A quiet shift is turning into a full-blown reroute of the customer journey: AI assistants are increasingly mediating how people discover brands, get answers, and complete tasks.
In a February 2026 briefing to CX Dive, experts warned that many organisations have built their digital experiences for an internet shaped by traditional search – and that playbook is now being rewritten by assistants embedded across apps, browsers and operating systems.
The “tap-and-swipe” era may be ending faster than expected
The provocation that’s making digital teams sit up is less about a new channel and more about a new interface.
The president of TELUS Digital Solutions, told CX Dive that they expect the familiar model of opening an app and tapping, swiping, or typing to fade rapidly – and summed up the alternative in one line: “You’re just going to want to use your voice.”
If that prediction lands, brands won’t just be redesigning screens; they’ll be redesigning how customers arrive in the first place.
AI assistants are becoming a distribution channel in their own right
The numbers underpinning the urgency are starting to stack up. A Semrush study argues that, for digital marketing and SEO topics, AI-search-driven visits could overtake traditional search visits by early 2028 – a signal of where attention may flow as interfaces change.
Meanwhile, Ahrefs analysed traffic from 3,000 sites and found that 63% received at least one visit from an AI source (in other words, “nearly two-thirds”), suggesting AI referrals are already touching a large slice of the web – even if they’re still a small proportion of total traffic for many sites.
The land-grab is already happening inside the assistants
As tech companies push assistants into everyday workflows, brands are beginning to experiment with being “native” where the questions happen – not just where their owned channels live.
The CX Dive reporting points to familiar consumer names such as Expedia, Spotify and Zillow exploring experiences built inside ChatGPT, hinting at a future where the assistant becomes a new storefront – and, in some cases, a new checkout lane.
From SEO to “agentic” optimisation
This is where the new vocabulary enters the room: agentic engine optimisation, answer engine optimisation, and generative engine optimisation.
The concept is straightforward – make your app and site discoverable and usable by AI systems – but the execution is more demanding than classic SEO because outputs are less predictable and less purely keyword-driven.
It’s an evolution, but also a break: if your “reader” is a machine acting on a customer’s behalf, the old rules of layout, copy, and navigation can become friction.
The biggest strategic mistake: pretending you control the whole journey
One of the sharpest warnings in the CX Dive piece comes from Zendesk. Jason Maynard says the “biggest mistake” he sees is teams “putting the blinders on”- because more of the digital experience will happen in an ecosystem brands don’t control.
In practice, that means optimising solely for owned channels and classic search visibility can become a form of self-imposed tunnel vision.
Why today’s websites can be painfully awkward for AI agents
Most sites are built to be seen by humans – and interpreted by Google’s search systems – not navigated by software agents trying to complete tasks.
The result is a clunky workaround: AI agents may resort to taking screenshots and working out where to click via X/Y coordinates, a resource-intensive approach that can be slow, incomplete and error-prone.
That latency doesn’t just frustrate the agent; it risks degrading the end customer’s experience, too.
Visibility becomes “winner-takes-most” – again
Anyone who remembers the scramble through AOL’s walled garden, then the rise of search engines, then Google’s dominance will recognise the rhythm: a new gateway appears, and brands compete to be legible inside it.
Dengel describes today as an “early, fragmented stage” – meaning experimentation across platforms matters now, before habits solidify. And he’s blunt about the stakes: “If you’re not the top three on Gemini or Claude or ChatGPT, you might see a huge fall-off in traffic.”
The first fix is boring – and unavoidable: “digital hygiene”
Before brands chase shiny integrations, it’s been argued that they should start with basics that make websites easier for AI agents to crawl and understand.
That includes structuring pages so hierarchy is obvious (headings, paragraphs, images, links, forms), ensuring HTML tags are correct, and improving metadata – especially clear labelling and descriptive alt text.
Even visual handling shows up on the checklist: “vectorising images” is cited as one way to help assets display consistently across platforms and machine-driven contexts.
AI.txt, semantic HTML, and APIs: making your site “machine-readable” on purpose
The operational shift goes beyond tidy markup. The CX Dive report says brands may create AI.txt files – positioned as a plain-text control surface similar in spirit to robots.txt – to guide how AI models and agents interact with web content.
It also highlights semantic HTML as a way to give systems richer context about what content is, not just what it looks like. And then there’s the strongest move toward machine-first design: open APIs, which are entirely machine-readable and bypass the fragile “human interface” layer altogether.
As Dengel puts it, the goal is to “direct the AI to the conclusions that you want” using your own data and information.
The real objective: get discovered there – then guide customers here
In the near term, many brands won’t be trying to “replace” their apps and websites – they’ll be trying to make AI-mediated discovery feed them.
A Professor at the University of Illinois Gies College of Business notes that an initial goal is working out how AI-driven search can steer customers into “proprietary channels” like your site and app.
The upside isn’t only traffic: done well, this kind of optimisation can also improve consistency of how your brand is represented across multiple assistant-led platforms, reinforcing value even when the interaction happens somewhere you don’t own.
A messy middle is likely – but doing nothing is still a decision
Even the advocates for optimisation aren’t claiming a clean overnight replacement of the web. It is expected to have a messy middle because the infrastructure isn’t fully there yet – meaning humans and machines will keep sharing the same interfaces for a while.
The difference is that brands that treat this as future work risk waking up to a present where assistants are the default starting point, and their digital properties are hard to read, hard to act on, and increasingly easy to bypass.
Conclusion: rebuild for the reader you didn’t design for
The takeaway from the February 2026 CX Dive reporting is stark: if AI assistants are becoming the first point of contact, brands must make their sites and apps not just attractive to humans, but usable by machines.
That means shifting from search-only thinking to agent-ready thinking – tightening digital hygiene, investing in semantic structure, considering AI.txt-style controls, and exploring APIs where it makes sense.
The destination is still being negotiated by consumer preference, and the route will likely stay messy for some time.
But the direction of travel is clearer: the brands that remain discoverable and actionable inside AI-led journeys will keep showing up – and the ones that don’t may find themselves digitally “present” yet practically invisible.
