The AI customer journey: how buying decisions are made in the age of AI search

Somewhere right now, someone is buying something you sell. They haven’t visited your website. They haven’t typed anything into Google. They’re having a conversation with ChatGPT, Claude or another answer engine, and by the end of it they’ll know what to buy, who to buy it from, and why.

If your brand came up, you influenced the outcome. If it didn’t, you never existed.

That conversation is the AI customer journey, and it’s where a growing share of buying decisions now get made. This piece is about the customer journey in the age of AI search: what a journey is, how the model has evolved over a century, what changes when research moves inside the engines, and what you can practically do about it.

And to be clear about scope, this isn’t about adding chatbots to your website or AI-personalised emails. It’s about where your buyers now do their research.

What is a customer journey?

A customer journey is the full path a person takes from first realising they have a problem to choosing, buying and using a solution.

Marketers map it as a sequence of stages, typically awareness, consideration and decision, connected by touchpoints: the searches, articles, reviews, conversations and visits where a brand can influence the outcome.

Journey mapping matters because buyers don’t experience your marketing as channels. They experience it as one continuous attempt to solve a problem, and the brands that understand each step of that attempt tend to win it.

A short history of the customer journey

The customer journey has been redrawn roughly once a generation, and every redraw followed the same trigger: the information environment changed.

1898: AIDA. Elias St. Elmo Lewis’s attention, interest, desire, action model became the funnel that still shapes marketing vocabulary today. It assumed brands broadcast and buyers received.

2009: the loop. McKinsey’s consumer decision journey replaced the straight funnel with a loop of consideration, evaluation and post-purchase experience, reflecting a world where buyers could research and compare on their own.

2011: the Zero Moment of Truth. Google gave a name to the research buyers do before ever contacting a brand: the searches, reviews and comparisons that happen between seeing an ad and standing at the shelf.

2020: the messy middle. Google again, describing how buyers loop between exploration and evaluation across dozens of sites and sessions before committing.

Notice the direction of travel. Each model handed more of the journey to the buyer and pushed more of it out of the brand’s sight. AI search is the next step in that same line, and it’s the biggest one yet.

What’s changing: the customer journey in the age of AI search

An AI customer journey is a buying journey that happens substantially inside AI answer engines: ChatGPT, Claude, Perplexity, Gemini and Google AI Overviews. Instead of visiting dozens of touchpoints over days or weeks, the buyer works through the funnel in the turns of a conversation.

It looks like this. Someone types “my commute’s an hour each way and I’m sick of driving, what are my options?” The engine lays out the category, including e-bikes.

They follow up: “what should I look for in a commuter e-bike?” Now the engine is setting their buying criteria. “Which brands are best in the UK?” That’s the shortlist, assembled from a handful of sources.

“Anything I should know about brand X?” That’s scrutiny. “Which would you go with for £2,000?” That’s the decision. Five turns, one sitting, and most of a purchase decision made.

This isn’t marginal behaviour anymore. By 2026, 46% of AI users start their purchase research on a standalone AI platform such as ChatGPT, Claude, Perplexity or Gemini, up from 25% in 2024 (L.E.K. Consulting).

Prophet’s 2026 consumer study found 73% of consumers now use generative AI, and pre-purchase research is its single most common use, cited by 61%.

Here’s the shift in one view:

Why it matters: five implications for brands

The criteria are set before you arrive. Early in the journey, the engine teaches the buyer what to care about. If those criteria match your strengths, every later turn tilts your way. If a competitor’s content shaped them, you’re playing uphill in a game you didn’t know had started.

Shortlists come from gatekeeper sources. When an engine names “the best providers”, it leans on a small set of directories, review sites and guides. Brands absent from those sources are absent from the shortlist, however good their own website is.

Scrutiny happens with nobody in the room. “Is brand X actually any good?” gets asked and answered without a salesperson, a landing page or a rebuttal. Whatever the web says about you, accurate or stale, is what gets said.

Your analytics can’t see it. There’s no session to track and no click to attribute. Nearly a third of AI users (31%) say their purchase decision was largely made before they reached a brand or retailer website, up from 26% two years earlier (L.E.K. Consulting). The journey didn’t get shorter. It got invisible.

The moment of truth has moved. It used to be the shelf, then the search results page. Now it’s the shortlist turn of a conversation, and most brands have never seen theirs.

What hasn’t changed

It’s worth being honest here, because the doom framing oversells it. Buyers still move through the same human sequence: notice a problem, learn the category, compare options, stress-test the favourite, commit. AI compresses the journey; it doesn’t rewrite the buyer.

The fundamentals still decide the outcome, too. Engines synthesise what the web already says about you, so being genuinely good, accurately described and well regarded in the places your category trusts still carries the day.

And buyers haven’t switched their brains off: 86% of US online shoppers who used AI for product research verified the recommendation somewhere else before buying (Product.ai, 2026). Your website, your reviews and your reputation still matter. What’s changed is the order in which they’re encountered, and who does the introducing.

How to see your own AI customer journey

You can get a useful first look yourself in an afternoon:

1. Write down five questions your buyers actually ask. Their words, not your keywords. Your sales team knows these by heart.
2. Ask them in fresh sessions across ChatGPT, Claude, Perplexity, Gemini and Google AI Overviews.
3. Follow up like a real buyer would. “Who are the best providers?” “What about [your brand]?” “Which would you choose?”
4. Note three things per answer: whether you appear, how you’re described, and which sources are cited.
5. Repeat on another day. Answers vary between sessions, so patterns matter more than any single response.

You’ll learn something from that exercise, and sometimes it’s uncomfortable. Its limits are real, though: one person’s questions, a handful of runs, no way to tell an anecdote from a pattern.

Measuring the journey properly means repeated runs, multiple buyer personas and every answer coded, which is the work of Tilio’s AI Customer Journey Analysis. We’re the first UK agency to analyse full customer journeys in AI search, not single prompts, and you can scope one for your category in about two minutes with the builder on that page.

Where the AI customer journey is heading

Three developments are worth planning for now.

From advising to acting. Assistants are becoming agents: browsing, comparing, filling baskets and booking calls on the buyer’s behalf. When the AI executes the shortlist rather than just suggesting it, being machine-legible stops being a marketing advantage and becomes a distribution requirement.

From generic to personal. Engines increasingly carry memory and context, so two buyers asking the same question will take different journeys. The typical journey becomes a distribution of journeys, which makes measuring patterns matter even more.

From visible to assumed. As answers become the interface, fewer journeys will leave any trace at all. The brands that map their journeys now, fix their drop-out points and keep re-measuring will compound an advantage the latecomers can’t see forming.

None of this requires panic. It requires the same discipline journey mapping has always required, pointed at a new place: find out what your buyers are actually experiencing, and close the gap between that and what you’d want them to hear.