AI shopping is changing how people can discover products. But a recommendation is not a completed purchase. Once a shopper reaches your store, they may still need to check a feature, compare alternatives or find a different size.
That creates a practical question for ecommerce teams: does your on-site experience help someone continue the search they started elsewhere, or force them to start again?
Why the next click deserves attention
In its September 2026 U.S. holiday forecast, Adobe Digital Insights projects a 130% year-over-year increase in AI-driven traffic to retail sites during November and December.
Forecast growth in AI-driven traffic to U.S. retail sites during the 2026 holiday season.
Source: Adobe Digital Insights, September 2026 U.S. Holiday Forecast. This is projected traffic growth, not AI’s share of visits or a forecast increase in sales attributed to AI.
This does not tell you how much AI-referred traffic your own store will receive. It does give you a reason to test the experience after arrival, alongside the work you do to become discoverable.
What happens after an AI shopping recommendation?
Imagine someone researching a waterproof backpack for a 16-inch laptop. An assistant helps them identify useful features and points them towards a product on your store.
They open the page. The colour is not right, the price is above their budget, or the laptop compartment needs checking. Their next action could be a search, a filter selection or a comparison with another product.
A fictional scenario, not a claim about every AI-referred visitor. An assistant’s conversation should not be assumed to carry over to your store.
The important point is continuity. You may not receive the shopper’s original conversation, but you can make it easy for them to confirm requirements and continue exploring.
Four things your store needs to get right
Preparing for AI shopping does not have to begin with another AI feature. First, examine whether the foundations of product discovery support the decisions a shopper still needs to make.
1. Product information that makes claims easy to verify
If a recommendation mentions waterproofing or laptop compatibility, the product page should make those details clear. Do not leave the shopper to infer a technical property from a lifestyle image or a vague description.
For the backpack example, check the stated protection level, compartment dimensions and relevant limitations. Where a detail is unknown, avoid turning an assumption into a product claim.
2. Search that handles both names and needs
A shopper might search for a recommended model by name. They might also try “waterproof laptop backpack” to look for alternatives. Your test set should include both.
Precise names and references need reliable matching. Broader requests may require better synonyms, richer product data or meaning-based search. Choose the response based on observed gaps, not the acquisition channel alone.
3. Filters that make requirements explicit
A page of related products is only a starting point. The shopper still needs a way to narrow the selection by the attributes that matter.
In this scenario, that could mean laptop compatibility, protection level, price and availability. Where those attributes are reliable and relevant, make them understandable and usable as filters.
4. Merchandising that respects the request
A promotion or high-margin product may deserve visibility. However, it should not outrank suitable alternatives if it fails a requirement the shopper has explicitly selected.
Apply hard requirements first. Then decide which suitable products deserve the most visibility.
The same need, two different experiences
Consider what happens when the original recommendation is not quite right. The difference is not simply whether the store displays alternatives. It is whether the shopper can tell which alternatives still meet the need.
Fictional interface examples, not measured results. Compatibility and protection claims require reliable product data. Laptop fit should be checked against actual dimensions, not screen size alone.
Test the journey, not just the landing page
You can start with a small set of realistic scenarios. For each one, define the initial need, the page the shopper reaches and the next decision they must make.
Record what fails, then assign the fix to the right layer. A missing attribute is a data issue. An ignored synonym is a query-understanding issue. A suitable product buried below irrelevant promotions is a ranking issue.
Measure what happens after arrival
Where your analytics can identify AI referrals, treat them as a segment to investigate rather than a reason to assume higher intent or better performance.
Start by checking which pages these visits reach. Then examine the actions that follow: product exploration, internal searches, filter use and progress towards purchase.
These signals need context. A second search may be normal comparison, not failure. Likewise, a difference in conversion between channels does not prove the channel caused it. Review landing pages, devices, product mix and sample size before drawing conclusions.
Do not optimise only for the recommendation. Make the next decision easier.
Frequently asked questions
What is AI-referred shopping traffic?
In this article, it means visits to a retail site that originate from a link in an AI tool. It is distinct from a purchase completed inside an assistant and should not be treated as a measure of sales on its own.
Does an AI recommendation replace on-site search?
A recommendation may point to a particular product, but the shopper may still want alternatives or need to change a requirement. Test whether search and navigation support those next steps on your store.
Do we need a new AI feature to prepare?
Not necessarily. Start by checking product information, known-item searches, filters and ranking. These tests help you identify whether a problem needs a data fix, configuration change or additional capability.
Will our store know what the shopper asked the assistant?
Do not assume that conversation is transferred. Any available context depends on the integration and data shared. Your store should make it easy for shoppers to express their requirements without relying on access to the original conversation.
What should we measure first?
Begin with identifiable referral traffic and landing pages, then examine subsequent searches, product interactions and purchase steps. Use those observations to choose journeys for closer testing, rather than relying on a single channel-level average.
Let’s discuss your customer queries, filters and merchandising priorities to identify where product discovery could work better.
Discuss Your Product Discovery