ChatGPT Ads and the evolution of consumer behaviour

An overview of the ChatGPT Ads open beta and the changing nature of the customer journey

You may have noticed if you've explored our website header menu recently that we’re helping brands get started with ChatGPT Ads.

So, we thought we’d give a brief overview here of how the ChatGPT Ads open beta currently works in the UK (and what we are seeing in the platform via our partner, StackAdapt).

It’s a fascinating time for the ad industry.

The background

We probably don’t need to belabour the point – consumers are shifting from keywords to conversations.

And these conversations matter to brands in part because of their scale. There are 2.5 billion prompts in ChatGPT every day and 70% of that usage is ‘non-work’.

Yes, referral traffic to websites might be at <1% of total traffic, but that’s precisely the point. Brands must go to the consumers (as ever).

This shift to conversations means that advertising is also moving from keyword- to intent-based.

The modern customer journey is no longer a straight line from a SERP to a checkout page. People are using ChatGPT for extensive research and product comparison. They validate their answers, explore their options, then return (perhaps direct) with conversion intent.

LLMs are actively shaping how brands get discovered in this new "agentic era." And it follows that optimising paid search for AI is now critical to campaign success.

The ChatGPT ad formats currently available in UK in early access beta program:

The ad formats are clearly separated from ChatGPT responses and they appear for users on the AI assistant’s Free and Go plans.

There are three types of ads currently being placed in the UK beta:

• Sponsored Answer Cards (typical ad, does what it says on the tin)
• Product Spotlight Ads (pulled from an advertiser product feed; shopping-style ad – see image below)
• Contextual Sidebar Ads (right-hand-side placement).

How the targeting and bidding works

Ads appear when they are relevant to a user prompt.

1. A user creates a prompt
2. OpenAI runs a brand-safety check
3. Relevant ads are retrieved
4. A relevance model selects the best match

This open beta uses a proto-auction model. The price depends on how many advertisers are relevant to a given prompt.

More niche and contextually aligned creative can result in lower effective CPMs. Advertisers can pursue both CPM and CPC strategies.

Selection logic is model-driven, and we can say that:

• Highly relevant creative has a strong advantage. Inputs may include the ad title, body copy, landing page content, and ad group keyword inputs used for retrieval support.
• Brands benefit from multiple contextual creative variations.
• The relevance scoring process is still changing regularly.

What can advertisers expect from the platform in the open beta?

The infrastructure is evolving, so there are obviously going to be a few pain points and bugs. There is currently limited reporting, focusing on impressions, clicks and conversions (if configured). There is country-wide (UK) targeting, with no option yet of regional strategies or demographic overlay.

That being said, things are moving very fast.

If you’re interested in getting involved, get in touch. The setup and approval process takes 1-2 weeks, and we help with managing tracking, campaign structure, creative effectiveness, product feed integration and performance analysis.

Brand and product discovery hasn’t undergone such a shift since the early days of Amazon. Now is the time to get testing!