From the first time I ran a car insurance query in ChatGPT, Perplexity, and Gemini (Bard?) in 2023, I knew, “yeah, they’re going to have to shove ads all over this”.
This was because years of status quo had been upended in a single response. In my AI results, incumbent organic insurance leaders, the price comparison websites, were seldom mentioned if at all, due to how AI interpreted the queries (why would it send to an aggregator when itself was meant to be an aggregator of information? The next logical step would be an insurer). This was so significant that it prompted a study in partnership with Similarweb (link), and the results proved it at scale. This was something different.
It was hardly a bold leap of logic on my part; just this one category (personal insurance) is worth billions in advertising revenue (just the small part of it I led was investing £40m a year). Disruption to this traditional (for the last 20 years, anyway) customer journey poses a huge threat to incumbents and a huge opportunity to those who play this new game well. If it didn’t get ads, it would potentially destabilise the current search ecosystem.
ChatGPT ads are coming
So, two years on, are we finally seeing the other foot drop? The “last resort” (as Sam Altman originally badged it) of advertising in ChatGPT is now “testing” in the US. The announcement came less than a week after Google announced their own version, so it’s not surprising, but are the two AI Advertising announcements the same? No.
Difference 1 – People are used to the commerciality of Google – many know the top links are ads now, and don’t really care, especially when they have high intent – it’s a call and response – if the ad looks to match the query, do they really care if it’s organic or not? In ChatGPT, we’re used to getting unfiltered responses, paying only for greater frequency or complexity.
Difference 2 – Tiers – Google doesn’t have tiers (not counting YouTube Premium, but those 4 people don’t count). This SaaS model is more akin to a Canva sub, or your Strava, except Canva stops at limiting your features, not pushing ads.
Difference 3 – How people use the tools. They aren’t asking for advice on Google; they are asking for a series of options. This is different; if you ask AI to give you the best mortgage for you, you expect it to do the due diligence, not just throw up the one that bid the most for ChatGPT Ads. Anecdotally, more people are asking long-form question type queries into Google now, but the results are a far cry from the depth you’ll get out of Perplexity, for example.
ChatGPT and Financial Compliance
Which brings us to another nuance of ChatGPT in Financial Services – OpenAI said they would start to limit the financial “advice” it provides. ChatGPT is getting its own EEAT (link to EEAT article) essentially. What does that mean for us, financial services marketers? Will it just start pointing to another aggregator or broker? In which case, does its value over search engines start to diminish? Will we see a rubber banding back to the status quo, or have things fundamentally changed?
These are all questions that will be easier the more we observe how real people act when presented with the ads. Until then, let’s dig into how OpenAI are trying to mitigate some concerns, what it’s going to actually look like when we start seeing the ads.
The principles outlined by OpenAI for their advertising are below. I may be misinterpreting them, so my personal take might be overly cynical, from a consumer perspective, but it does give a feel for what might be possible as an advertiser:
Mission Alignment – Ensure AGI benefits all of humanity.
Answer Independence – Adverts will not influence the answers ChatGPT gives the users.
Conversation Privacy – Conversations will not be shared with advertisers, and they will never sell your data
Choice and control – you will have control over how your data is used.
Long-term value – The platform is not and will not be optimised for time spent in ChatGPT.
I’ve commented elsewhere on what I think of these principles, but for the sake of brevity, let us just say that some of the above principles would make traditional targeting of intent or audience a bit more difficult than other advertising channels.
Can we predict how this will turn out?
As stated, it’s easy to just think of this as another Google / Microsoft Ads and act accordingly, but it’s not the same. In search, ads disrupt the experience in their eyes. Or at least not in a way that people aren’t used to. We’ve been lobsters in this steadily warming pot for 20 years, it’s only marketers that really notice the bubbles.
So, in which channels do ads disrupt the experience? Prime Video, YouTube, Spotify, Netflix, Disney, ITV. Perhaps these are more appropriate parallels to look at when we see the impact, or even the commercial models (could ChatGPT end up making you watch an ad before you can query? Will a conversation suddenly drop in a sponsored segment before continuing?) . Given that some of these channels are themselves are relatively recently experiencing true paid presence for the first time, it would be worth using them to help predict behaviour.
Or maybe people won’t even notice, or indeed, appreciate it if they do – let’s hope so, for the good of our advertising.
What to do now?
Just as we’ve been saying not to neglect your SEO because this shiny AI has come along, don’t neglect organic AI because Paid will soon be an option. It’s quite feasible that organic strength will influence ad performance (as we see with things like landing page relevance and ad rank in “traditional” paid search).
That said, carving out some test budget now wouldn’t be a bad idea. Speak to your agency if you have one, ask them their thoughts. We’re already doing this with clients.
In the next 30 days:
If you don’t know already, audit your current LLM presence. It is your baseline.
Have an account that *isn’t* pro so you’ll actually be able to see ads when they appear
Prepare ad messaging that might help address the most common questions leading to your product. Think how that might look in the ChatGPT interface under the organic responses. This may need to be a bit more nuanced than standard PPC ads – “learn more” rather than “buy now” (in FS anyway, E-commerce will likely be more direct).
In the next 2-4 months
Carve 5% or so of the budget as a test allocation, assuming ads have started to work correctly over in the US. Pull them from your experiment budget if you have one, don’t sacrifice effective channels for this bird in the bush.
Keep an eye on any regulatory response and customer response in the US. They are the canary in the mine (what’s with all these bird analogies?) for us. Particularly watch for things like SEC guidance that might be related to potential FCA guidance on declarations, transparency, and “Balance” in the messaging.
If you’re a big player, watch what the fintechs are doing, they are often more experimental. If you’re a fintech, watch what the incumbents are investing in and move quickly to get there before it saturates and gets expensive.
In 6 months
It’s a bit bold predicting this far ahead in the face of AI, but I’d expect a lot of these elements in place by summer so you aren’t left behind.
Work on your organic AI recommendations using channels like Digital PR – much like we’re seeing with Google AI Overviews – “entities” that it’s confident address the question get a hard link, instead of text only. It’s reasonable that it could have similar effects on paid ads. A strong level of trust and brand presence is as important as ever, and could play a significant part in your paid performance (as it does now). Don’t risk tunnel vision – if this works for ChatGPT, we’ll see expansion on Perplexity, etc. Some of those might even do it better. Your agency should be keeping an eye on this for you.
Reassess your organic visibility on Google, and watch for impact on other channels – this will inevitably impact channels, none exist in a vacuum. Attribution will be as sticky as ever, especially as much of this will be upper funnel. Watch for the signal in all the noise.
