The machines decide what gets seen, but brands still have to be worth choosing 

Why agencies don’t have to pick between AI and brand building, and why the old rules matter more now that a machine is reading the file. 

A man with dreadlocks sitting at a desk in front of a computer, intently focused on the screen.

The machines decide what gets seen, but brands still have to be worth choosing    

Why agencies don’t have to pick between AI and brand building, and why the old rules matter more now that a machine is reading the file. 

A man with dreadlocks sitting at a desk in front of a computer, intently focused on the screen.

By Conor Nichols, Special Reporter

"If you aren’t in the AI summary, you’re not in the decision."
"The brands that will succeed are those that are actively sought out."

We’ve all heard these arguments, but usually as single areas of focus.

This is reflected in the industry with agencies either doubling down on brand but largely through the lens of ‘linear storytelling and traditional advertising formats’ or those going in all in on an AI powered future ‘obsessed with performance creativity and machine readability. 

Jessica Tamsedge, CEO of Dentsu Creative UK, thinks both camps are missing the point. “We recognised there was a sweet spot that was needed more than ever,” she says. “A point of view on brand building that makes sense of that machine-readable landscape, but also builds active preference, choice and emotional relationships with consumers.” 

This is the thinking behind Dentsu Creative’s new report, ‘Word of Mouth, Word of Machine’, a playbook for brands that want to be seen, understood and surfaced by LLMs without losing their humanity. It’s aimed at brands, but it also sets out what agencies can do for them. Its closing line sums up its core argument: “The machines will decide what gets seen. The humans will decide what gets believed.” 

The false dichotomy

Jessica Tamsedge

Jessica Tamsedge
CEO
Dentsu Creative UK

The report is built on the idea that the machine-versus-human debate is a false one. “Machine readable versus human referenceable: these things are the same thing,” says Tamsedge. “There’s a symbiotic relationship. The training data starts with people.” 

The behaviour change that underpins this false dichotomy is already here. According to the report, 79% of Brits have acted on AI advice in the real world, 58% of searches now end in zero clicks, and 41% of consumers already use AI for shopping-related tasks. Brands that don’t feature in the AI summary, the report warns, increasingly “won’t feature in consumers’ purchase decisions”. 

A black background with two pie charts, one in pink and one in purple, with percentages and text indicating that 58% of searches now end in zero clicks, because people find all the information they need in the summary.

This is where the report’s five principles come in: clarity, machine readability, consistency across time, earned “human referenceability” and human authority. The reaction from clients has been telling, Tamsedge notes. “We’re seeing far more CEOs, not just CMOs, coming to the table for these conversations to ask, what do I actually need to stand for and put out into the world?”  

Flooding the zone 

A lot of global CMOs have told her they want to “flood the zone.” This means producing more, being everywhere and essentially feeding the machine. She thinks this is the wrong instinct.

"Actually, restraint and focus are needed, which, as a media as well as creative agency, we’re well placed to advise on."

Part of this restraint is knowing where the machines actually look. “People have an idealised notion of the types of communities we now need to flood with content,” she says. “Reddit, for example, is incredibly important, but it’s not the only platform.” Tamsedge’s advice is to be “forensic” in working out the breeding ground for the models and then working out how to get into it. 

The report adds a data point. It argues LLMs reward brands that perform strongly in AI recommendations tend to have “regular, recent and repeated mentions” across publishers, forums and expert commentary, and it cites Dentsu’s own analysis finding that 39% of influential sources referenced by LLMs had been recently published or updated. This suggests a one-off viral hit wouldn’t be enough for LLMs. 

The playbook’s two axes are visibility and credibility. This means making sure the right voices are showing up, and Tamsedge’s point is that they aren’t necessarily the loudest ones.

"It’s not just the social-media-friendly voices that are provocative, surprising and feed stopping. You now need trusted voices that the LLMs will determine are authoritative."
A circular diagram with the words "brand clarity" in the center, surrounded by five principles for building brands for an AI future. The diagram is divided into five sections, each representing a different principle of brand clarity. The principles are written in a clear and concise manner, making it easy to understand and follow.

The report argues that LLMs nudge towards the centre so influence will be built through reliability as well as volume or virality. Add brand consistency and assets that build “a structured representation of your brand”, and, Tamsedge says, “all of that is science you can manage.” The report’s aim is to codify this for clients, “because it’s a minefield right now.” 

On Reddit, the report quotes Sharb Farjami, the platform’s VP of Global Agency and Partnerships, who argues that people still look for human validation when a purchase matters. Reddit’s own survey found its users trust it more than LLMs for their final purchase decision: twice as much in the US and nearly four times as much in the UK. 

Tamsedge touches on more examples of how brands can show up in communities like this. Brand employees or ambassadors can run “Ask Me Anything” sessions or “Lifting The Lid” content, which lets the brand play “a different kind of archetype”. The shift she describes is away from the hero or the saver, and towards balancing “brand meaning with utility”. 

It also means a bigger role for PR, a fundamental for brands pre and post the dawn of gen AI. “PR is again a 101 practice” she says. “Make sure you are talked about. Make sure your message punches above its weight and travels further than what you can pay for, because the paid search ranking visibility game is sort of over as we know it.” "

The archivist problem  

There’s also the problem of memory. Humans forget old campaigns but LLMs don’t. “They are collecting decades of data, as long as the internet’s been around and brands have been advertising on it or even had websites on it,” Tamsedge says. “That is building a picture on brands’ behalf of what they’ve actually been putting out to the world and what other people have been saying about them.” 

The report gives an example from Dentsu's own media team. A colleague “recently helped a supermarket brand haunted by past campaigns”. The retailer’s current message was about convenience and quality, but AI summaries kept undercutting it with an aggressive low-cost push from ten years earlier, one the report says is “no longer on brand, or justifiable”. 

That’s the report’s point about how machines handle change. LLMs “are far less adept at understanding nuance, subtext, or strategic reinvention”, it argues, so inconsistencies are “less likely to be interpreted as evolution” than “averaged”, or presented as contradictory perspectives. Old material isn’t necessarily demoted just because it’s old. 

Tamsedge says she’s seen the same pattern more broadly. “We’ve worked with retailers that have moved from value, cost and discount messaging to wanting to stand for more premium, choice and elevated distinction,” she says. The trouble is that LLMs “don’t forget”. 

This makes bold relaunches harder. “If that picture centres around a very specific brand positioning, you will be harder pushed nowadays to launch a new brand direction, purpose and mission and expect a clear, cohesive picture for your audiences.” Her framing for CMOs is that they now need to be “archivists as well as visionaries”. 

Tamsedge stresses how vital this is.

“We used to talk about a lot of this as romantic marketeer land: go into your DNA, look at where you came from. These aren’t nice-to-haves anymore. They are vital to making sure you are consistently structured, so that an LLM knows what to think of you and represents you coherently.” 

Found versus chosen 

For Tamsedge, getting a brand surfaced is table stakes. “This will only get you into the AI Overviews,” she says. “It will only get you into the response.” 

 Tamsedge suggests the shopper who just wants the cheapest, closest, most reasonable option and will care less about who the most talked-about brand is. The report goes further and imagines a “rational consumer” era, where emotion and mental availability, “all the old idols of marketing science”, are under threat when machines do the buying. Tamsedge’s answer is the opposite of giving up on brand:

"You have to stand for something such that people will still put the brand name in the search bar."

That means creating things only the brand can offer. “You’re never going to win on ‘best trainers for back to school’ if you’re Nike, for example, on brand alone,” she says. “You’ve got to double down on Nike Run Club, for instance.” Her own example is Waitrose, where Dentsu has built branded properties such as a live cookery school. “If people are excited enough by it and care enough about it, they’re not searching for the best entertainment thing in this category. They’re saying, ‘I want to see more about Waitrose, Waitrose’s cookery school.’ Immediately you’re cutting through the dynamics and the filters the LLMs put on, because the consumers want more and are choosing you.” 

This is why Dentsu is working a lot within sport, entertainment and culture. As the international marketing agency of record for the NFL, it helps the league with its own brand and with picking brand partners, because partner brands “want to be associated with cultural experiences that are not replicable by a category search”. 

The report is upfront that this is old advice. “If that feels like brand building 101, it’s because it is,” it says of its first principle, clarity. The stakes are higher. The report’s pitch is that if your brand story is messy, AI will scale the mess. 

Agents as just another channel 

As for the next step that most of the industry is anticipating, AI agents acting on behalf of consumers, Tamsedge is relatively calm. “I think brands will treat this like another branded channel,” she says. It needs to be disclosed as a utility, as a technology touchpoint, but “you’ll want it to convey the same branded qualities as any other brand experience. If you’re a brand that’s about creativity, compassion or responsiveness, you’ll want all of those things to be present.” 

Tamsedge points to Dentsu’s work with Adobe as an example of how this is being put into practice. Dentsu is Adobe’s primary international creative partner and co-created Gen Studio Dentsu Plus, a content supply chain technology. It is also working with Adobe on what she calls “brand intelligence”: a way of building a brand’s rules into its content and AI-driven experiences, so that everything produced automatically draws on “the non-negotiable aspects of the brand”. 

The idea is consistency at scale. “Everything has to feel like this. Everything has to meet these criteria,” she says. “Otherwise, it’s not going to feel like it’s from brand X.” 

A close up of a person's face with a purple and blue background. The person's eyes are visible, and the image has been edited to give it a unique look.

What it means for agencies 

For agencies, Tamsedge’s view is that you don’t need a completely new breed of person. “The biggest thing for me is in strategy, and it’s not a new breed; it’s a turbocharged version of our existing strategists.

Her test for any brand platform, or “connecting idea”, is whether it works for machines, humans and the purpose the brand is trying to live up to, or whether it’s an abstraction. Her example of the latter is: “‘We exist to create lovely moments of joy and happiness’ means nothing.” She says the brand strategist has to be far more fluent in the AI and agentic audience “to know what a connecting idea should look like in reality, and how much harder it has to fight to get to work that will be talked about, remembered and actively sought out in a sea of sameness and AI-mediated responses.” 

She’s also wary of the instinct to create an AI department. “There’s a risk if you create an AI strategy specialist, or an AI brand codifier. If you create too many separate divisions, you run the same risk we did in the early 2000s: here’s the digital department, here are the people who think about it from a channel perspective. Everything still comes back to an audience experience and a brand intention.” 

For agencies without a Deibu 

Not every agency has the resources of a network like Dentsu. The group has its own proprietary tool, Deibu, which maps a brand’s LLM visibility, performance, trusted voices and consistency over time and produces an audit diagnosis. But Tamsedge insists the starting point can also be much simpler

“In crudest terms, anyone can search their brand, or different category-specific searches, on ChatGPT, Gemini or Claude to see how they’re performing.” Results are personalised, so they’ll vary, but “it gives you a sense of how you’re coming across through an agent. That’s a really helpful, easy, low-barrier starting point.” 

The second step is an audit of your own back catalogue: paid, owned and earned. “What’s sitting on your websites, what’s not been updated? AI is not discriminating when it’s trawling everything... just because it was put out a long time ago doesn’t mean it’s demoted in the ranking.” Back to the archivist mindset, she says: “What have we been inadvertently putting out to the world?” 

The last word goes to people 

So, what should brands do now to be on the right side of the machines-decide-what-gets-seen, humans-decide-what-gets-believed divide? Tamsedge’s answer is a mix of the technical and the old-fashioned. Understand, and be “not mystified by”, why you do or don’t show up in an AI answer. Then be “so singularly clear” on your purpose that it pushes you to create experiences people will seek out. 

“Then nothing beats human choice,” she says. “You cut through all of the algorithmic intervention and mediating, and you as a customer go, ‘I want to spend time with this brand.’” 

A poster with the text "This is not a two-audience problem, the relationship is symbiotic" written in bold, black font against a white background. The text is centered in the middle of the poster, emphasizing the importance of the message.

It’s a useful reminder for an industry that’s spent three years worrying about being automated. It’s also a reminder that the brands benefiting most from the machines are the ones that did the hard thinking before they arrived. The report’s own version is that brands will “scale through AI, without surrendering their souls to it”. In other words: the work the machines can’t do is still the work. 

 

A man with dreadlocks sitting at a desk in front of a computer, intently focused on the screen.

Published by the IPA, the IPA AI Magazine helps member agencies make sense of the fast-changing AI landscape, with practical insights, expert perspectives and real-world applications of generative AI.
Managing Editor: Conor Nichols
Design: Jenna Betts

© 2026 The Institute of Practitioners in Advertising. All rights reserved. No part of this site may be reproduced without our permission.