The Future of Ecommerce Websites in the Age of Agentic AI: What Changes, What Doesn't, and Where Competitive Advantage Will Lie

AI is changing how consumers discover and research products, and much of what gets written about it reaches straight for the biggest claims: the death of the website, the death of Google and the end of marketing as we know it. But is that really where we're heading?

This article focuses on consumer ecommerce, where the impact of agentic AI is being felt by almost all advertisers in some way. Many of the underlying themes will also apply in B2B, where AI is increasingly being used to support activities such as supplier discovery, procurement and RFP creation. However, the buying dynamics are sufficiently different to warrant separate discussion.

Agentic commerce, where AI discovers, researches and ultimately transacts on behalf of customers, may be the long-term ambition of Google, OpenAI and other AI providers. Today, however, those capabilities remain limited and are not yet operating at scale.

Despite that, there is little doubt that AI is already beginning to reshape digital commerce. Across our interviews with senior ecommerce, marketing and platform leaders, there was broad agreement that agentic commerce will become increasingly important over the coming years. Interestingly, most believed the impact would be greater across ecommerce as a whole than within their own sector, suggesting many organisations still see the greatest disruption happening elsewhere.

To understand what is changing in practice, we combined our own testing and client data with those interviews to build a practical view of how customer journeys are evolving today—and where they are likely to change next.

The useful way to think about agentic commerce isn't whether AI can buy products. It's which parts of the customer journey AI is already changing, which parts come next, and what that means for the role your website plays at each stage.

Over the next few years, the technology will continue to advance, but AI won't reshape every part of the customer journey at the same speed. Discovery has already changed, research is changing rapidly and product selection is beginning to shift. Purchase and post-purchase are likely to take considerably longer, constrained as much by customer trust and behaviour as by technological capability.

The short answer is that the website won't disappear, it’s role will change. Value will gradually drain from the transactional middle of the journey and pool at the two ends: being the recommended answer at the top of the funnel and owning the customer relationship at the bottom.

1. Right now... The rules around research and discovery are being re-written by LLMs

As more consumers have become familiar with LLMs, particularly ChatGPT, discovery is shifting away from traditional search.

For marketing professionals, the challenge is measurement. Customers begin their journey on the LLMs, but revert to familiar behaviour at the point it’s time to browse, so little of it is trackable in terms of website analytics. 

Across our client base, AI-referred sessions have grown 187% over the period between January 2024 and June 2026, though from a very small base of under 1% of total sessions.

That understates the shift: many people act on an LLM recommendation before searching for the brand or typing the URL directly. LLM usage grew 2–3x in 2025 alone, to over 800m weekly active users (OpenAI public data).

Shopify's AI-driven traffic is up roughly 8x year-on-year to the end of Q1 2026 and AI-driven orders up 13–14x; OpenAI say about 20% of their near-billion weekly users now arrive with shopping intent, mostly 'help me find' and 'help me compare' queries. Fast growth from a small base, but not yet the whole journey.

Since AI Overviews launched on Google in August 2024 (in the UK), the share of our clients' organic traffic exposed to them has climbed to roughly a third of all organic search queries. AI Overviews answer within the results page, some clients' Google Ads spend has risen 25–45% in response to declining organic traffic.

This is impacting marketers today and many are scrambling to know what to do next. Right now, if an LLM cannot parse your brand and product name, price, inventory position, and customer rating, you may be invisible to it, however good the page looks to a human. More than half of the marketing leaders we interviewed named clean, machine-readable data as a top near-term priority.

Generative Engine Optimisation (GEO) follows many of the same principles as SEO, and many marketing leaders are re-positioning SEO and PR teams to focus on wider GEO as much as search. We covered the 'how' in our LLM marketing series

Where the pre-agentic shopper spent most of their research time on your website, the LLM leans on third parties, such as Reddit, Wikipedia and review sites anywhere your brand is mentioned. It’s now less about link-building and more about reputation-building. The most proactive marketing teams are building deliberate citation strategies on exactly these sources.

A B2C subscription brand told us that Reddit and forums now feature heavily in its GEO optimisation work, and a premium consumer brand leader described how it was more important than ever that they are monitoring their Trustpilot score and sentiment. 

The result of this ‘agentic shortlisting’ means many customers are arriving at sites further through their purchase journey than before. They arrive pre-qualified, having already compared, filtered and largely decided away from your website.

So, what should CMOs be doing today?

Firstly, CMOs need to build the capability to understand and influence LLM-driven discovery.

1. Track presence, not just traffic. Measure how often your brand is cited, appears in shortlists and is described by leading LLMs. Treat it like rank tracking today: monitor it consistently and look for changes over time.

2. Understand where AI Overviews are changing behaviour within search engines. Identify the landing pages and query types losing clicks so that any increase in paid spend is a conscious commercial decision rather than a reaction.

3. Treat GEO as a dedicated capability. Give it an owner, a roadmap and clear objectives. Build structured data, strengthen third-party citations and improve the signals that shape how LLMs understand and recommend your brand.

4. Review your attribution model. Expect more journeys to begin outside channels you can measure directly and avoid judging performance solely through last-click attribution.

2. What comes next? Agents will browse, select and build the basket, but adoption comes category-by-category, not as a single wave

Every brand leader we interviewed agreed on one point: agentic purchasing is coming. The debate wasn't whether it happens, but where it happens first.

Analysts project agentic commerce reaching $4tn by 2030 yet known AI referrals still account for only around 1.5–3% of tracked website sessions across our client base.

AI agents can already research products, compare options and build a basket. What they cannot yet do, at any meaningful scale, is complete the transaction.

A leading LLM told us that this is a deliberate decision to “deprioritise the bottom of the funnel and the checkout experience and stay much more focused on discovery and mid-to-top of funnel”, citing “…the quality bar, the security bar and the cyber risk are all so fraught that it just didn’t feel like the right deployment of our resources.” This doesn’t mean it won’t happen… but it’s not a strategic priority for them right now.

When we asked our interviewees how far it could go over the next three to five years, estimates varied but almost all of the leaders we spoke with felt the impacts would be felt more elsewhere than in their own sector, where they understood the barriers around trust, economics and risk.

One interviewee summed up the mood perfectly: "Agentic commerce is like teenage romance. Everyone talks about it, but almost nobody is actually doing it."

Our interviews kept surfacing the same four hurdles, and these constraints are commercial, legal and, behavioural as much as they are technical.

1. Trust and control are the central issue. Shoppers maybe relaxed about an agent reordering pet food, but far less so about it making judgement calls on expensive products with their money. Two things sit underneath this:

   • Payments are still designed around people. As one senior commercial leader  at a leading LLM put it: "I don't know when people will be comfortable letting an agent use their credit card.

   • Returns introduce uncertainty. Shopify identified this as a major barrier in more considered categories, where customers worry about an agent committing them to a purchase they'll later need to unwind.

2. The economics still must work for merchants. If agent-led checkout costs more than a merchant's existing channels while converting worse, adoption becomes commercially irrational. That's why the first wave of agent-led checkout was withdrawn within months.

3. Nobody yet owns the risk when something goes wrong.  When an agentic transaction goes wrong, no one yet agrees whether the user, the agent developer or the merchant are the liable party. As one global home-furnishings retailer put it, customers already blame the retailer when a partner creates a poor experience. Handing more of the journey to an agent only increases that risk.

   
   

That’s not to say adoption will not happen, (if you’d asked people 15 years ago would people be willing to buy a holiday on their smartphone, the answer would likely have been no...) it just won’t arrive as a single wave.

	High consideration and infrequent purchase	Low-consideration and repeat-purchase
Examples	Engagement rings, cars, furniture, luxury fashion, luxury travel	Coffee pods, pet food, groceries, consumable refills
What the customer gives up	A lot – Control (risk), emotion, the desire to feel the choice on something that matters	Less – The risk is lower if it goes wrong, these are repeat purchases where the act of buying brings little joy
Where the agent will likely play a role	Research and shortlists - the human still completes the purchase	Goes end to end first - discover, select, buy, reorder
What matters most to you as a brand	Share of recommendation - being included and endorsed when the shortlist forms are of paramount importance	Agentic ready - clean feeds, structured data, protocol-readiness - being buyable by an agent at all

So what can a CMO do to prepare? 

• Determine where you sit: Rank your categories and products by how much the customer gives up by delegating (cost, risk, emotion). This may be a mix of products across both ends, so resource this product-by-product rather than as one company-wide posture.

• If your product is low-consideration and repeat-purchase, plan for integration: machine-readable feeds, readiness for protocols like ACP, and a view on whether you participate in agent checkouts or insist on owning your own.

• If your product is high-consideration and infrequent-purchase, win the research phase: how you get recommended, endorsed and shortlisted when the agent does the comparing - the human will still complete the purchasing journey, but only from the agent's shortlist.

3. So, what does this mean for your website? Value may, over time, migrate to the two ends of the funnel

While much of the purchase journey maybe possible for AI, all businesses should now consider how best to approach both the top and bottom of their purchase funnel.

At the top, what you prioritise depends on where a product sits on the scale of how willing your customer is to delegate the decision making to the machine. At the bottom, all businesses should be optimising towards lifetime value: delivering a service or product so good that customers (or agents) want to return time and time again.

Consideration for how to approach the top of an AI commerce funnel	High consideration and infrequent purchase	Low-consideration and repeat-purchase
What you’re competing to be	Recommended - cited, endorsed and shortlisted when the agent compares	Buyable - discoverable, parseable and transact-able by an agent at all
Where to invest	Earned endorsement in the sources models trust - reviews, relevant forums (e.g. Reddit), editorial and expert content — plus brand and reputation signals	Clean feeds, structured and machine-readable product data, protocol-readiness, disciplined GEO
Cost of neglecting	Absent from the LLM-generated shortlist the human buys from	Invisible to the agent - you never enter the consideration set

The obvious point is that many businesses are not solely selling high consideration or low consideration. For example, a business may sell coffee machines as well as coffee pods. This is therefore less about picking a single business-wide strategy but making a deliberate decision, product line by product line on how far you are willing to integrate with the agent layer.

If every agent interaction sees websites in your category the same through data, conversion optimisation is no longer a differentiator. The only place left to feel different is around and after the transaction, and it's what stops the next reorder going back out to an open shortlist the agent can reopen at will.

Whatever AI is able to achieve throughout the purchase funnel, the delivery of service, product quality and experience you ultimately deliver to a customer will not be automated.

On your website, surfaces such as your account, checkout, confirmation and thank-you pages are high-intent, first-party moments that may have historically been neglected in favour of conversion optimisation. But what if these pages are now the place to differentiate and to win customers? 

Whatever your category, the job here is the same: optimise for lifetime value. For low-consideration, repeat-purchase products, that means making the next order effortless, so the customer never needs to reopen the shortlist. For higher consideration purchases, it means an experience good enough that they return to you directly rather than starting again inside an LLM.

What should CMOs focus on right now to prepare? 

• Own the customer relationship entirely: Drive repeat purchases through a first-class customer and product experience that encourages both consumers and agents to return.

  • Audit the journey for differentiation.  If product selection appears identical across your category within an agent, what unique customer experience do they miss after checkout compared to other channels?  What will encourage repeat business?

  • Treat agent-free surfaces like accounts, checkouts, confirmations and thank-you pages as premium real estate.  If your team hasn’t paid much attention to them until now, it’s time to understand their impact on the customer journey and how they contribute to long-term customer value.

• At the top of the funnel, position yourself as the go-to source of unique valuable information that cannot be found elsewhere.

  • As agentic commerce accelerates, how deeply should you integrate?  Focus on feeds and protocol readiness where products can delegate, and endorsement and “share of recommendation” where they can’t.

  • Clearly differentiate why an agent or a human consumer should choose your brand over competitors.

What does this all mean for private equity investors?

The most prepared businesses are treating this as two distinct jobs, focusing on clean data, feeds and endorsement at the top of the funnel and differentiated, loyalty-building experiences post-purchase. The most advanced are framing this at a product category level rather than as one company-wide bet.

• How much of our revenue is structurally exposed? If agents take over more of how customers choose which of our categories are most at risk of being commoditised and what share of revenue and profit sits in them?

• Are we a brand customers will still ask for, or are we becoming interchangeable? As choice moves to agents, do we have the reputation and customer relationships to stay the preferred option, or do we risk being reduced to a low-margin supplier competing on price?

• Do we have an AIO/GEO/LLM strategy, and who owns it? Is this on management's agenda as a genuine strategic priority with a named leader and a plan, or is it still being treated as a marketing project?

• Is our investment proportionate and sustainable? Are the costs we're already absorbing to defend our position a bridge to a stronger model? Are we investing enough, too much, or in the wrong things compared to peers?

• What protects our customer relationship and repeat revenue over the next three to five years? If the way customers buy becomes standardised across our whole category, what keeps them coming back to us rather than starting the choice from scratch each time?

Brands will remain central to consumer choice, but the way they build and defend competitive advantage is likely to change. As AI takes on more of the discovery and research process, the businesses that adapt to those new decision-making mechanisms will be best placed to win.

If you'd like to understand how visible your business is inside AI tools today, or help preparing for a more agentic commerce driven future, get in touch here.

All views expressed in this post are the author’s own and should not be relied upon for any reason.