AI Could Become Advertising's Next Walled Garden
AI assistants are expensive to run, and advertising is a proven way to pay for attention. If ads arrive, marketers will need measurement that sees past them.
Consumer AI has an economics problem that traditional software doesn't have in the same way.
With conventional software, you build the product and maintain the infrastructure, and each additional user can be relatively cheap to serve. Scale has long been one of software's great advantages.
AI changes that. Every conversation takes computation. Every prompt has to be processed and every response generated. The more people use these products, and the more complex the conversations, the more infrastructure they need. Growth brings a bigger bill.
How do you pay for all that attention?
If hundreds of millions of people end up spending a meaningful part of their day inside AI assistants, an obvious business question follows: how do you make money from that attention?
One answer has worked at massive consumer scale before. Search engines and social platforms built products people used constantly, gathered enormous audiences and then built advertising businesses around them. AI platforms could follow the same path.
That doesn't mean every company will. Anthropic, for example, has publicly said Claude will remain ad-free, relying instead on subscriptions and enterprise revenue. Whether every AI company keeps its current business model over the next five or ten years remains to be seen.
But if you plan media years ahead, ignoring AI as a possible advertising channel would be shortsighted. And that raises the harder question: how would you measure it?
AI doesn't leave a click trail
The old digital playbook was built around clicks. Someone searched, clicked an ad, visited the site and maybe converted.
AI doesn't necessarily work that way. Picture someone asking an assistant which running shoes to buy, which software their company should use, where to travel or which financial product fits their needs. A brand could influence that decision without generating any of the clicks marketers are used to.
The person might take the recommendation, keep researching elsewhere, visit a store, search for the brand later or buy through another channel entirely.
Now attribution gets messy:
- The AI platform reports one version of performance.
- Search claims the sale because the person searched the brand.
- Retail media claims it because the purchase happened on a retailer's site.
- Paid social touched the same customer days earlier and claims it too.
We've seen this before
A powerful new platform enters the media mix, builds its own reporting environment and asks advertisers to trust the numbers inside its walls. That's the story of every walled garden so far, and it's why platform-reported results can add up to more than the sales you actually made. (See how this plays out for Meta and search in is Meta getting the credit it deserves?)
That's why it makes sense to think about AI advertising measurement before the budgets become meaningful. You would need measurement that:
- Accounts for impressions and influence, not just clicks
- Doesn't depend on individual-level tracking or last-click rules
- Captures carryover, since an AI recommendation may pay off days later
- Measures cross-channel effects, such as AI exposure driving brand search
- Estimates incrementality: whether the exposure actually changed business results
The good news is that this kind of measurement already exists, and it doesn't need anything from the AI platform beyond spend and delivery. We explain why in AI ads don't break marketing measurement.
The question that matters
Whether AI assistants become ad platforms is an open question. Some will, and some may not.
The more useful question is whether your measurement will be ready if consumer attention moves there. Wherever attention goes, advertising usually follows, and wherever advertising goes, attribution problems aren't far behind.
Frequently asked questions
Will AI assistants show ads?
Some may and some may not. Running AI at scale is expensive, and advertising has funded other mass-audience products like search and social. Anthropic, for example, has publicly said Claude will remain ad-free. Business models can change over time, so it's worth planning for both outcomes.
Why would AI advertising be hard to attribute?
An AI assistant can influence a decision without producing a click. The person may research elsewhere, visit a store or search the brand later. Click-based attribution would give that sale to search, retail media or whichever channel was last.
What is a walled garden in advertising?
A walled garden is a platform that sells ads, controls the data about who saw them and reports its own performance, with limited outside verification. Google, Meta and Amazon are the usual examples.
How should marketers prepare to measure AI advertising?
Use measurement that works from spend, impressions and business outcomes over time and estimates each channel's incremental effect, rather than relying on clicks or the platform's own reporting.