As marketing agencies incorporate generic AI applications into their work, they are adapting their pricing models to take the cost of AI tokens into account.
Generative AI signals and responses don’t come for free. LLM developers such as OpenAI use tokens as a means of measuring AI computation; The text you type in the prompt and the output generated by the model are counted. The more tokens processed, the more computation is used and the higher the cost. And while tokens cost a fraction of a cent, they add up fast; A recent Coca-Cola advertising campaign required 70,000 signs and millions of tokens, According to its manufacturers.
“There’s no one model that’s a silver bullet… you have to use a lot of different models and they all have different token economics,” explained Jonny Rohrbach, co-founder and director of partnerships and operations at Silverside AI, a production studio backed by the founders of indie agency Pereira O’Dell.
Marketing services companies need to decide whether to consume AI overheads, or pass it on to customers. So far, there is no single answer to that question. Digiday spoke to media houses, creative agencies, production shops, full-service agencies and consultants to understand the situation on the ground.
‘We can’t charge customers for something we don’t know will work’
At full-service agency Merge, token costs are billed to clients on a metered, case-by-case basis, said Kyle Smith, its chief technology officer. Production shop Big Spaceship follows a similar structure, treating calculations as just another budget item, similar to catering or equipment rental. “We’re treating it like production costs,” said CEO Taryn Crowthers.
Silverside, which created vodka brand Svedka’s Super Bowl ad using Zen AI, uses a subscription model with “seat” pricing similar to that operated by companies like Salesforce, according to Rohrbach.
However, full-service agency RPA takes the opposite approach – it absorbs its clients’ token costs. Lisa Herdman, SVP and executive director of video investments, suggested that the benefits of the technology to customers are still too uncertain to build a pricing structure. “We’re still in the testing phase… we can’t expect to charge our customers for something we don’t know will actually work for them,” he said.
Chris Neff, Anomaly’s global chief AI officer, was similarly wary of passing on AI overheads. “It seems like a money grab,” Neff said.
‘The more committed they are… the cheaper we can make it.’
Pencil, the AI marketing platform from marketing services group BrandTech, is able to access economies of scale in agreements with LLM developers like Anthropic and OpenAI. Customers are charged a “generation credit” equivalent to a single chat response, an image creation or a second of video, each taking up one token. CEO and co-founder Will Henshall said they are not charged for tokens used on set-up and client onboarding.
“The customer is committed to a certain number of generations in pencil; they can spend in any model or use cases and [it] Price is based on quantity. The more they commit, the more we can go and negotiate with model providers, and the more affordable we can make it,” Henshall said. Customers are assigned a set number of tokens based on an initial cost estimate; If their needs exceed that amount, they “top up” with another installment.
Henshall said this pricing method keeps things simple for Pencil clients as it competes with creative and production agencies. “If we charge according to usage, our incentives are tailored to the customer,” he said. Tiered pricing, in part, ensures that Pencil’s largest customers are not subsidizing the token prices received by its smaller customers, and the company offers different pricing plans for smaller customers, reserving legal compensation for larger partners.
‘We do not want to create obstacles for customers’
Media agencies are adopting a different approach. For example, Horizon Media launched Blue, its general AI media planning and strategy platform, in December 2025. According to EVP and technical lead Krish Kuruppath, the majority of its token cost is attributable to client onboarding and the company’s ongoing development of Blue; He declined to say how much Horizon spent on tokens in a given year.
Since launching last year, Horizon has brought 40 clients onto the Blue platform, each with about a dozen logged in users. Kuruppath said the use of tokens will become a significant cost issue for the company once its user base reaches “tens of thousands.”
Currently it charges customers a “nominal” fee to start using Blue. “It’s cost recovery, not revenue,” said Domenic Venuto, chief product and data officer. “We don’t want to create barriers for our customers. But at the same time, we don’t want to eat away at our margins either.”
For example, media agency Kepler offers its own AI-enabled planning tool, called KipAIR. Peter Rice, vice president of data strategy and analytics, said it has not made wholesale deals with AI providers on token costs. Rice suggested it didn’t make sense to pass the cost of AI overheads onto an agency that typically works with clients on a retainer basis.
“We want this to be integrated into the way we deliver for our customers,” he said. “Impact is the metric we care about. Token usage is the fuel to deliver that impact.”
Full-service agency Lerma/ takes a different approach. According to chief data and intelligence officer Josh Archer, the company has a bulk-buy agreement with an LLM provider (Archer did not specify which provider). As such, the cost of the AI token is included in its initial pricing estimate for clients – at cost.
“We’re not marking it down, we show it as a line item, similar to other costs,” Archer said. Should the agency undervalue and find itself with a surplus of tokens, he said it would provide additional assets to clients. “We try to give more. This is a great opportunity for us to push ourselves,” he said.
‘You want to compete on merit’
BrandTech’s wholesale deals for tokens invite comparisons with one of advertising’s most controversial practices: major media buying. If an agency can acquire large quantities of tokens at a discount, what’s to stop them from passing that cost on to clients at a markup?
James Londel, founder and CTO of consultancy Evelyn, suggested that pursuing a dominant media model would likely prove a dead end for agencies. “The big economic shift is not symbolic arbitrage. It’s labor compression. If AI runtimes meaningfully replace reporting teams, duplicate dashboards, and middleware connectors, the financial upside dwarfs any marginal wholesale discount,” he said in an email.
Henschel agreed. “You want to compete on merit, not lowest price,” he said.
Even if they aren’t attempting to add a price premium, agencies that forgo token costs run the risk of “focusing solely on cost and margins rather than focusing on the value and architecture that is right for our agency and client,” said Jess Lewis, Crossmedia’s global chief technology and data officer.
It’s possible that token pricing could become part of agency pitches in the coming months and years. Ruben Schreurs, CEO of Ebiquity, told Digiday that he expects the consultancy will in time be asked to audit an agency’s token usage, just as it audits media spend now. “If it’s part of the contract… then yes, we will audit it,” he said.
Knowing that client CFOs will keep a close eye on any pass-through costs, agency leaders are unwilling to reduce the use of AI to savings of dollars and cents.
It’s more effective for agencies to focus on how AI can be used to benefit a client’s business, Schreurs said. “It is less about inputs and more about how incremental business returns are being delivered,” he said.