far from the only one.

AI can write reports, generate code, produce images, analyze data. But the systems that govern how we use technology, laws and standards, internal policies, professional norms, take years to adapt.

And these major innovations are called general purpose technologies. Things like electricity, computers, or the internet.

And the biggest changes don’t come from the technology alone. They come from everything that has to reorganize around it: companies and workflows, regulations and governance.

And that adjustment takes time. It’s already happening, but it’s not a quick process. It’s a systemic change. Governments are creating new laws, companies are building new policies, and industries are trying to figure out what responsible use means.

So the messy part we’re in right now, unfortunately, is normal. Every general purpose technology creates a period where the capability exists, but the institutions that stabilize it haven’t caught up yet.

Lightly edited for clarity.

AI makes writing, analysis, design, coding cheap. Things that used to take hours or days can now happen in minutes.

And when the cost of producing something collapses, the systems behave differently.

You see it in the noise explosion. When ideas are cheap to produce, organizations get flooded with them, and then the bottleneck becomes attention.

You see it in misinformation. When convincing content becomes cheap to generate, credibility gets harder to judge.

You see it in idea convergence. AI can raise the average quality of ideas, but it can also pull people toward the same answers.

You see it in trust systems. When content itself can’t be trusted, people rely more on signals like reputation.

You see it in institutional lag. Technology moves quickly, but the rules and systems around it take time to catch up.

And none of this means that AI is bad for organizations. It means the advantage used to come from the ability to produce, but increasingly it is coming from the ability to filter, decide, coordinate, and build trust. Because when execution becomes cheap, the systems around that execution become your advantage.

Lightly edited for clarity.

So if the hard part isn’t doing the work, then deciding what work to do becomes the hard part.

AI makes producing work faster, cheaper, easier, which moves the bottleneck. The constraint becomes about organizing, not just executing.

Because when the cost of making something drops, everyone makes more, right? There’s more ideas, more code, more reports. But now you have a flood of output, and who decides what gets shipped, who integrates it, who makes sure it works across teams?

And that’s where the bottleneck goes as coordination. That’s where the leverage point is.

For example, cloud computing made launching software cheaper, but the winners weren’t just the ones with servers. They were the ones with faster decisions and better products, and tighter feedback loops.

So if you’re thinking AI will give you an edge just by producing more, you’re wrong. The edge comes from how you organize all that output.