Digiday+ Research: The marketer’s guide to AI applications, agentic AI, AI search and GEO/AEO in 2026

Yet this rapid evolution has also brought new challenges: how to best integrate AI into internal workflows, the complexity around agentic AI and the ramifications of AI search.

In this report, Digiday explores how marketers are navigating the opportunities and challenges AI brings as it becomes an indispensable piece in their toolkits.

Digiday+ Research surveyed 142 brand and agency professionals about their use of AI and current and future investments in the technology. Digiday+ Research also conducted individual interviews with marketing and technology executives responsible for AI investments and applications development. They included executives from:

• Code and Theory
• Huge
• Left Field Labs
• M7 Innovations
• Tinuiti
• IPG Media Lab

Digiday’s survey — which has been conducted annually since 2022 — found that marketers’ adoption of AI technology has risen significantly. In 2022, 44% of brand and agency pros said their companies were investing in AI technology. That percentage rose to 57% in 2023 and 71% in 2024, before hitting 86% in 2025.

AI’s growing importance for marketers has also become evident in the number of companies that have created chief AI officer positions over the past two years. In 2024 and 2025, brands General Motors, Mastercard and ZocDoc appointed AI chiefs, as did agencies Golin, Luckie & Co. and Horizon Media.

“In every industry there’ll be a percentage of companies that figure [AI] out, then a large percentage of companies that don’t. And I think the economic upside to figuring it out makes for such a big competitive gap,” Wesley ter Haar, co-founder and recently appointed chief AI officer of digital agency Monks, told Digiday in April.

Consumer adoption of AI has seen significant growth, as well, and, as a result, many brands are now regularly using AI in consumer-facing applications. PetSmart, for example, relaunched its member program using AI to tailor deals for customers based on their past purchases, and Guitar Center launched a chatbot called Rig Advisor to help customers select the right products to suit their needs.

However, as AI technology evolves and becomes more complex, many of the marketers Digiday spoke with for this report said that training employees on how to best use AI tools lags behind overall adoption.

Dan Gardner, co-founder of creative agency Code and Theory, said that’s particularly apparent when it comes to upskilling and reskilling team members. “Anybody can learn a new tool. Upskilling and reskilling is multiplying the value of your human ingenuity,” Garder said. “For example, a designer is trained in communicating design. Using an AI tool to design a little easier is not making them more skilled. They’re just using a new tool. The way to upskill is to multiply the value by which they understand communication design. There has not been enough emphasis on the new way to work versus implementing tools.”

Matt Maher, founder of independent research and development firm M7 Innovations, said that, while individual users may be comfortable with AI tools, companies generally aren’t using the tools to their full capacity. “Users are definitely more knowledgeable at a baseline level, but there is a delta between understanding tools like ChatGPT, which has 800 million weekly active users, and Gemini, which has 400 million a month, and then using them to their utmost potential,” Maher said.

“When a company adopts [Anthropic’s] Claude and uses APIs for all of its internal software, and [Microsoft] Copilot for essentially everyone, it feels like a big machine,” Maher added. “It’s almost a failure of imagination of how much you can actually use these tools if you push them to their limits. … Big tech isn’t great at showing people the amazing things they can do. … There’s still a gap, even though, at a baseline, we’re all getting used to AI.”

When organizations — including marketing teams — make moves to adopt AI technology before they’ve fully developed a plan on how that technology should be used, a gap often results between how much is being invested in AI tools and the return on that investment, according to Marc Maleh, global CTO at design and technology agency, Huge.

“You have massive investments on AI and the basics of all of them is the infrastructure layer — TPUs [tensor processing units] from Google, GPUs [graphics processing units] at Nvidia — somebody has to pay for that,” Maleh said. “Every time an agency or brand wants to deploy a model, the Googles and the Amazons of the world need to find a way to monetize the infrastructure layer, and all of the layers within the AI ecosystem.”

Paying to use that infrastructure is becoming a greater concern for brands, Maleh explained. “What if I want to turn on 500 more seats of Claude code? What does that look like financially? Am I going to get that money back if I’m only getting a 30% productivity increase?” Maleh asked.

“There’s a realization about the economics of GPUs and TPUs because money is going into those things,” he added. “All of a sudden, those models have to get monetized in a real way. AI was the shiny object and continues to be that. Brands thought, ‘We want the press release, so let’s worry about the GPU and TPU charges and the API calls later.’”

Digiday’s survey found that the majority of marketers continue to implement AI technology into their workflows by using out-of-the-box AI tools, rather than building tools in conjunction with existing large language models such as Google’s Gemini or OpenAI’s GPT, or building and training their own LLM in-house.

Eighty-five percent of survey respondents said their company is using out-of-the-box AI tools. Less than half of respondents (40%) said their company is building proprietary tools with an existing LLM, and only 19% said they’re building and training their own LLMs.

The expense of building customized AI tools through an existing LLM or building and training a proprietary LLM, along with the learning curve associated with implementing either of these options, are the likely reasons why most marketers are choosing to use out-of-the-box AI tools. Smaller companies also may not be able to afford an AI team dedicated to creating custom tools.

Huge’s Maleh noted that several new out-of-the-box AI tools have become available to marketing teams within the past year. “What’s actually happened is there’s more available out-of-the-box models now,” Maleh said. “Whether it’s an Adobe or a Google, you can start with a base model that somebody else has already invested time in creating, and then customize it to your needs. That’s a lot of what Google’s Cloud Platform has with out-of-the-box tools like Vertex.”

Google Vertex AI is an AI development platform that uses Google Cloud’s infrastructure to let users build their own custom AI or machine learning models. The platform offers pre-built models that serve as a base for users to build custom tools and capabilities.

Maleh said another change to the AI landscape that has taken shape over the past year is a democratization of AI models and collaboration among some of the big industry players, such as Adobe and Google’s recent partnership in which Google’s Gemini, Imagen and Veo models are integrated into Adobe’s creative tools. “Now, if I’m using Adobe’s Firefly but I want to use Google’s Nano Banana as my asset generation model, I can do it within the Firefly console,” Maleh explained. “A year ago that wasn’t the case. It was Firefly or nothing. … We went from a place where a lot of platforms were walled gardens to where it’s more opened up.”

M7 Innovations’ Maher said that some tech companies are lowering the barriers around their AI services and allowing brands to build on top of existing features. “What I’m starting to see now is a tech stack,” Maher said. “There’s not gonna be one to rule them all. I’ve seen brands say, ‘Copilot is our base, and we stack Claude on top with a bunch of really smart APIs.’ Or, ‘We use Adobe Firefly to create and we have Canva to complement.’”

This can result in significant savings for brands, Maher added. “I’m starting to see cost efficiency gains — partnering with the bigger companies, but creating their own version or sandbox. … And you’re saving a lot of money because you’re not having to build it from scratch,” he said.

Digiday’s survey found that, for the second year in a row, copy generation is marketers’ most-used application of AI. Seventy-two percent of respondents selected copy generation as the top AI technology their company used in 2025, up from 66% of respondents who said the same in 2024.

The survey also found that more marketers used AI for multimedia generation in 2025. Sixty-four percent of respondents said their company used AI applications for multimedia generation, including image, video and music creation, in 2025 — a 17 percentage point increase from the 47% of respondents who said the same in 2024.

For example, global fashion retailer H&M and athletic apparel brand Puma both used AI applications to create imagery for recent ad campaigns. In March, H&M included images made with AI-generated digital recreations of a real model in social media posts and product imagery. In the same month, Puma released an ad campaign generated entirely by AI.

Despite the fact that AI applications are being used by more marketers, both H&M and Puma faced criticism for removing the human element from their creative work. For instance, influencer Morgan Riddle called H&M’s use of AI twin models “shameful” for taking away job opportunities that could have gone to real people.

In another use case, global cosmetics company L’Oreal’s CREAITECH lab leverages AI to generate localized visuals and campaign assets from simple text prompts using models including Google’s Imagen 3 and Gemini. This has helped the company scale localized messaging and creative for TikTok and Instagram across 20 markets in Europe, the Middle East and Africa.

“We can take the same product shot and seamlessly place it in a Japanese garden, on a bustling Parisian street or in any other relevant setting,” Antoine Castex, group data and AI enterprise architect at L’Oreal, said in an April interview with Google.

The move has helped reduce costs and accelerate storyboarding, concept generation and visual testing for the company, according to Thomas Alves Machado, L’Oreal’s global content director for generative AI.

“The productivity side continues to be a big one for the internal tooling or unlocking things that [companies] could never do before with existing data and humans,” Huge’s Maleh said. “The KPIs are going to become more defined too. Are you trying to get more revenue, increase margins or just [have] that flash in the pan, like a press release?”

M7 Innovations’ Maher suggested that marketers should first identify the problems they’re hoping to solve with AI and then match AI tools to a solution. “I often say we’re at the stage now where AI has a lot of solutions that are looking for problems,” Maher said. “What most brands fail to do is to look in the mirror and say, ‘What are the problems we are trying to solve?’ … The term technical debt is popping up more because a lot of [AI] platforms offer 25 great features, but a company only needs two or three.”

In Digiday’s survey this year, 18% of respondents said their companies use AI for applications in the “other” category. Among those respondents, many said they use AI tools for data reporting and coding.

Skincare brand Beekman 1802, for example, uses AI for data analytics. The brand partnered with AI analytics company Bezel to analyze customer data and create customer personas for its targeting and marketing efforts.

“We were able to throw all of our data at a large language model and really understand deeply who this consumer is and how many subtypes of consumer we have,” Beekman 1802’s Chief Digital Officer David Bake said during Glossy, Modern Retail and Digiday’s AI Marketing Strategies virtual event in October.

And at Amazon, developers began leveraging AI tools to code more securely in 2024. During the company’s Q2 2024 earnings call, Amazon CEO Andy Jassy said that AI-powered software development assistant Amazon Q Developer had helped Amazon to migrate 30,000 product applications from Java 8 or 11 to Java 17, saving over 4,500 years of development work and $260 million annually from performance improvements.

“Asking an AI assistant to write some code is not very different from asking it to translate English text into French with the correct semantics,” Jason Andersen, vp and principal analyst at global technology research and advisory firm Moor Insights and Strategy, told Digiday. “In both cases, the requestor still needs to be in the loop and provide the context of the situation to request the work.”

Adam Simon, former managing director at IPG Media Lab, said that, no matter how brands and agencies implement AI tools, their main goal should be to improve performance with human oversight. “Just because [generative] AI finds some new version of the creative, it doesn’t mean that you have to necessarily run with the AI-generated one,” Simon said. “You could take that back to your creative team and have them create their version of it. We want to keep humans in the loop for the most high-profile, consumer-facing assets. But these tools will continue to get better.”

When Digiday asked brands and agencies which specific types of AI technology they’re using in different workflows, the survey found that generative AI has higher rates of adoption than predictive AI.

Generative AI uses advanced machine learning techniques to produce new content like text, images, code and audio, while predictive AI uses data analytics to forecast upcoming events such as customer behavior, potential fraud and purchase probability. Predictive AI has existed longer than generative AI, but survey results suggest that generative AI is more applicable to marketers’ workflows.

Digging deeper, Digiday’s survey found that marketers use generative AI the most for creative production (82% of respondents said they use generative AI in this way), followed by marketing (at 81% of respondents), and external and internal communications (at 75% and 56%, respectively).

In one example of what this looks like, Unilever partnered with marketing services group Brandtech on its Beauty AI studio, an in-house AI system that creates assets for paid social, programmatic display and e-commerce. The tool has reduced creative production time, according to Selina Sykes, global vp and head of marketing transformation, beauty and wellbeing at Unilever.

“It’s a different way of working. We used to send briefs off and get content back. Now it’s this agile, iterative approach,” Sykes told Digiday. She estimated in the summer that, on average, Unilever was using the system to create 400 creative assets per product. “Before, we’d be doing 20 assets per campaign, and now we’re doing hundreds,” she said.

Digiday’s survey found that brands and agencies use predictive AI most for measurement and KPI analysis. Forty-eight percent of respondents said their company uses predictive AI for this purpose.

For example, Kroger Precision Marketing Powered by 84.51°, the combined division of Kroger that houses its data science and retail media divisions, uses predictive AI for KPI analysis. The company recently launched an AI-generated email digest for its suppliers that takes the form of customizable weekly reports that analyze both short- and long-term performance through several KPIs.

Huge’s Maleh said the company has expanded its use of AI tools to include both generative and predictive AI. “Design is using it for image generation. Within our creative department, we’ve gone from tools not just being used for pitch decks, but actually being used in production,” Maleh said. “Our data science and data engineering team built a product called Live and within Live we’re creating digital twins that learn, adapt and guide you about a brand. So, using it [AI] for rapid insight, strategy and persona development and concept testing.”

However, Maleh added that AI has yet to be used to its full potential. “As an industry, we have not fully embraced them [AI tools],” he said. “Whether it’s writing emails or similar activities, the human touch still goes a long way.”

Eric Lee, CTO at creative technology agency Left Field Labs, said that AI tools often overlap in their applications, and that, at the end of the day, brands are looking for results. “Increasingly, the lines are blurred,” Lee said. “The distinction between a predictive model versus what is run through a language model — there’s a lot of crossover. People largely don’t care what’s under the hood. They just want it to work. Whether it’s a language model or another model helping them get there, what’s the right tool for the job in the most effective way?”

Digiday’s survey found that marketers are using AI less frequently in workflows that include media buying and planning and financial analysis. About one-third of respondents said they use predictive AI for media buying and planning (35%) and financial analysis (34%). Meanwhile, only around a quarter said they use generative AI for media buying and planning (25%) and financial analysis (21%).

But some industry executives have said they think AI has potential to be used more for media buying and planning. At the Digiday Media Buying Summit in October, for instance, industry pros discussed ways in which AI could be applied for media buying and planning. “Just starting with an AI tool to analyze the [request for proposal] gets a good summary for the team,” said one agency executive who spoke with Digiday on the condition of anonymity. “In certain cases, with tight deadlines, it could be a very good first crack at what we might be able to do. And who knows, it might come back with something you didn’t even think about … I don’t think I’d rely on it to write the RFP, but it’s a great sort of first look.”

Adam Simon, former managing director at IPG Media Lab, said smaller brands and agencies might benefit from using AI tools in the media buying and planning process more so than their larger counterparts. “Meta and Google have great platforms for helping to automate A/B testing and optimize targeting, spend and creative,” Simon said. “You may not see that on a large agency or brand, but it’s transformative in small businesses and marketing teams where one person is trying to manage campaigns across multiple platforms, and they don’t have the time to commission and plan out elaborate A/B testing.”

Agentic AI has become a real buzzword in the marketing industry. But Digiday’s survey found that the technology has a ways to go to catch up to predictive and generative AI when it comes to marketer adoption. More than half of survey respondents (54%) said their companies don’t use agentic AI within their workflows.

Agentic AI and AI agents use the components of traditional and generative AI to anticipate needs autonomously and execute against them. They can learn from experiences, adjust future actions based on new information and take action on a user’s behalf, rather than relying purely on human commands like other forms of AI.

For example, digital agency Monks recently partnered with technology company Nvidia to produce an experimental 30-second film for athletic apparel brand Puma that was created entirely with agentic AI. Monk’s ter Haar described the process to Digiday. “It’s written by AI agents. It’s mood-boarded by AI agents. The director of photography is an agent, and those agents have worked together to pull that script together, pull that mood board together,” ter Haar said.

Despite such use cases, there are still a number of barriers facing agentic AI when it comes to widespread adoption among marketing teams, especially compared with predictive and generative AI. One of those barriers is trust. Agentic AI runs without human feedback, which requires that users of the technology trust that AI agents are running tasks correctly. If agentic AI hallucinates in one step of a process, subsequent steps will be affected. (AI hallucination refers to when an AI application generates false information that it interprets as factual.)

Huge’s Maleh said marketers need to carefully weigh which tasks agentic AI is capable of completing correctly on their behalf and what types of data should be shared with the technology. “There is a governance concern, especially with agentic and data access,” Maleh said. “What do I want to give an agentic AI the rights to do on my behalf? … You may want an agentic experience, but what are your beliefs in regard to the data layer and access to that data? What do you want to allow the end consumer to take control of?”

Left Field Labs’ Lee said using agentic AI for smaller tasks builds trust and comprehension over time. “Over the past year, we have seen more micromanaged agents, where you have an agent that can do a lot, but a layer of education and trust needs to be built before people are ready to say, ‘Yes, go manage my inbox, or execute this code for me,’” Lee said. “Building up that trust is a process, just like it would be with a human at the end of the day.”

Beyond building trust, a more critical barrier to furthering adoption of agentic AI among marketers is the complexity of the technology itself. Because agentic AI runs multiple tasks autonomously, rather than functioning as a single application, it often must be linked to multiple other tools like APIs and even LLMs.

“We’re also emerging from a mindset of collective extraction — let’s get as much information as we possibly can. That’s the old paradigm,” Left Field Labs CEO Sarah Mehler said. “Now it’s about what’s the right amount of information necessary to complete this for agentic AI to be valuable. Agentic AI is really hard to do, and not everybody can. There are very few examples of it actually working well in the world.”

M7 Innovations’ Maher said with the fast-evolving nature of AI applications, marketers need to prepare their in-house teams now for future iterations of AI technology. “AI has moved from being an assistant to being more of an operator. We see this with ChatGPT’s Atlas, Perplexity’s Comet and Gemini in Chrome,” Maher said. “But how does a brand prepare its website for an agent. … Front of house, it’s making sure you’re agentically ready. Back of house, it’s how you can stack things [tools] in different ways to make you more efficient as a marketer.”

The number of AI-powered search applications is growing. Google introduced its AI Overviews search function in 2024, followed by AI Mode in 2025, while OpenAI rolled out its ChatGPT chatbot in 2022, later integrating a conversational search feature. Anthropic’s AI assistant Claude and “answer engine” Perplexity also feature conversational search results.

While AI-generated search results are still relatively new in comparison to traditional search results, marketers are deeply feeling the effects. As consumers benefit from increasingly detailed search results driven by AI, brands are seeing significant decreases in site referral traffic.

Digiday’s survey found that more than one-third of brand and agency pros (37%) said their companies have seen decreases in upper-funnel search traffic as a result of AI, and 21% said their companies have seen decreases in lower-funnel search traffic as a result of AI.

This phenomenon has come to be known among marketers as zero-click search — consumers find answers to their search queries directly on the results page and don’t have to click through to additional websites for more information. Zero-click search is more common in AI-generated search results that typically provide information in conversational summaries, like Google’s AI Overview.

In a December 2024 survey of 1,100 consumers conducted by Bain and Dynata, 80% of users said they relied on AI summaries at least 40% of the time, leading to an estimated organic traffic decrease between 15% and 25%. And a 2025 eMarketer report estimated that AI search agents could cause a 38% drop in ad exposure during discovery, 47% during consideration and 30% at conversion.

Knowing how and when a brand appears in AI-generated search results is also an obstacle for marketers, according to Jen Cornwell, senior director of AI, SEO and innovation at performance marketing agency Tinuiti.

“Because we don’t get data outputs from the [AI search] platforms themselves, it’s been really difficult,” Cornwell said. “We’ve moved into zero-click, where there is minimal attribution for visibility inside of one of the AI search platforms. Yes, you can get a citation. You could get a link and look at your referral traffic, but that is what we’re assuming. It is a fraction of the visibility that some of these brands are getting.”

Some third-party companies are looking to help marketers combat this with tools that monitor and boost search performance, according to former IPG Media Lab’s Simon.

“We’ve seen a slew of new companies like Profound, Evertune and Bluefish that are great at being AI brand monitoring firms. Startups like Scrunch that help navigate the space and optimize existing content for AI-powered search and eventually for agentic commerce,” Simon said. “If you have the budget and the resources to work with a company like Scrunch, it can be helpful. Long term, we are going to see better tools and communication from the platforms, like OpenAI, Google and Anthropic, about how to engage with them.”

It’s worth noting that Digiday’s survey also found that nearly one-third of respondents (32%) said their company’s upper-funnel search traffic has remained unchanged as a result of AI, and 36% said their company’s lower-funnel search traffic has remained unchanged.

AI-generated search is still a nascent technology, but growing consumer use means that brands will likely be adjusting their business practices around it for the foreseeable future, according to Left Field Labs’ Lee.

“Shopper behaviors are changing,” Lee said. “When you talk to AI about getting a recommendation, it feels more trusted than the ads at the top of a Google search. So the nuance of how brands talk about their product, and how AI then talks about their product, is really interesting to look at. That journey for marketers is incredibly insightful.”

It’s also important to note that traditional search has not entirely fallen by the wayside. Tinuiti’s Cornwell said shoppers regularly return to standard search tools. “Consumers are still going back into a traditional search engine at some point in their search journey, maybe to extend the research that they did inside of a ChatGPT,” Cornwell said. “AI search is actually a smaller part of what brands need to worry about. Brands need to focus on the fundamentals. How can we add value? Because there is still value in organic search. It is losing value over time, and we do need to plan for zero-click. But how do we make sure that we build toward both?”

As AI-driven search evolves, GEO (generative engine optimization) and AEO (answer engine optimization) are acronyms being used interchangeably to describe how publishers, marketers, and e-commerce sites ensure that AI crawlers can easily understand enough information about a brand so it surfaces in AI-powered answer engines. Essentially, GEO and AEO are to AI-generated search what SEO is to traditional search.

When Digiday asked marketers about how their company is optimizing GEO or AEO strategies, no dominant strategy emerged from the survey results. The greatest percentage of survey respondents (34%) said their company’s GEO or AEO strategy involves highlighting more content featuring answers to potential consumer questions.

But Digiday’s survey also found that just over one-quarter of respondents (27%) said their companies have not changed strategies. Meanwhile, 16% said they’re not familiar with GEO or AEO.

“AI platforms act almost as a pseudo sales person,” Tinuiti’s Cornwell said. “Because ultimately the goal is to get the AI agent to recommend your brand. In order to do that, you need to appease the agents.”

With the rise of AI-generated search, some traditional search platforms are incorporating AI shopping applications for customers. For example, in November, Google introduced tools in Google Search and its Gemini app aimed at making browsing, comparing and buying products more natural for shoppers. The biggest update was to Google Search’s AI Mode, which lets people describe what they are looking for in everyday language instead of using keywords.

This focus on colloquial language is a common strategy among brands as they rethink their search strategies based on how they show up in AI-generated results. Another strategy for brands has been prioritizing content like blogs and landing pages that use more conversational language.

“There’s a lot of talk about AI search, and not enough talk about this shift to conversational search,” Cornwell said. “Regardless of platform, people are already doing that inside of traditional search. … A lot of best practices right now are things that SEO was doing 10-15 years ago. Really basic, like an FAQ, a bullet list or an HTML table. Those are formats that all of the LLMs like to see.“

Personal care brand Suave has tailored its own strategy as the industry shifts to focus on GEO and AEO. It started updating its website in September to optimize for GEO. “We’re in the middle of revamping our entire website and web presence for Suave,” said Rafael Lopes, Suave’s vp of innovation and brand equity. “We are [using] agentic AI, when it comes to SEO and copywriting, to make sure that everything we have related to product benefits and ingredient lists and how we talk about our products is AI-friendly.”

Similarly, big-box retailer Target has prioritized five key aspects with its AI search strategy: price, product, promotions, availability and policies. According to Ranjeet Bhosale, vp of digital product management at Target, the retailer’s website is updated to make this data “machine readable” and ready for AI engines to pull from — whether internally (through Target’s agents) or externally.

Huge’s Maleh said many of the marketers he talks with don’t have a plan to optimize their GEO and AEO strategies. “A lot of them ask, ‘Are people actually doing this?’” Maleh said. “We talk to them about the numbers around Google search — down or plateauing — and the explosion of people using Gemini, ChatGPT to get search results, and whether their content is ready for it.”

Maleh said brands shouldn’t think of optimizing for GEO and AEO as starting from scratch, but rather enhancing their existing content and search strategy to include GEO and AEO. “A lot of brands don’t understand that there are similarities, but you have to put the content structures in place to have it crawlable,” Maleh said. “Start with the simplest tools, like media intelligence, to do the heavy lifting.”

“It’s really about leveraging existing content to start, as opposed to thinking that you need new content to appear in these places,” Maleh added.

Several of the industry executives Digiday spoke with for this report also mentioned that brands should adapt their earned media and organic social media strategies for AI-generated search. “All of the LLMs eat this diet of earned media and I predict a larger investment in earned media,” Tinuiti’s Cornwell said. “In 2026, I think we’ll see a larger influence in LLMs from organic social.”

M7 Innovations’ Maher specifically pointed to social news aggregation and discussion platform Reddit as the perfect mix of search and social opportunity. “Reddit is a platform that has a mature ad ecosystem and it has 22 billion human-created posts that are used as authority in these large language models,” Maher said. “Posts that are six-,12- or 18-months-old still get sourced first within AI search engines and Reddit continues to be No. 1, if not in the top three, of the sources cited in large language search.”

“That’s why Reddit is so powerful,” he added. “I can take my ad spend on Reddit and build authority with large language models, and that is the alchemy that brands want.”

The way former IPG Media Lab’s Simon sees it, there is still room for growth when it comes to agentic AI. “The LLMs don’t need to be better, but the agentic capabilities need to be better. Their hooks into other platforms and systems need to be better to enable them to carry out tasks in a more useful way,” Simon said. “We will see development on that front over the next year or so, but we have another two or three years worth of development on the product side, before the features, integration and distribution catches up with the capabilities.”