Wesley ter Haar: brands must be usable for the AI agent
Monks' Chief AI Officer expects cannibalisation of generic search advertising
Marketing Report interviews Wesley ter Haar, Co-Founder and Chief AI Officer at Monks
How is Monks responding to the arrival of advertising on ChatGPT in the Netherlands?
We have a clear responsibility towards our clients: how do we translate changes in culture, consumer behaviour and technology into a business advantage?
Consumers have a limited amount of time and attention. As their orientation, advice and purchases increasingly run through LLMs, the marketing budget has to move with that.
ChatGPT Ads is one part of it but three layers come together here: paid placement, visibility through trusted sources and signals (GEO) and agent-ready product data and processes. The real step is from findable to executable: a brand must not only appear in the answer but be usable for whatever an agent does next. That is where we help our clients.
When is advertising on ChatGPT advisable and when is it not?
If you operate in a category where expertise, personal preference and product comparison matter, a mature LLM and agent strategy is important. Advertising is still a small part of that. This is a good moment to test and learn; I expect this shift to continue.
If your audience is barely active there or your product information is not in order, I would not move substantial media budget towards it yet. Then the work starts with those basics.
How does an advertiser know what results ChatGPT advertising has delivered?
OpenAI now reports the familiar basics, such as impressions, clicks and conversions. The more interesting question is how much of the funnel ultimately takes place inside the LLM, or is carried out from the LLM by agents.
Measurement methods will then have to move with it. How much incremental sales does it deliver, which actions are actually completed and can we eventually be paid on that result? An attributed conversion is not yet demonstrable incremental sales.
What do you make of the rates ChatGPT charges and how can they be justified?
At the moment you are paying tuition and paying for the hypothesis that an ad inside an LLM environment has more impact. That added value still has to prove itself.
A higher cost per click can be perfectly defensible if it demonstrably delivers more business. The question is how much extra result you get back for your investment. That is where the pressure will come to lie.
How do GEO and LLM advertising relate to each other? With good GEO, do you still need advertising?
I expect brands to be compared explicitly on their offer, price, availability and evidence more often as consumers leave more of their considerations to LLMs and agents. That makes the information a brand is judged on more important.
GEO is the strategy for creating and optimising the information, content and context that LLMs and agents use. LLM advertising adds a recognisably paid marketing message to that.
Good GEO can reduce dependence on advertising but it does not automatically make paid visibility redundant. You have to test where it still delivers extra value.
How do you expect the media mix of advertisers to change? Which media will ChatGPT advertising come at the expense of?
My estimate is that it will initially come mainly from innovation and test budgets. If the results are positive and spending is scaled up, I expect the cannibalisation to come mainly from generic search advertising, affiliate marketing and parts of retail media and social media.
If consumers start running their choices and purchases through LLMs and agents on a large scale, a bigger shift is needed. Then the focus moves further towards CX, where the C stands for consumer as well as computer.
Everything depends on trust and reliability. What effect does advertising in LLMs have on that?
Free digital products are often paid for with advertising revenue. That is a reasonably well-known model. Yes, LLMs are more personal but advertising in this environment does not have to be more irritating than a non-skippable pre-roll.
For trust, what matters most is that it stays clear what is paid for and that an advertiser cannot buy the independent answer. I also expect more interesting ad formats to emerge that connect better to what someone is trying to do. Those can be more pleasant for the consumer than the current interruption model.
Initially there will be no personalised advertising in the European Union because of the GDPR. Opt-in still makes it possible. What effect does that have on the advertising options and how can advertisers respond?
The GDPR does not prohibit personalisation but it does set conditions for it. Consent is not a free pass. At ChatGPT, personalised advertising is initially not available in the European Economic Area.
So for advertisers the first opportunity lies in the context of the conversation. Someone says what they are looking for, what matters to them and which trade-off they are making. You can already connect to that very relevantly.
Ultimately I expect people using a free product to prefer useful advertising over random advertising. The best version adds something: a fitting offer, a relevant comparison or help with the next step. That is where the value exchange has to be.
Do you think branded content will appear inside LLMs that influences the results? Why or why not?
Yes, I expect many more interactive possibilities there. A product page that adapts to your question, a configurator inside the conversation or a personal product demonstration after you have uploaded a photo. That gives brands much more room to make their story relevant.
I can also imagine paid services emerging that let brands manage their official knowledge base and make up-to-date information available to LLMs and agents. That can improve how a brand is represented. Buying a place in the independent answer goes a step further. I expect the separation between that answer and the paid experience to remain in place for now.
How do you rate the way OpenAI is dealing with the Dutch market as an incoming ad seller?
The Netherlands looks like a good test market to me: relatively small, digitally mature and critical. Self-service is now available to Dutch advertisers as well, so the option to test it yourself is becoming more concrete.
I think it is too early for a grade. Above all I would look at how easily advertisers can get started, what support they get and how quickly experiences from the market are translated into a better product. Let us look again at the end of this year; by then we can judge more than the introduction.
What are your expectations of advertising on ChatGPT?
In the short term I mainly expect a lot of testing and learning. Advertisers and OpenAI both have to discover which formats suit the way people use an LLM.
In the longer term I expect a larger part of the purchase process to take place inside the conversation, or to be carried out by an agent. An ad can then lead directly to comparing, configuring, booking or buying. That makes it more important for brands that their offer is actually executable: current prices, available products and processes an agent can work with. That is the bigger change I am watching.
How do you think Anthropic (Claude), Google (Gemini) and Mistral Vibe will deal with advertising?
Anthropic has explicitly opted for an ad-free Claude and earns from subscriptions and business contracts. That fits the emphasis on work and productivity.
There is also an important difference with many earlier internet services: businesses and consumers already pay directly for AI that does valuable work. Advertising is mainly interesting for financing usage people do not want to pay for. Not every AI company has to make the same choice there.
For Google, advertising is the obvious route; it is already developing ad formats for its AI search experiences. The separate Gemini app is a distinct strategic choice. I can imagine Google using an ad-free experience there as a competitive advantage for the time being.
At Mistral I see more logic for now in paid professional use, in line with the positioning of Vibe around work and coding. To become a large advertising business you first have to build a lot of structural consumer usage.
Why do advertisers need Monks for advertising on ChatGPT, when in theory they could do it themselves?
If you only want to switch on a campaign, you can in principle do that yourself. Our role becomes interesting when you want to move as a brand with the bigger change in how people find information, form preferences and make purchases.
More questions then come together. How is your brand understood by an LLM? Which information makes your offer convincing? Can an agent work with your products and processes? And does that ultimately deliver extra business?
That is where we bring strategy, creative, media, data and technology together. The advertising is one part of that. Our responsibility is to translate that change into a business advantage for clients.
Follow Marketing Report on LinkedIn
Subscribe to the Marketing Report Newsletter
Send us your press releases, columns, interviews…
Join Marketing Report in our LinkedIn Groups
