ChatGPT Ads in Sweden so far: results, buyer journey and measurement

How have the first weeks gone?

On August 24, sponsored messages began appearing below the AI answers for Swedish users on ChatGPT’s Free and Go plans. A week later, OpenAI opened its self-serve tool Ads Manager to businesses in Sweden and 30 other European countries.

The ad agency AdRelevance, which runs Google Ads for several of Sweden’s largest e-commerce companies, has tested the platform since launch. In an article on the trade site Ehandel.se, CEO Karl Lindberg writes that ROAS for some clients is on par with or above Google Ads. According to him, however, product feed campaigns have so far performed better than other ad types.

But the evidence is still limited. Lindberg shares no figures, and the results apply to a limited number of clients on a platform that is changing at a rapid pace.

What counts as a conversion in ChatGPT Ads?

How many conversions a ChatGPT Ads campaign reports depends on the time windows the advertiser has chosen. If the user converts within 30 days of clicking the ad, it counts as a conversion for the campaign.

It is also possible to attribute conversions to the ads on the sole condition that they were part of the customer’s buyer journey.

If this is turned on, purchases, inquiries and other conversions that happen within a day of someone seeing the ad are counted, even if they didn’t click it. This applies to purchases in an online store as well as inquiries and other goals that have been set up.

In Google Ads search campaigns, an ad click is normally required for a conversion to be attributed to the ad. If you compare ROAS from ChatGPT Ads with ROAS from Google without taking the settings into account, there is a risk that the figures are calculated differently. By default, Meta counts purchases up to seven days after a click and one day after a view. On that point, ChatGPT Ads is more like Meta than Google’s search ads.

This autumn, OpenAI has also changed what is included in the conversion total. In August, purchases and other view-through conversions were reported in a separate column, according to PPC Land, but since September they are included in the total if the view window is turned on. If you compare a report from August with a later one, it can look as if results have changed, even though only the counting method has. Conversions also show up only after one to two days, so the most recent days’ figures are often incomplete.

Where in the buyer journey do the ads appear?

When someone asks ChatGPT a question, it is often about a need, such as what solves a certain problem, and rarely about a specific product. The ad therefore reaches people who are still comparing and researching, while a Google search ad often reaches someone who is already close to buying. Lindberg therefore places the channel between Google Ads and Meta Ads. According to him, many people still go to Google to complete the purchase.

For your evaluation, this means direct purchases show only part of the visibility the channel generates overall. If someone discovers your brand in ChatGPT and then buys via Google, the purchase is credited to Google. That is why it is worth tracking how many new customers the channel reaches and whether branded searches increase when you advertise in ChatGPT.

Discovers

The customer discovers that something is missing or could be better.

Meta Ads

Sponsored

Compares

The customer does research and weighs options.

ChatGPT Ads

Which headphones block out noise on the train?

Sponsored

Buys

The customer knows what they want and looks for the right store.

Google Ads

headphones [model] best price

Ads

Less competition for attention in ChatGPT

The fact that product feed campaigns have performed better for e-commerce companies is partly down to the format.

Google Shopping shows several product ads, often in a rotating carousel. In ChatGPT, by contrast, only one advertiser is shown per answer in the format AdRelevance has tested, which means less competition for attention. The SEO tool SE Ranking reached the same conclusion in a review of the US version of ChatGPT: one ad was shown per answer, and roughly one in four commercial questions got an answer with an ad.

According to Lindberg, clicks often cost a little more than in Google Ads, but OpenAI has not yet published any comparison figures for click prices. When each click costs more, it also needs to lead to a purchase more often for the investment to pay off. That is why it can be smart to start with products that already sell well, have good margins and a competitive price.

Titles, descriptions and attributes in the product feed determine which conversations a product is matched to, so it pays to fill them in carefully. The advertiser can also add context hints, meaning words and phrases that can be linked to the campaign. The system then weighs them in as signals when choosing which ad to show. There is, however, no report showing which conversations led to an impression.

What happens next?

OpenAI is developing the ad platform quickly, with new ad formats and changes in how results are measured. That is why it is important to decide how impact will be measured before the test starts. It then becomes easier to tell whether ChatGPT Ads actually brings in new customers or whether the ads are credited with purchases that would have happened anyway. Setting a clear baseline from the start also makes it easier to evaluate the channel’s impact as the platform evolves.

Want to know how ChatGPT Ads would fit your advertising? Read more about our ChatGPT Ads service or about how the ads work in the chat.