I’ve spent the past week in Boston with colleagues from UP THERE, EVERYWHERE, attending HubSpot’s annual conference.
As last year, the big focus was artificial intelligence - particularly large language models - LLMs, AI agents and the ways in which they are disrupting marketing, sales and search.
At times, some predictions sounded almost apocalyptic. Traditional search is declining. AI is producing more content than humans could ever consume. Websites are becoming less central to the purchasing journey. Before long, agents may even be making purchasing decisions on our behalf.
Is this LLMAGEDDON—the end of marketing as we know it?
Possibly. But that doesn’t mean the end of marketing. My overwhelming feeling after UNBOUND 26 is that we are entering an extraordinarily exciting, fresh and creative period.
Here are my five biggest takeaways.
HubSpot changed the name of its conference from INBOUND to UNBOUND.
It is only one letter, but it signals something much bigger.
“INBOUND” described a particular marketing philosophy: create useful content, attract people to your business and gradually convert them into customers.
That model isn’t dead, but the customer journey is no longer so orderly. Discovery now happens everywhere—inside ChatGPT, Claude, Google, YouTube, Reddit, LinkedIn, podcasts and online communities, as well as on company websites.
The traditional sales funnel is being pulled apart.
The move from INBOUND to UNBOUND reflects a marketing world with fewer boundaries and a far more complex playbook. There are more channels, formats, data, tools and technologies—and everything is changing faster.
Marketers can find this intimidating or liberating. I prefer the latter. We are no longer restricted to a standard campaign formula. The opportunity to identify, reach and engage highly specific audiences has never been greater.
Search is undergoing its biggest transformation since the arrival of Google.
At UNBOUND 26, we heard claims that the proportion of searches taking place through LLMs had grown from around 6% to 15% in a year. Some estimates suggested that as much as 25% of search activity now happens through an LLM.
The precise figures depend on how we define “search,” but the direction of travel is clear. People are increasingly asking ChatGPT, Claude, Gemini and other AI platforms questions they might previously have typed into Google.
The important distinction between SEO, AEO and GEO is not simply what the acronyms stand for, but how the search is performed.
With traditional SEO—search engine optimisation—someone enters a few keywords. The search engine indexes and ranks relevant webpages, and the user is presented with a list of links. The objective for a brand is to appear as high as possible in those results and persuade the user to click through.
With AEO—answer engine optimisation—the user is more likely to ask a complete question. Instead of simply offering links, the system extracts information from one or more sources and presents a direct answer through a featured snippet, voice assistant or AI-generated overview. Google says its AI Overviews combine its Gemini model with its existing search index, ranking systems and Knowledge Graph, using web results to support the generated answer.
With GEO—generative engine optimisation—the process becomes more conversational. The user may ask a complex question, add context, challenge the initial answer and progressively refine the request. The AI does not simply locate a webpage. It synthesises information from multiple sources and produces a new response, recommendation, comparison or shortlist.
The boundaries between AEO and GEO are inevitably blurred. Google now generates answers, while ChatGPT and other LLMs can search the live web. The crucial change is that users increasingly receive a synthesised answer before deciding whether they need to visit a website at all.
So where do these answers come from?
There are two different layers. LLMs draw upon patterns learned from the enormous collections of text used during their training. When live search is enabled, they can also retrieve and cite current information from the web. OpenAI, for example, says ChatGPT Search may rewrite a question into several targeted searches before selecting relevant sources.
Those sources extend far beyond corporate websites. They include news organisations, specialist and trade publications, academic research, government sites, company documentation, Wikipedia, independent reviews, blogs, LinkedIn, Reddit, YouTube and other online communities.
Research from Muck Rack analysing more than one million AI citations found that approximately a quarter came from journalistic sources, while niche industry publications were particularly influential for specialist questions. Another analysis of more than one billion citations across major AI platforms found YouTube to be the most-cited platform overall, followed by Reddit—although the mix varied substantially between platforms and types of question.
This matters because an LLM may treat a company’s website as the best source for specifications or warranty information, but turn to trade media, YouTube demonstrations, Reddit discussions and independent experts when asked whether that company or product can be trusted.
The objective is therefore no longer simply to rank highly on a search-results page. It is to become part of the information ecosystem from which the answer is constructed.
You need clear and technically accessible information on your own website—but you also need credible people elsewhere on the web talking about you.
Imagine that you want to buy a new pair of training shoes.
It is perfectly feasible to ask an AI agent for a recommendation, compare the available options, find a nearby retailer and perhaps even complete the purchase without visiting the manufacturer’s website.
Now consider buying an expensive piece of laboratory equipment.
An AI agent might conduct the preliminary research, compare suppliers and produce a shortlist. But the buyer will probably still visit the vendors’ websites before making a final decision.
They will need to check specifications, performance data, compatibility, delivery times, service arrangements, warranties, regulatory information and terms and conditions. They will also want reassurance about the supplier’s experience and reputation.
Complex B2B purchases usually involve several people, each asking different questions. A scientist, procurement manager, compliance specialist and finance director will not evaluate the same purchase in the same way.
B2B content will therefore need to serve two audiences.
The first is the AI system looking for clear facts, credible evidence, specialist expertise and well-structured information. The second is the human reader looking for detail, reassurance and the confidence to make an expensive—and potentially risky—decision.
The evidence influencing that decision will also extend far beyond the company website. Trade publications, independent experts, customer reviews, YouTube demonstrations, LinkedIn commentary and specialist online communities may all help determine how an AI system describes and evaluates a business.
For B2B companies, reputation must therefore be built across an entire digital ecosystem.
One of the most interesting presentations was based on new research from the B2B Institute at LinkedIn and Ipsos. Its central argument was simple: B2B brands do not have a reach problem—they have a credibility gap.
The average B2B purchase now involves around ten stakeholders and takes 272 days. More strikingly, 40% of B2B deals are lost not to a competitor, but to no decision at all.
Buyers are not simply comparing products. They are evaluating risk and asking: “Can I defend this decision internally?”
Trust is what enables a buying group to move forward.
The presentation introduced a useful “Credibility Stack.” The brand establishes a point of view; employees humanise it; customers and peers validate it; and credible creators, experts and industry voices scale it.
The research also challenged the idea that creator marketing is primarily for consumer brands. Creator-led B2B advertising was predicted to generate 1.2 times stronger consideration and 1.5 times greater brand impact than average short-form video.
But credibility does not come from paying someone popular to repeat a corporate message. The most effective creators are educators, specialists, executives and commentators who help people understand something new. Authentic, opinionated and human content can outperform material that feels overly corporate or rehearsed.
In an age of almost infinite AI-generated content, being seen is no longer enough. Brands must become believable.
Despite all the technology on display, presentation after presentation returned to the importance of human connection.
This is especially true when creating great content.
AI can produce competent copy in seconds. But when everyone uses the same tools in the same way, everything begins to sound the same. The words may be polished, but the thinking becomes predictable—and trust begins to erode.
I use LLMs in my work as a writer perhaps ten times an hour. They help me research, explore arguments, test headlines and proofread documents.
But they do not replace my writing skills. They augment them.
The creative edge still comes from human experience, judgement, curiosity and originality. There is already plenty of poor writing in the world. Good writers—and good writing—will become more important, not less.
The same is true of business. People still want to meet people, attend real events, exchange ideas and receive recommendations from people they trust. Relationships matter.
At UP THERE, EVERYWHERE, we have a phrase for this: Always Add Human.
That may turn out to be one of the most important principles of the AI era.
One of the highlights of UNBOUND 26 was hearing Tom Brady, arguably the greatest quarterback of all time and the winner of seven Super Bowls.
What stayed with me was not simply his success, but the practice, repetition and failure behind it. His advice for anyone wanting to succeed - ‘Get good at failing’. Treat every setback as a learning opportunity - that’s the way to success. He treated every practice with the intensity of a Super Bowl. When the big moment arrived, he was ready because he had rehearsed it so many times.
That is how businesses should approach AI.
Nobody has a finished playbook. We need to experiment, test ideas, make mistakes, learn and practise. Waiting until every question has been answered is not a strategy. By then, the organisations that started earlier will already be far ahead.
Brady also spoke about gratitude—focusing on what had gone right rather than continually dwelling on what had gone wrong.
We can obsess about what AI might take away—or concentrate on what it allows us to do better.
So, is LLMAGEDDON the end of marketing as we know it?
Yes, quite possibly.
And that might be exactly the opportunity we need.
Want to explore these changes further? Visit upthereeverywhere.com for more insights and observations from UNBOUND 26.