Your customer is disappearing behind AI. Zero-party data won’t bring them back

For twenty years, the big promise of digital marketing was a direct relationship with your customer. Know who they are, see what they do, talk to them one to one. Whole DTC categories were built on it.

That relationship is quietly disappearing. And what it’s disappearing behind is an AI assistant.

The customer you can’t see

More and more buying journeys now start, and sometimes nearly finish, inside a conversation you’re not part of. Someone asks ChatGPT, Gemini or Perplexity which sunscreen to buy. The assistant compares, shortlists and explains. By the time the customer reaches your website, if they reach it at all, the decision is mostly made.

You can see the shift even in ordinary search. SparkToro’s 2026 research, built on Similarweb clickstream data, found that 68% of US Google searches in the first four months of this year ended without a single click. The share of searches sending any click at all has fallen by almost a quarter since 2024, and when an AI Overview appears, click-through rates drop by nearly 60%. Inside a chat assistant, there often isn’t a results page at all.

No pixel fires inside that conversation. No cookie, no session, no email capture. The question they asked, the brands they compared and the reason you didn’t make the shortlist all happen out of sight.

The fight over who owns the customer

We’ve already seen how much this matters to brands. In September 2025, OpenAI launched Instant Checkout, letting people buy from merchants without ever leaving ChatGPT. By March 2026 it had been scaled back, and purchases now send people to the merchant’s own site. One widely reported reason: brands didn’t want to become anonymous fulfillment centers.

Brands won that round. They kept the checkout. But the comparison, the shortlist and the moment of choice still happen inside the assistant. You can keep the sale and still lose sight of the customer.

Retail brands know this feeling well. The customer at the Target or Ulta till has always belonged to the retailer. AI is extending that to everyone. It’s turning every brand into a wholesaler, with someone else owning the moment of choice.

The industry’s answer: zero-party data

The marketing industry’s answer to losing sight of the customer has been to ask them directly. Zero-party data is Forrester’s term for information a customer intentionally and proactively shares with a brand: quiz answers, preference centers, onboarding questions, loyalty profiles.

It has real strengths. It’s built on consent, it’s explicit, and it’s yours. For personalization and CRM, it’s genuinely valuable.

But as an answer to the disappearing customer, it has three problems.

1. It only reaches people who’ve already found you. Zero-party data needs a touchpoint: a site visit, an app, an email sign-up. The customer who chose a competitor inside a chat never got as far as your quiz.
2. It’s what people say, not what they do. Declared preferences are attitude, and stated intentions are a notoriously poor guide to what people actually buy.
3. It’s loyal-skewed and blind to competitors. The people who fill in your preference center are your most engaged customers. Your competitors’ customers never will.

Put bluntly, zero-party data is a survey with better branding.

First-party data has the same limit. It’s a mirror: it shows you the customers you already have. As Byron Sharp’s work shows, brands grow by winning more buyers, mostly light buyers who barely think about you. They’re precisely the people least likely to be sitting in your CRM, and the most likely to be asking an AI assistant instead.

Attribution was already struggling

None of this is happening to a healthy measurement system. Since 2021, iPhone apps have had to ask before tracking people across other apps, and a large share of people say no.

Add journeys that now happen inside AI conversations, and user-level attribution ends up measuring a shrinking, unrepresentative slice of buyers: the ones who still happen to be trackable. That isn’t measurement. It’s looking where the light happens to be.

What you can still see

Here’s the good news. You may not be able to see the individual customer, but you can still see the market.

Even a customer who’s invisible to you leaves traces in aggregate: what scans at the till, what people search for, which sites they visit, what gets said about brands in your category, what your competitors actually spend, and how often AI assistants recommend each of you.

Measured across a whole category, that trail tells you more about where your brand stands than user-level tracking ever did. Three things make it work.

1. All the signals. Paid, Owned, Earned and Search, including LLM visibility. Not just the channels that happen to be trackable.
2. The short and the long term together. The activation that lands this week, and the brand building that pays out over months and years. (More on that in why you need both.)
3. Your competitors on the same lens. Share is relative. Market-level behavioral data is the one view where every brand in the category shows up like-for-like, because nobody has to opt in for their sales to scan.

What this means for your team

• Accept you won’t see every customer. Rebuilding individual tracking in a world designed to prevent it is an expensive way to measure a shrinking minority.
• Use zero- and first-party data for the relationship. Personalization, service, loyalty. Not for judging whether your brand is winning.
• Measure the market for decisions. Momentum against your competitors, tied to revenue.
• Put AI visibility on the scorecard. If an assistant shortlists three brands in your category and you’re not one of them, nothing in your own data will ever tell you.

You may never see every customer again. You can still see the market. Measure that.