ChatGPT Ads in Switzerland: From keywords to context

In traditional search advertising, a person’s intent is often inferred from specific search queries and other signals. In ChatGPT, however, users describe entire situations and circumstances – including their goals, budgets, requirements and the constraints shaping their decisions.

For advertisers, it is therefore no longer only important what someone is searching for, but also why, in what situation and with what requirements. An ad no longer simply needs to match a search term. It needs to be helpful in the specific situation and offer a meaningful next step at the right moment.

Perhaps the biggest difference from traditional search advertising lies in how users express their intent – and which signals are relevant to ad selection.

Someone searching for car insurance on Google today might use search terms such as “car insurance Switzerland”, “comprehensive cover for an electric car” or “compare car insurance”.

In ChatGPT, the same search could be much more detailed:

“I’m buying my first electric car for around CHF 45,000, drive approximately 15,000 kilometres a year and would like comprehensive coverage. Good protection against parking damage and a straightforward claims process are important to me. Which insurance might be right for me?”

For advertisers, this makes a significant difference. The query contains more than individual terms. It describes an entire decision-making situation: the type of vehicle, budget, usage, desired insurance cover and specific requirements.

This is precisely where the advertising logic behind ChatGPT Ads comes in. When selecting an ad, OpenAI considers factors including the context and intent of the current conversation, as well as information from the ad and landing page.

Advertisers can also provide so-called Context Hints within a thematically focused ad group. These provide additional information about what an offering provides, who it is relevant to and the situations in which it may be useful. Context Hints do not work like exact-match keywords or fixed targeting rules; instead, they give the system additional context for assessing relevance.

What matters, therefore, is not simply whether a particular keyword appears, but whether the advertised product or service, based on the available signals, matches the identified need and specific situation.

This makes it increasingly important for companies to understand exactly: What do we actually sell? To whom is our offering particularly relevant? In which situations does the need arise? What questions do potential customers ask? And how do they describe their problem or requirements?

The more precisely companies understand and articulate these connections, the stronger the foundation they create for being considered in a relevant conversation.

Keywords therefore remain a relevant signal. What increasingly matters as well is the context in which a need arises and how users express their intent within the conversation.

For ad delivery, what increasingly matters is how well an offering and its relevant use cases can be described.

Based on our experience so far, ad inventory within a ChatGPT conversation is more limited than on traditional search results pages. Advertisers therefore have fewer potential opportunities to serve ads – making it all the more important to understand in which situations their offering is genuinely relevant.

This is not just about the ad itself. It is also about the information a company can provide about its products, services and relevant use cases.

A company that merely states that it offers car insurance, for example, provides relatively little context. A company that can additionally explain precisely which vehicle types, risk profiles, needs or life situations specific benefits are relevant to creates a much stronger foundation for contextual matching.

This gives greater importance to first-party knowledge that extends beyond traditional CRM data: structuring a company’s own product and service knowledge so that both people and AI systems can understand when an offering is relevant – and why.

First-party data and knowledge can therefore become an important differentiator between competitors.

At present, the focus is primarily on standard ChatGPT ads and feed-based product ads. Ads are currently shown exclusively to users on the ad-supported Free and Go plans. Plus, Pro, Business, Enterprise and Edu remain ad-free.

Standard ChatGPT ads can be displayed below a ChatGPT response based on the context of the conversation and include elements such as a headline, description, image and landing page. They therefore address precisely those situations in which users have already described a specific problem, need or goal.

Feed-based Shopping Ads rely more heavily on structured product data. When commercial intent is identified, relevant products can be displayed with additional information such as an image, title or price.

These formats demonstrate particularly clearly just how important the underlying data is becoming. Product feeds need to be technically sound, but they must also describe products in a way that makes their relevance to different search and decision-making situations clear.

Additional formats and buying options are expected to follow. Which of these will become available when is currently still open.

As interesting as this new advertising logic is, ChatGPT Ads should not be launched with the expectation that they will immediately replace an established lower-funnel channel.

ChatGPT’s advertising offering is still at an early stage. Reporting and campaign management capabilities are currently far less mature than those of long-established advertising platforms.

Available metrics already include impressions, clicks, spend, CTR, average CPC, average CPM and conversions. More advanced competitive metrics such as Auction Insights or Impression Share are not currently available.

This makes both benchmarking and optimisation more difficult.

What’s more, not every offering is equally suited to a conversational advertising environment. Initially, ChatGPT Ads are likely to be particularly interesting in areas where people conduct extensive research, compare options or need guidance before making a decision – such as complex B2B solutions, higher-priced products, insurance, financial services or services that require extensive consultation.

Our assessment is therefore: test, yes – but do not shift budgets blindly.

Precisely because the advertising offering and best practices are still evolving, we do not believe the real first-mover advantage lies in going live a few days earlier than the competition.

The advantage lies in starting with a solid setup and quickly building robust learnings.

Companies should therefore put the following foundations in place now:

• Open an Ads account and initiate Advertiser Verification and billing early.
• Structure product and service data completely and accurately and, where relevant, set up product feeds.
• Clearly define which audiences and use cases individual offerings are relevant to.
• Collect and document typical customer questions, needs, requirements and decision-making situations.
• Prepare ad groups and Context Hints around relevant use cases.
• Make product knowledge from Sales, Consulting, CRM and Product Management systematically accessible.
• Prepare ads, landing pages and content so that they communicate not only product features, but also their concrete benefits.
• Define in advance which interactions and conversions should genuinely count as success and prepare the corresponding tracking.

Based on our experience so far, advertisers should allow up to two weeks from opening an account to going live. This time can already be used to prepare the campaign setup: targeting, Context Hints, ad creation and tracking can be set up in parallel so that everything is ready once the account is approved.

We have already gained experience in markets where ChatGPT Ads became available earlier and are now analysing these learnings for the Swiss market as well.

When getting started, what matters is understanding when an offering is genuinely relevant, to whom – and structuring that knowledge so the system can work with it.