The booking happened. You never met the buyer.

I want to describe a small moment that used to be so ordinary nobody named it. Someone reads about your hotel, clicks through, lands on your site, and in the twenty seconds before they book, they become a person you know. A session. A cookie. An email address at checkout. A preference. A place in a sequence. You did not do anything to earn that. It came attached to the click.

That moment is what just became optional.

What Google shipped this week

Three things landed in Google's AI Mode travel announcement on August 27, and they are worth separating because they have different consequences.

Flight price tracking, covering more than 300 partner airlines and travel sites, available in over 180 countries. Points and miles redemption rates, with Alaska Airlines, Hawaiian Airlines, American Airlines, Choice Hotels, Hilton and Wyndham at launch and Accor, Flying Blue, Hyatt, LATAM and Lufthansa Group named as coming. And hotel booking completed inside AI Mode, checking out through Google Pay, with ten launch partners. The booking piece is United States and English for now.

Two of those are information features. The third is a checkout. Only the third one changes anything structural, and it is the one that got the least attention, because a checkout inside a chat window sounds like a convenience feature rather than a change in who owns the customer relationship.

Read the last two rows as the same row arriving at different speeds. Reports this week also had ChatGPT Work completing a haircut booking, which Max Weinbach confirmed. A haircut is not a strategy. It is a proof that the plumbing works.

A click was always an introduction

Here is the thing the zero-click argument got slightly wrong for the last two years.

We treated the click as traffic. A number in a dashboard, going up or down, worth defending because more of it was better. That framing made the fight about volume, and it made the response about winning citations so the volume would come back.

But the click was never really traffic. It was an introduction. It was the mechanism by which an anonymous person with an intent became a known person with an identity, and everything downstream of that, your nurture, your retargeting, your lifecycle email, your win-back, your entire owned channel, was built on the assumption that introductions arrive for free as long as you rank.

“Losing clicks costs you traffic. Losing the introduction costs you the ability to ever contact that person again.”

Those are not the same problem and they do not have the same solution. You can lose a lot of traffic and still be fine if the people who matter still arrive. You cannot lose the introduction and make it up on volume.

Why first-party data capture stops being automatic

For most companies, first-party data capture has never been a program. It has been a byproduct. Nobody funded it, nobody owns it, and nobody has a number for it, because it arrived as a side effect of things that were funded for other reasons.

Run the honest audit and it usually looks like this. Most of your list came from organic and paid traffic converting on your own pages. A smaller slice came from events, content downloads and partnerships. A slice you would rather not discuss came from purchased data. And the growth rate of the whole thing tracks your traffic almost exactly, because the first bucket dominates.

Now shrink the top bar by a fifth over three years and watch what happens to everything downstream. Your email program does not fail. It just quietly stops growing, then starts aging, and the day somebody notices is roughly two years after the day it started. This is the same delayed, hard-to-attribute damage we described in the agentic commerce conversion gap, and it is dangerous for exactly that reason: nothing breaks loudly enough to trigger a response.

The B2B version is quieter and already here

If you sell software to enterprises, none of the travel specifics apply to you and the mechanism absolutely does.

Your version does not involve a checkout. It involves a buyer who asks an engine which vendors to consider, gets a shortlist, reads about you inside an answer, spends four weeks forming an opinion you never observed, and then arrives at your demo form with no referrer and no session history. From your side that looks like direct traffic from a stranger who already knows things. The shortlisting happened somewhere you cannot see, and it happened without an introduction.

Rebuilding first-party data capture around outbound

This is the part where I am supposed to tell you outbound fixes it. It does not fix it. It does something narrower and more useful, which is worth being precise about.

Outbound is the only channel where the introduction happens first and everything else follows. You do not wait to be found, you do not depend on a surface staying the same shape, and the identity is the starting condition rather than the reward. In a world where discovery increasingly does not hand you a person, a channel that starts with the person is structurally more valuable than it was, even if its reply rates have not moved an inch.

That is an argument about portfolio construction, not about performance. It is close to the point we made about cold email as a touchpoint rather than a channel, pushed one step further: the touchpoint is now also your most reliable identity source. Four changes follow from that, and they are all boring.

1. Give list growth a number and an owner, separate from traffic and from MQLs
2. Build target account lists top down from your ICP, not bottom up from who visited
3. Treat the reply as the conversion event, because the reply is the identity
4. Report sourced pipeline by account rather than by session or last touch

Number two is the one that changes how a team actually spends its week. Bottom-up list building, where you enrich the people who showed up and work them, has been the default for a decade because showing up was reliable. When showing up gets less reliable, the list has to be built from the market rather than from your logs, which is slower, less satisfying, and no longer optional. Number four is the one your CFO will care about, and it is the reason to do this before somebody asks rather than after.

If outbound is going to carry more weight in the portfolio, it has to be good enough to deserve it, which is a real prerequisite and not a throwaway line. A channel you have been running at half attention does not become load-bearing because a slide says it should. That is the case we make for treating cold email as an owned demand channel with real infrastructure behind it rather than a tactic somebody runs between other jobs.

What I would not do about it

Three overreactions, and I expect to see all of them by Christmas.

Do not pull out of the surfaces where transactions are moving. The ten brands in Google's launch list are not making a mistake by being there. Being present in the place the buyer transacts is correct, the revenue is real, and refusing to participate to protect your data capture is choosing to lose the sale to protect the ability to email someone about the sale you lost. Participate, and be clear-eyed about what participation costs you.

Do not buy your way out of it. The temptation when organic capture shrinks is to replace it with purchased data, which degrades your deliverability, your sender reputation and your relationship with the people you contact, all at once and permanently. The ecommerce version of this mistake is the same shape as the B2B one, and both are usually made by someone under pressure to hit a list growth number they were handed without a method.

And do not build a strategy on a rollout schedule. Hotel booking inside AI Mode is United States and English today, and it is excluded from the European Economic Area, which is part of the wider pattern we mapped out in the guide to a fragmenting search product. The correct response to a capability that is live in one market is to prepare, not to restructure. Preparation here means knowing your capture number and having an outbound function that works. Both of those are worth doing even if none of this accelerates.

You are not behind on this. Almost nobody has a first-party data capture number, and the companies that will look prescient in two years are not the ones with a special insight into agentic commerce. They are the ones who started counting something they had been getting for free, noticed the trend early, and had somewhere else to get it. Start by counting. The rest gets easier once the number exists and somebody has to explain it every month.

See where you are cited today

A free snapshot audit of your rankings and AI citations before we ever talk.