When the Whole Customer Journey Happens Inside the LLM

In 2026, your best customer might never visit your website. They’ll ask an assistant a question, compare options, get a recommendation and complete the transaction inside the same chat window. No search query, no scroll through your lovingly optimized homepage, no app download.

Just: “Book me something,” or “set up a monthly delivery of my dog’s food from Amazon,” and it’s done. That’s not a hypothetical. OpenAI just made waves announcing that it will be introducing ads into ChatGPT, creating an opportunity for brands willing to embrace this new search reality.

CMOs at some of the world’s biggest brands are already wrestling with it in boardrooms: Do they want to be a destination people deliberately visit, or a utility embedded inside large language models (LLMs) and chat interfaces? You can already see the split.

A hospitality giant might insist you get loyalty perks only if you book directly through its app or website. That’s a classic “destination” move: protect the walled garden, force the customer to come to you, and keep tight control over pricing, offers and data. Meanwhile, marketplaces and aggregators are leaning the other way.

For a platform that already lives on top of everyone else’s inventory, becoming a utility inside ChatGPT or another assistant is a natural extension. If you’re Zillow or Etsy, the value proposition is, “Wherever the user is making a decision, we’ll be there.” Both instincts are rational. Both miss the more interesting middle ground.

What’s actually changing is not just where people buy, but where the customer journey happens. Search traffic was already getting squeezed by zero-click answers and search result pages that behave more like “responses” than ranked lists.

Now, LLMs threaten to move even more of the journey “upstream,” into a space where brands have almost no direct visibility. Ask an assistant: “What’s the best credit card for frequent travelers?” or “Find me a family-friendly hotel in Chicago under $300.”

The assistant may summarize reviews, Reddit threads, product specs and pricing data, then present a short list and a default choice. By the time a user hits your site — if they ever do — the real persuasion has already happened. That forces brands to ask a new strategic question: how much of your experience are you willing to expose to general-purpose LLMs?

There are at least four major trade-offs: Competitive exposure: To answer comparison questions, assistants need structured data about your products, availability and terms. The richer that feed, the easier it is to line you up against your competitors. Data leakage: Every interaction is more training data.

The more you let assistants observe your pricing logic, promotions and support flows, the more they can learn the playbook and apply it to everyone else. Control over pricing and offers: If assistants can complete a booking or purchase in-line, you’re trusting another platform to honor your rules around discounts, inventory and upsells.

Ownership of the relationship: Is the customer “yours,” or are you just a brand-shaped slot inside someone else’s interface? Some brands will respond by building higher walls. They’ll allow LLMs to surface basic facts — retail locations, opening hours, top-level product info — but insist that booking, checkout and support all happen on their own properties.

The bet: better margins and stronger loyalty, even if they sacrifice some convenience-driven demand. Others will do the opposite. They’ll treat LLMs as a full-funnel channel and expose deep functionality: browsing inventory, booking, returns, account changes and even loyalty redemptions. For them, the risk isn’t invisibility; it’s commoditization.

Once assistants become a de facto “auction” layer — choosing which brand to recommend or transact with based on opaque relevance signals instead of bid prices — the danger is being swapped out like any other seller in a marketplace. The most interesting work in 2026 will happen between those poles.

Finding the Middle Path A middle-path strategy starts by deciding which intents you’re willing to let live entirely inside assistants, and which ones you deliberately pull back into your own environment. Maybe you allow “find” and “compare” to happen in chat.

Maybe you let assistants handle low-stakes, high-frequency transactions like reordering a staple item, booking a one-night airport hotel or checking a shipment status.

But you reserve high-value relationship moments for your own surfaces: loyalty enrollment, complex bundles, personalized upgrades or interactions where you can showcase your brand’s differentiated experience. That middle path presents a measurement problem.

The hard question for any growth leader is whether LLM-driven activity is genuinely additive or is just rerouting demand you would have captured elsewhere. If discovery happens in chat, and conversion happens in your app, you need a way to stitch that journey together so teams can quantify incremental ROI and de