How marketers can prepare for advertising on Chat GPT and other AI agents

When Google released Bard last year, I thought it was a good time to revisit the prognostication from my 2010 book, Everything I Know About Marketing I Learned from Google.

When Google released Bard last year, I thought it was a good time to revisit the prognostication from my 2010 book, Everything I Know About Marketing I Learned from Google.

I took a victory lap of sorts in this April 2023 article owing to the final chapter of my book, aptly titled "Future-Proofing" in which I called for "app-sisstants" or "search-and-act engines" to become the next big breakthrough in organizing the world's information and making it universally accessible and useful – which is technically still Google's mission, although you wouldn't be blamed for thinking it's OpenAI.

While we're not calling them app-sisstants today – "agents" is the term de rigueur – we are seeing innovation and adoption at a scale that may seem incredibly rapid, but not when you consider that it's been 14 years since I predicted it.

In that last chapter, I interviewed Dag Kittlaus, the founder of Siri – which had not yet been bought by Apple – and he laid out an example of how an app-sisstant could work. He said, “With travel, Siri will know if you got screwed on [an airline flight] connection and will proactively find hotels nearby." Apple intelligence, anyone?

Now, in the spirit of keeping myself honest (and humble-ish) I should point out that I also heralded Hunch as an app-sisstant that portended the future of what we now call agentic AI, and let's just say my hunch was wrong on that one.

But back to posting my W's, in that same chapter, I talked about the various ways these app-sisstants could be monetized. I said, "I’d think a paid subscription model would lend itself better to objectivity." And that is certainly how the first wave of agenctic AI apps started.

But then I noted that "perhaps Google, like it has tried to do with virtually every other platform marketers want to infiltrate, will find a way to appropriately serve ads in apps like Hunch and Siri."

Indeed, it seems like this will be the path forward for app-sisstants, or agents, with Perplexity officially announcing an ad model three weeks ago and perplexing the market with $50 CPMs. (Although as I commented on X, when we first launched fixed placements on Ask Jeeves they went for $50-100 CPM – and that was 20 years ago!)

So maybe it wasn't that big a leap for me to make the prediction this week in Street Fight that OpenAI will launch ads on Chat GPT in 2025. What was ironic though is that the day my interview was published – and the Q&A had taken place earlier, mind you – The Financial Times dropped the bombshell that OpenAI was exploring advertising.

To be clear, I'm not here to say, "I told you so" (although I did!) but I am here to say that all marketers should be readying their plans to ensure they capitalize on the opportunity.

This is how I closed my book and dare I say all the takeaways still apply:

However marketing is defined in the future, applying each of the Googley Lessons we’ve learned will be critical to success in this brave new world of search-and-act engines.

1. Relevancy Rules. If you want to be in the considered set of any app-ssistant, you’ll have to make your company, product(s), and service(s) relevant to the task or decision at hand.

2. Tap the Wisdom of Crowds. The recommendation algorithms will be taking into account what others think about you. To ensure your brand gets the nod, you’ll need to listen and respond to the crowds before rallying them behind you.

3. Keep it Simple, Stupid. As smart as machine-learning can be, it will always favor simplicity. Make it easy for apps – and their users – to know what you do and where to bucket you.

4. Mindset Matters. While it will become readily apparent when and where people are in buy-mode, seeking out those apertures today will ensure you stay top of mind – and in business, for that matter – when app-ssistants take over.

5. Be Where Your Audience Is. In the hopes of inserting your brand into every possible relevant category of decision-making, you’ll need to deconstruct it and distribute it wherever your audience is, physically and virtually.

6. Don't Interrupt. With information talking to information, you’ll be tempted to interrupt the process for an overt commercial message. Don’t.

7. Act Like Content. App-ssistants speak digital, not analog. The inputs they seek will be acting like content—digital content.

8. Test Everything. Ambient findability is unchartered territory. The only way to truly know what will work – and keep working – is to test everything.

9. Track Everything. The steps in the app-ssistant’s decisioning process may be opaque, but the inputs and outputs will all be accountable. Tracking everything will give you visibility into when and why your brand was recommended, or not.

10. Let the Data Decide. When thinking through what products or positioning will give you a leg up in the recommendation algorithms, don’t make an ass out of “u” and “me.” Let the data decide.

11. Brands Can Be Answers. Whether it’s a human or app-ssistant asking the questions, brands are answers too. If you want to be the answer, though, you’ll have to know which questions you can answer.

12. Your Unique Selling Proposition is Critical. You’ll also have to make sure you can answer the questions better than anyone else with your unique selling proposition.

13. Your Competition is Broader Than You Think. App-ssistants don’t think along conventional category lines. People won’t instruct them to fetch the best rental car. They’ll ask them how to get from point A to point B. If you’re in the conventional rental car category, recognize you’ll be competing with everything from public transportation to exercise.

14. You Can Learn a Lot from a Query. If you thought there was a lot to learn from today’s search queries, imagine the rich insights you’ll get from the various types of recommendations people seek or instructions they give their app-ssistants.

15. Sex Sells. While app-ssistants likely won’t respond differently to sexually-charged messages, people always will. If the answer they get doesn’t make them feel good, they’ll give it a thumbs-down.

16. Altruism Sells. Similary, app-ssistants may not be able to tell the difference between good and evil, but their programmers (and users) will. Surely, they’ll tilt the scales in favor of companies behaving like the former.

17. Show Off Your Assets. Also key in this landscape for marketers will be having a bevy of digital assets to address any potential unmet need that a decision or recommendation engine might surface. This goes beyond transactional queries. Create assets that will address instructions for fetching entertainment and/or information.

18. The More Shelf Space, the Better. Syndicate your assets far and wide via API and other formats so app-sisstants can access them on the shelves of review sites and other points of aggregation that will be used to make decisions.

19. Make Your Company a Great Story. No matter how good app-ssistants get at making recommendations, we’ll always turn to our friends and families for their opinions. And the best app-ssistants will incorporate the opinions of friends and family into their recs. Give them a reason to talk about you. Give them a story to tell.

20. Don't Rely on Search Engine Marketing Alone. Don’t obsess over SEM. The future is SAM—Search-and-Act-Marketing. And, remember, the app-ssistant will often operate at the bottom of the funnel. People still need some stimulus to seek out a recommendation or give an instruction.