How AI is changing online customer journeys this holiday shopping season
- Louise Arnold
- 2 days ago
- 3 min read
AI-assisted discovery is changing how customers research products and the paths they take once they reach a retailer’s website.
Evidence from the 2025 holiday season support this shift.
Traffic from AI sources to UK retail websites increased by 329% year on year over the 2025 holiday season (Adobe).
The share of traffic from third-party AI search channels doubled globally during the 2025 holiday season (Salesforce).
Shoppers referred through AI search converted nine times more often than those arriving through social media, suggesting that many are reaching retailers with a clearer idea of what they want and a stronger intention to buy (Salesforce).
This doesn't mean retailers should expect a sudden wave of AI-generated traffic. It does mean AI is adding another route into the customer journey, potentially changing where shoppers enter the site and how quickly they reach purchase.
Instead of starting on a homepage and browsing through categories, more shoppers arrive directly on a product page after comparing products, prices and reviews elsewhere.
They may already know what they want, which variation they need and what they expect to pay.
This could make some customer journeys shorter, more focused and quicker to convert.
Customers may arrive further into the journey
A traditional customer journey might include several stages:
visiting the homepage
browsing categories
using search and filters
comparing products
selecting an item
moving through basket and checkout
An AI-assisted shopper often completes research and comparison before reaching the retailer.
Their journey could begin deeper within the site and move quickly towards:
live stock checks
product options
promotions
basket updates
payment
This does not mean every customer will suddenly behave in the same way. Many will continue to browse, search and compare products on the site. But the overall balance of customer journeys is beginning to change.
The same traffic can create different pressure
The volume of traffic is only part of the picture. A thousand shoppers browsing cached category pages will place very different demands on a platform from a thousand shoppers checking stock, applying promotional codes and attempting to checkout.
More customers arriving ready to purchase, places pressure on the systems supporting:
stock and availability
pricing and promotions
customer accounts
baskets
payments
delivery and fulfilment
Overall traffic could remain within expected levels while the most transactional parts of the platform experience greater demand.
Purchase-ready customers may be less forgiving
Customers arriving through AI-assisted research will also have links to competing products and retailers.
They have potentially completed much of the buying decision before they reach the site.
That makes the experience they encounter especially important.
A slow product page, incorrect stock result or failed checkout could send them straight back to an alternative they have already considered.
What could this mean for peak planning?
It’s not possible to predict every AI-influenced or emerging agentic journey.
You can, however, question whether historic traffic patterns still tell the whole story.
Last year’s visitor volumes remain a useful starting point. But the balance between browsing, direct product entry and purchase-focused activity may need closer consideration.
A useful load test should do more than prove that a website can support a target number of concurrent users.
Load Testing should recreate a realistic mix of customer journeys, traffic proportions and drop-off points.
The aim is not to predict every possible journey. It’s to understand how changing customer behaviour could move pressure to different parts of the platform this holiday season.
Recognising that the journey mix may be changing is only the first step. In Part 2, we look at how peak preparation changes, can help protect the critical paths customers rely on when demand is highest.
