Horizons Q3 2025 Transcript

Go back to the Horizons Q3 2025 Summary Page

Hello, and welcome to Horizons. This is the second Horizons we’ve run in this format, which is rather nice — no expense spared once again, and we’re all very excited.

My name is David, and this is Lindsay. For those who haven’t joined a Horizons session before, it’s useful to outline what we’re trying to achieve. The whole point of Horizons is to understand what’s coming in the next 12 to 18 months that will have a fundamental impact on media, marketing, and advertising. It’s about anticipating change: shifts in audience behaviour, emerging trends, evolving channels, and the strategies of major players like Meta and Google that will shape your day-to-day work.

Another key focus is what we call “Brand You”. Horizons is designed to give you insights you can take away, not just abstract information. We also want this to be interactive: we’ll be sharing new and exciting developments that will affect you, and our aim is to make you the most interesting person in your office. To encourage that, we’ve lined up polls and a Q&A session, with questions open throughout the event and addressed at the end.

And yes, as last time, no expense spared — we’ve even been in make-up since 6am. We also have $250 of flight vouchers to give away, so you can travel… well, not quite anywhere in the world, but as far as $250 will take you. To be in with a chance, you’ll need to take part in the polls and Q&A. Those who engage most will have the greatest chance of winning.

Agenda

Here’s what we’ll be covering today.

We’ll start with WhatsApp — a really interesting space at the moment, which we hinted at during the last Horizons. We’ll then move on to Paid Search, where the concept of zero-click searches is becoming increasingly relevant as AI changes the search landscape. We’ll also touch on LinkedIn, which is evolving in ways worth discussing. And finally, we’ll look at the extraordinary developments in Artificial Intelligence — and how these are transforming digital marketing.

As you’ll see, this agenda is heavily digital. At previous Horizons we’ve also covered out-of-home and connected TV, but given the pace of change in digital right now, we’ve chosen to dedicate this session entirely to it.

Before we begin, though, we’ll do a quick recap of some of the key topics from the last Horizons.

Recap: Meta

Three months ago, we discussed Meta’s shift towards incremental attribution: focusing on incremental gain rather than last-click conversions. We highlighted that this would likely push up CPCs and CPAs, since advertisers are pursuing harder-to-reach audiences. But we also expected this to be offset by higher overall revenue.

The results have been encouraging. Across our client base, we’ve seen positive outcomes from incremental attribution, delivering anywhere between 12% and 24% additional revenue. That’s a meaningful uplift. Yes, CPCs and CPAs increased — but the incremental revenue more than compensated.

It’s remarkable how quickly this has become apparent. It was only three months ago that we were speculating about this approach, and already the results are strong. For existing clients, we’ll be discussing how to expand these tests further.

Part of this shift is driven by Meta reducing the number of niche audiences available. Many very specific audience groups are being deleted, with those capabilities rolled into broader audience segments. While this may initially feel restrictive, those hyper-specific audiences often overlapped heavily and weren’t particularly accurate. By leaning into broader targeting and focusing on creative as the differentiator, incremental campaigns become much more effective. They reach new people — not just the audiences already warm to you.

There’s also a structural shift taking place. Meta has introduced limits on the number of audiences you can run at once, which forces a rethink in how campaigns are structured. You may no longer be able to hand-pick multiple niche audiences, but you will still have some flexibility — and creative variety will become more important than ever.

That doesn’t necessarily mean producing endless new videos; it can be as simple as using different text overlays, alternative intros or outros, or variations in copy and framing. The point is to ensure variety, because Meta’s own research shows that once someone sees the same ad variant more than four times, efficiency drops dramatically.

This puts pressure on the creative process. Creative teams need to be able to generate assets quickly, respond to data, and test variations — otherwise you risk hitting frequency caps or simply handing control over to Meta’s AI. And at present, Meta’s automated creative variations are not good enough. The real opportunity lies in creative teams applying storytelling skills in multiple ways, cutting and presenting content differently to maintain freshness and engagement.

In short: Meta is evolving, and while targeting options are narrowing, creativity has become the critical lever for success.

Recap: Performance Max (PMax)

Performance Max has also evolved rapidly in recent months. Since last October, Google has ramped up development and transparency, and in the past three months we’ve seen significant changes.

One of the most notable improvements is the level of visibility now available. Advertisers can see performance down to keyword level and even which videos users have watched on YouTube before converting. This clarity builds trust in the platform, showing the value of different asset types and how they move people along the funnel.

Performance itself has been very strong. That said, there are challenges: the sheer strength of PMax results can sometimes feel “too good to be true”, raising questions about the quality of conversions. However, the granular keyword data reassures us that the traffic being driven is genuinely high quality. Long-tail, in-market brand terms are being picked up effectively, often providing better insights than traditional search query reports.

This visibility is shifting our trust levels. Historically, we might have been cautious about putting too much spend into PMax, but the improved data is making a stronger case for investment. In fact, ROI on PMax is often higher than on Meta, even though we’re still typically spending more heavily on Meta.

Looking ahead, Search and PMax are converging. Google is increasingly recommending that assets include images, and the boundary between Search campaigns and PMax will blur further. This raises important questions about how to structure Google activity: where do you want control over keywords, and where should you allow automation to broaden reach? The reality is that exact match no longer guarantees true exactness, and broad match is largely being absorbed into PMax.

This means careful planning is required to avoid overlap, duplication, or wasted budget. Different products may suit different approaches — but structure and clarity will be key.

Now let’s turn to WhatsApp — one of the most exciting emerging channels we flagged last time.

My own son first brought it to my attention, as he was using WhatsApp channels to follow Formula One teams. He’s not allowed on traditional social media yet, but WhatsApp gave him a way to connect. And over the past three months, this has accelerated. Communities and channels within WhatsApp have grown, with Meta pushing more updates and turning it into an increasingly active environment.

For clients, of course, this introduces another channel to manage, raising questions about moderation, content creation, and prioritisation. Do you divert effort from TikTok to WhatsApp? These are the conversations we’ll need to have.

Broadly speaking, there are two main opportunities on WhatsApp:

Commerce. For example, Tata Click, an Indian lifestyle fashion retailer, uses WhatsApp broadcasts for major sales events, generating half a million dollars in a single day. Engagement rates of around 57–58% are extraordinary — levels usually only dreamed of in email CRM. Other brands such as Skullcandy have used WhatsApp to address cart abandonment, reducing drop-offs by 50–60%.
Even transport is being transformed: Delhi’s chaotic bus network now allows users to message their destination via WhatsApp, pay directly, and receive a QR code ticket. Not quite as slick as tapping a phone in London, but revolutionary nonetheless.
Community. Brands are building highly engaged groups. For example, a UK menswear company created a WhatsApp community for its super-users, which not only drove sales but also helped them raise £1 million through crowdfunding. Pure Gym has created smaller communities focused on specific classes or activities, while Jellycat has used WhatsApp to complement its TikTok presence.
What’s fascinating here is the difference in approach. Jellycat’s TikTok channel, with 2 million followers, features highly polished content. On WhatsApp, with 500,000 followers, content is far simpler — often just photos with polls or emoji reactions. Yet engagement levels are similar, and the instant feedback provides powerful insight into which products will gain traction.

WhatsApp groups also feel more personal and authentic, offering the kind of insider knowledge and open discussion once associated with Twitter. Many industry groups are now sharing knowledge privately on WhatsApp, creating a richer, safer space for conversation. We’ve even won new business through recommendations in WhatsApp groups.

It’s early days, but the potential is clear.

Google and Zero-Click Searches

Finally, let’s revisit Google and the rise of zero-click searches.

This is where someone types a query, receives an AI-generated response (for example, via Gemini), and finds their answer without clicking through to a site. The implications are significant.

For one, fewer clicks mean ad inventory shrinks, driving up CPCs. Secondly, organic traffic drops, as fewer people click through to websites at all. This creates an inflationary cycle, as brands spend more on paid activity to make up for declining organic reach.

The numbers are striking: 60–80% of informational searches are now zero clicks. That means as many as eight out of ten people searching for information don’t click on anything — neither ads nor organic listings.

This is not just a Google phenomenon. Other AI platforms, from ChatGPT to Perplexity, are also shaping user behaviour. Different platforms are carving out niches: Claude for high-quality content, Perplexity for research, and so on. And all of them contribute to the zero-click environment.

For brands, the challenge is to adapt. The solution lies in:

AI Engine Optimisation (AIEO): ensuring your brand appears in AI-generated answers, difficult though this is to measure today.
Paid Search focus: concentrating spend on transactional queries where user intent is high and conversion rates are stronger.

While CPCs may rise, the quality of clicks that remain is likely to improve. These users will be more qualified, having already consumed information through AI before clicking through.

It’s a shift, but not unlike previous disruptions — such as Google overtaking early search engines. Brands that move quickly to adapt will benefit. And while quality of AI-generated information is still variable, the direction of travel is clear.

Reddit is also worth noting. Historically a difficult platform for advertising, it is improving — albeit slowly — as human moderation increases. We’ve been experimenting with it for over a decade, and while it’s still inconsistent, the potential is growing as user behaviour changes.

And that brings us back to the present: Meta and PMax continuing to evolve, WhatsApp emerging as a new force, and Google reshaping search through AI. It’s a fast-moving landscape, but one full of opportunity.
Things are speeding up, and there’s much more to come. They’re also receiving injections of cash from the likes of OpenAI, because they’re providing the learning data and infrastructure behind these systems. So, they’re being pumped full of money, which gives them more resources to make all of this work.

And now ChatGPT is launching ads. That’s currently being tested, something we also predicted. Diversification is coming, and that’s a good thing — it gives us a healthier mix across the market.

Speaking of evolution, let’s move on to the world of LinkedIn.

LinkedIn is a fascinating one at the moment. To give some context: around last year, LinkedIn launched a product called Wire, which has since been rebranded as Brand Link. Essentially, this is LinkedIn’s big move into video. They’ve been signalling this for over a year — video, video, video has been the message.

What it allows is the placement of pre-roll ads in front of content from top publishers and thought leaders — the very top echelon of LinkedIn. For example, you might see pre-rolls before Bloomberg content or before videos from influential voices. LinkedIn even demonstrates this by showing you how to skip an ad (although most people know how to do that already!).

Initially, it was just the top four or five players. But this is now expanding: influencers themselves can take part, enabling pre-rolls to run before their content, with a revenue-share model so that influencers also get a cut.

In other words, LinkedIn is adopting a YouTube-style model. And this makes sense. YouTube has restricted advertiser access in some ways, and LinkedIn is now stepping into that space, providing highly targeted environments and audiences with stronger affinity.

The benefit is obvious: rather than your ad simply being part of a scrolling feed, where it may be skimmed past, pre-roll placements in front of quality content have a far better chance of holding attention.

And this isn’t just a B2B channel. LinkedIn users are people doing their jobs — often in higher income brackets — so even if you’re selling consumer products such as theatre tickets, it’s a valuable environment. One point worth stressing: avoid defaulting to the “LinkedIn blue” look. Because so many B2B brands use blue in their ads, simply steering clear of it will help you stand out.

When it comes to creating ads, there’s no difference between LinkedIn and other platforms: authenticity, good creative, and correct specs matter. LinkedIn recommends vertical 9:16 video and suggests keeping ads under two minutes. They call that “short” — which feels long, but in this environment, it works.

The creative approach is also important. For example, if your ad is running before a thought leader’s video — which might be natural, unpolished, face-to-camera content — do you mirror that style, or stand out with something more obviously polished and ad-like? We’ve seen both approaches work: organic-style content often achieves strong engagement, while polished creative can drive stronger conversion. The key will be testing which format works best in this new environment.

Ultimately, it comes down to data, testing, and creativity. Feeding insights back into creative teams so they can adapt quickly will be crucial.

Artificial Intelligence

Now, inevitably, we need to talk about AI. I don’t think we’ll ever run a Horizons without it.

Back in February, we discussed the rise of AI influencers and how Meta, TikTok, and Snapchat are building studios to help brands and individuals create their own. The pace of change since then has been phenomenal.

AI Models and Digital Twins

Take the example of H&M’s digital twins. The models you see in some of their campaigns are AI-generated versions of real people. These digital twins are owned by H&M, but the real models also retain rights and can license their AI likenesses elsewhere. They get paid commission for the use of their digital versions.

One model summed it up perfectly: “She’s like me, except without the jet lag.”

This raises huge questions around ethics and rights. We’ve already seen controversy, such as AI influencers created to represent African communities but built by white developers in the US. The debate around diversity, authenticity, and exploitation is only beginning.

But the technology is becoming more accessible. Soon, it won’t just be big brands creating AI models — individuals will be able to create digital versions of themselves too. This could actually offer benefits, particularly for privacy. For example, younger generations may prefer projecting AI avatars that reflect their personality and opinions without showing their real faces.

There are creative and commercial implications too. Will AI commoditise content creation, creating “mass market” versions of products while artisan creators charge a premium for handmade or truly original work? It could lead to a split economy, where AI dominates mass production but leaves space for niche, high-value creativity.

Fully AI-Generated Influencers

Beyond clones of real people, we’re also seeing entirely AI-generated influencers. These avatars can livestream, sell products, and interact with audiences 24/7.

In Asia, brands like Brother Printers have pioneered this, with AI bots generating continuous sales. Company teams wake up to see how much money their AI “staff” have made overnight.

The results? Humans still outperform AI in terms of revenue, often generating three to five times more. But humans need rest and wages, while AI works non-stop once created. That efficiency makes it impossible to ignore.

Of course, much of what AI influencers say is poor-quality “tripe”. But the technology is improving rapidly, and the hyper-realism will only increase. Already, AI avatars can comment, welcome viewers, and respond to chat messages in real-time. Apart from the occasional glitch, they’re eerily convincing.

This creates both opportunity and risk. Brands must find ethical, transparent ways to use this technology — enhancing customer service, accessibility, and user experience — without misleading audiences or devaluing human creativity. Used responsibly, it can improve inclusivity (e.g. multilingual chatbots) and deliver better customer journeys. Used irresponsibly, it risks damaging trust.

AI in Manufacturing

AI isn’t just changing marketing — it’s reshaping manufacturing.

Alibaba, for example, has flipped the traditional e-commerce model. Traditionally: design → manufacture → market → (hopefully) sell.

Now, AI is used to generate designs, imagery, videos, and marketing materials before anything is produced. These prototypes are then listed online, and only if enough customers commit to buying will the product actually be manufactured.

This reduces waste but also raises questions: is creativity being sidelined in favour of whatever the algorithm predicts will sell? If people only choose from AI-generated options, are we losing originality? Creativity, after all, rarely comes from asking people what they want.

We may see a future where AI dominates mainstream production, while true innovation comes from artisan or contrarian creators operating outside of the algorithmic mainstream.

Q&A

As we move to Q&A, a few key themes emerged from the polls.

Trust and real conversions remain the biggest concerns, especially with PMax and AI-driven changes. PMax, for clarity, is Google’s system that spreads your creative assets across its entire ecosystem — Search, YouTube, Display, and more — using AI. Crucially, you now have much more visibility into where your ads appear, sometimes even more granular than search query reports. That transparency makes it far more usable.
Zero-click searches are another major concern. With traffic dropping as AI answers replace clicks, the key is to understand your brand’s visibility in AI-generated results (AIEO). Benchmarking this is the first step. Then, you need to measure the quality of the traffic that does reach you, which should improve as clicks become more qualified. Last-click attribution models will not cope well with this shift — brands must take a broader view.
WhatsApp advertising and spam. The question was: how do you avoid ads feeling spammy in a personal messaging environment? The answer: don’t spam. It sounds simple, but it comes down to relevance, value, and restraint. Cart abandonment reminders feel less intrusive, as they respond to a user’s own action. Constant generic promotions will be blocked. This is where first-mover advantage matters: build value and trust now, before the channel becomes saturated with noise.
WhatsApp for B2B. There is huge opportunity here in private communities. Instead of gating content behind forms, brands can invite target prospects into valuable, safe spaces for knowledge-sharing and networking. These communities can fuel brand affinity and create direct engagement far more effectively than traditional tactics.

Closing

Finally, looking ahead: our next event will focus entirely on AI. We’re planning a more panel-style session, bringing together voices from the brand side and the AI world to tackle questions of ethics, branding, and practical use cases. We’ll confirm the date soon (likely mid-October), and would love to hear from you if you’d like to attend.

The pace of change is extraordinary. In just the past six months, we’ve gone from speculative conversations to running AI agents that scan huge amounts of data overnight, producing reports at 3am so humans don’t have to. The curve is only steepening.

Events like this, and the conversations they spark, are essential for us all to keep evolving together. Thank you for joining us — and do stay for questions.