Physical Address

304 North Cardinal St.
Dorchester Center, MA 02124

Marketers have a new audience to worry about — large language models


Modern marketers know that every new piece of work they put out into the world in service of their brand can be met by many different audiences.

There is a target group of consumers that marketers hope their latest campaign or activation will resonate with. There are competitors who follow every move of their competing brands. There are online culture warriors waiting to turn an advertiser’s misstep into a cause célèbre. And now there’s one more group that marketers need to keep in mind – the big language models, the underlying technologies that underpin generative chatbots and AI applications.

With AI platforms like ChatGPT and Perplexity gaining traction, tech firms are creating new ways to understand how large language models perceive their brands—and what AI responses say about them. One of the latest developments comes from Profound, an SEO startup focused on AI search platforms, which today presents a way to estimate conversation volume rather than just analyze AI output.

Just as Google provides data on traditional search volume and trends, Profound hopes its updates will identify user intent when interacting with chatbots — something that’s still not possible with the current nature of generative search. One way brands can use the Profound panel: To inform how they create, test and optimize content based on what people are chatting about with bots across platforms. It could also help them better monitor the competitive landscape based on the companies people mention when using generative search.

“If you ask 21,000 questions to major AI answer search engines about fast food in 30 days and McDonald’s comes up 50% of the time, that’s useful,” said Digiday CEO and Profound co-founder James Cadwallader. “Brands are very interested in this. They pay for it. McDonald’s is a very important figure for them. But that doesn’t reflect how many people on the internet are looking for fast food in these AI answers.”

If there’s a dataset of queries for, say, “Toshiba microwaves,” Cadwallader said Profound’s predictive model can estimate how many searches mention other microwave brands. He wouldn’t disclose Profound’s sources for the external data, but said it’s logged, privacy-compliant, and still in beta.

To analyze the meaning and context of chats, Profound uses billions of semantic embeddings in real time and then uses a proprietary sampling algorithm to identify conversational patterns with statistical significance. He also developed new approaches to optimizing temporal queries, as traditional methods break down when dealing with the asynchronous nature of AI conversation flows.

Some SEO experts like Kevin Indig already see the benefits of using Profound’s data volume, API access, and citations within Profound as appealing. He said brands could use it to create or optimize content based on what people are chatting about with bots on different platforms. One example Indig gave was its client Hims, which is currently using the Profound platform to fill content gaps or enhance existing materials.

“We noticed that there was a lot of demand for topics related to diet programs,” said Indig, an advisor to the brand’s growth teams. People are looking for the best diet or what is the easiest diet, but in a more sophisticated way than on Google. We realized that while we have weight loss content, which is a massive topic for Hims, we don’t specifically compare diet programs.”

Although the solutions differ in detail, each aims to help marketers better understand how LLMs like Gemini, ChatGPT and Meta’s Llama display their brands. Profound is not alone in its launch. Brandtech-owned marketing firm Jellyfish unveiled its own “Share of Model” solution earlier this month. Jellyfish’s first clients using the new tool include whiskey brand Danone and Pernod Ricard Chivas Brothers.

Understanding search behavior

Changes in user search behavior have required “a completely different way of thinking about optimization,” said Jack Smyth, director of AI solutions, planning and insights at Jellyfish. His company’s tool uses APIs from various platforms to give a “bird’s eye view” of LLM branding across video, images and text assets used to train AI models.

Users can also upload creative assets—catalogues, websites, or social media campaigns—and evaluate how well they match how individual AI models think about different categories and concepts. This could help boost product recommendations like AI search platforms create new features for e-commerce. “Share of Model” also integrates with Brandtech’s other AI content creation tools, such as Pencil Pro.

“Our view is that anything you post on any platform is now part of someone else’s training set,” Smyth said. “Every piece an advertiser creates is now a brief for someone else’s model.”

Web users are increasingly turning away from traditional search or browsing tools in favor of artificial intelligence “agents” such as chatbots. According to YouGov, 66% of 18-24 year olds and 51% of 25-34 year olds regularly refer to AI tools for product recommendations.

As usage grows, so do the stakes. Platforms like ChatGPT and Google AI Overviews answer questions by searching vast amounts of information gathered from the wider internet. However, summary answers may not be consistent with a carefully crafted brand position in the market – and may be completely wrong.

“Consumers are looking for clarity in a sea of ​​options. As brands, we need to ensure that our products are represented in these critical AI-driven reactions,” said Gokcen Karaca, head of digital and design for Chivas Brothers, Pernod Ricard’s whiskey brand.

Artificial intelligence applications are a priority for marketers in the new year. A recent survey of more than 700 marketers by German agency network Serviceplan Group found that 81% said AI was their top priority – above branding, measuring marketing ROI or CRM investments.

Why the impact of the LLM is important for marketers

“LLMs will increasingly influence customer behavior and the Share of Model platform allows us to track and compare the perception of each LLM,” Catherine Lautier, vice president and global head of media and integrated communications for the Danone brand, said in a statement.

Other companies that have developed similar tools include Hubspot, which in August, it launched a free tool called “AI Search Grader”. and recently added Confusion to the list of AI platforms that can be analyzed. Another is BrightEdge, an SEO marketing company that recently introduced a way for brands to measure how they appear in Google’s AI reports.

At $72,000 per year—the minimum price for an annual subscription to the Jellyfish model platform share—these solutions are a fraction of what some companies spend on their overall AI investment. Coca-Cola, for example, committed to investing more than $1 billion in Microsoft’s AI solutions in April.

Opportunity costs need to be considered, Smyth said. Ignoring how LLMs interacted with a brand’s digital output and how they could influence potential customers could cause problems as the use of generative AI tools becomes more widespread.

“People may be less likely to interact in a chat interface, but they may use an agent to shop,” Smyth said. “Or you may need to think about creating a website that is not primarily for people; it’s actually designed as a store of information for models to go and create proactive recommendations across platforms.”



Source link

Leave a Reply

Your email address will not be published. Required fields are marked *