Paul Hewett.
The AdNews end-of-year Perspectives, looking back at 2025 and forward to next year.
And see the AdNews 50 plus page state of the market report, Forecast 2026
Paul Hewett, CEO, In Marketing We Trust.
In 2025, AI-driven search became one of the biggest shifts defining marketing. And I've spent enough years in measurement to know when the methodology doesn't add up.
The AI visibility tools that flooded the market this year looked sophisticated. The dashboards were slick and the metrics were precise - but precision isn't accuracy, and I've come to question whether these tools measure what marketers actually need to know.
ChatGPT, Perplexity, Gemini, Claude, DeepSeek - they've changed how people find information with millions getting answers without clicking through to websites. The pitch from vendors made sense: if you can't see what's happening in those conversations, you're flying blind. So here's a dashboard. Here's your share of model. Here's how you stack up against competitors.
Budgets followed. I remained sceptical.
We know AI-driven search is going to continue to be an area of focus for marketers moving into 2025. So a key focus this year has been to audit these tools properly.
The problem runs deeper than bad tools
I set up two platforms with identical parameters. Same brand, same prompts. One platform said our client was the most visible brand in the category. The other ranked them second-to-last.
Same inputs. Opposite answers.
That was the moment I stopped hedging - these tools are measuring something, it's just not anything useful.
The key thing to consider is that most of these platforms query LLMs through APIs rather than the consumer interface. The API gives you responses based on training data. The actual product pulls from real-time sources, remembers conversation history and personalises based on context. They're not measuring the same thing.
That’s why, even though we ran identical prompts through both, the results were wildly different.
But even if someone builds a better tool, there's a more fundamental issue. We've been trying to measure AI search like it's a static channel. Except AI search isn't static - it's a conversation.
Every answer depends on what came before, how the question was phrased, what the model remembers about that user.
Trying to sample this with periodic API calls is like listening to 30 seconds of random customer phone calls and claiming you understand what your customers want. You might pick up a few words but you’re likely missing the point.
This wasn't just a search problem in 2025. This year, we saw a proliferation of AI tools, new dashboards and CMOs under pressure reaching for anything that promised clarity, even when nobody had checked whether the numbers meant anything.
This all laddered up to it feeling like a year of buying confidence we hadn't earned.
GEO is not SEO
I've lost count of how many times I heard this year that Generative Engine Optimisation is "basically just SEO." It isn't.
SEO works outward from your website. You control your content, structure, backlinks - shaping signals that search engines can observe.
GEO works inward from citations. LLMs can pull from everything including Reddit threads, review sites, forums and news articles. These are sources you don't control. One outdated review or forum post can shape how AI talks about your brand, no matter how good your website is.
For any marketer looking to build their skills in this space in 2026, remember - SEO is about authority you build but GEO is about reputation you earn. And sometimes inherit, whether you like it or not.
What actually matters: accuracy
Thinking back on a year of incredible projects, one comes to mind - particularly for what it taught me.
We worked on a major rebrand for an Australian company this year, giving it a new name, positioning and completely new platform. Months later, the LLMs still do better with the old name.
Ask ChatGPT directly about the new brand and you get vague, incomplete answers. But ask a general question about the industry, and the old name comes up naturally, positioned as a leader.
It's like testing human memory - unprompted recall versus prompted. The old name has years of mentions baked into training data. The new name is still building its reputation from scratch.
What I’ve learned is that this isn’t a visibility problem, it's an accuracy problem. And it applies to any brand that's rebranded, changed products, updated contact details, or shifted positioning. LLMs surface old information because that's what their sources contain.
The sources are identifiable, encompassing everything from structured data, Wikipedia, review platforms, to industry publications. The good thing is, you can audit them and can fix what's wrong.
It’s tedious work but it moves the needle in ways that chasing dashboard metrics doesn't.
The Invisible Middle
But even if you get accuracy right, there's a harder problem - part of the customer journey is now invisible.
I've started calling it the ‘Invisible Middle’. That's where intent forms, where people weigh options, where decisions actually happen. It used to show up in search queries, website visit and content engagement. Now it happens inside AI conversations nobody can see.
Think about when someone asks ChatGPT for a recommendation. They go back and forth, compare, decide. Your analytics show nothing.
This isn't a gap we'll close with better tracking. It's a structural shift in what marketers can know. Next year, we're going to have to get comfortable with inference. Mix modelling, incrementally testing, econometrics. Approaches that feel like a step backwards to people raised on click-level attribution but are actually the way forward when you can't observe everything.
Where that leaves us
LLMs are a channel now. That’s not a 2026 prediction, it’s a 2025 reality.
Marketers need to get to grips with it and there’s genuine urgency to do so. We spent 2025 chasing visibility metrics from tools that can’t reliably measure what they claim to measure.
In 2026, the focus needs to shift. We need to focus on accuracy over visibility, to fix what LLMs get wrong about your brand rather than obsessing over share of model and accept that part of the customer journey now happens somewhere we can’t see.
That’s not a reason to wait. It’s a reason to take this more seriously than we did in 2025.
Have something to say on this? Share your views in the comments section below. Or if you have a news story or tip-off, drop us a line at adnews@yaffa.com.au
Sign up to the AdNews newsletter, like us on Facebook or follow us on Twitter for breaking stories and campaigns throughout the day.