We tested AI shopping agents from Amazon and Perplexity to see how they fare helping prepare a family for Iceland’s famously fickle Laugavegur Trek. For now, highly capable LLMs have the clear upper hand.
By David Golub and Francois Silvain
Undertaking a multiday hiking trip in Iceland isn’t for the faint of heart – or the poorly packed.
Making sure to have the right gear for any family vacation can involve quite a few stressful buying decisions, all the more so when headed to a remote region well-known for frequent, intense shifts in weather conditions.
We wondered: Can AI shopping agents help a family get ready for such an adventure?
To set up a meaningful test, we chose a challenging but approachable destination: the Laugavegur Trek in the Icelandic Highlands, a multiday hike through rain, snow, river crossings and barren volcanic landscapes.
Personal shopping agents promise to reduce decision fatigue and streamline online shopping. But to evolve beyond just smarter search, such agents need to demonstrate context awareness and support reliable, trustworthy buying decisions.
Our Laugavegur test case through the land of fire and ice allowed us to ask three key questions reflecting these adoption benchmarks:
- Can shopping agents build a reliable packing list?
- Can the agents make good product recommendations?
- Can we purchase at least 1 item with an agent?
As two outdoor enthusiasts with decades of hiking experience, we wanted to know if these agents could save time, reduce anxiety or introduce us to new and better equipment, or even suggest gear we hadn’t already considered.
With the summer season coming up, we put Amazon’s Rufus and Perplexity Pro to the test, with detailed summaries below. Our quick takeaway?
Highly capable LLMs (large language models) such as Perplexity, have the clear upper hand over bots like Rufus, at least for the moment, in providing expert-level guidance for complex purchasing, even if Perplexity’s buy flow remains a tad glitchy.
Prepping for Iceland Sets a High Bar
Iceland’s Laugavegur Trek cuts a stunning 35-mile path through Iceland’s Highlands, from the geothermal springs of Landmannalaugar to the lush Þórsmörk valley. For many, it’s a bucket-list adventure of a lifetime.
While breathtaking, this trail is notorious for volatile weather — hikers face cold rain, sudden sunshine, snow crossings, and powerful winds, sometimes all within hours, even in July. River crossings add another challenge as glacial meltwater levels can rise rapidly.
This unpredictability creates the perfect testing ground for AI shopping agents. Unlike routine purchases, gear selection for Laugavegur demands expert-level knowledge about layering for warmth and waterproofing to stay dry.
As experienced hikers and AI professionals, we were curious how our accumulated trail miles might stack up against AI recommendations.
Can these systems simplify the shopping process and make an Icelandic adventure more accessible to families?
We were eager to find out.
Let’s Meet the Shopping Agents
For our Icelandic hiking gear experiment, we selected two leading AI shopping assistants that represent different approaches to help consumers make purchase decisions.
Amazon Rufus
Amazon Rufus launched in early 2024 as Amazon’s in-house AI shopping assistant, fully integrated within the Amazon shopping experience.
Rufus is designed to answer product-related questions, provide recommendations and facilitate product discovery within Amazon’s vast ecosystem. It leverages Amazon’s extensive catalog, customer reviews and Q&A content to help shoppers make informed decisions.
What makes Rufus interesting is its deep integration with Amazon’s purchasing workflow. Buyers can journey from curiosity to checkout without leaving the platform.
Perplexity Pro
Perplexity Pro offers a contrasting approach, deploying Perplexity’s AI-powered search engine across multiple retailers, including Amazon, Best Buy, Walmart and Shopify merchants, in addition to accessing the common crawl internet content.
Pro subscribers ($20/month) can access features like “Buy with Pro” for one-click checkout and “Snap to Shop” for visual product search. This unfolds in a familiar LLM chatbot experience enhanced with a shopping cart workflow.
Unlike Rufus, Perplexity positions itself as platform-agnostic, prioritizing unbiased recommendations over retailer loyalty.
Why We Chose these Agents
We landed on these two assistants because they are well-known and represent fundamentally different UX patterns in AI shopping: Amazon’s closed-ecosystem approach versus Perplexity’s cross-platform strategy.
We wondered if our challenge would be better served by deep integration with a single retailer or an agent accessing multiple stores. For outdoor gear specifically, where specialty brands often offer the best choices, we suspected this distinction could prove meaningful.
Would Rufus’s deep knowledge of Amazon’s catalog and customer feedback provide more nuanced recommendations? Or would Perplexity’s ability to search across multiple retailers, including outdoor specialty shops, ultimately deliver better gear suggestions?
As we discovered, the two share opposite ends of an overlapping interface strategy: Rufus is a chatbot that directs sales into the Amazon ecosystem while Perplexity is a search-LLM that shifts sales into an e-commerce side-panel.
The contrast would help us understand which approach creates more value for consumers preparing for specialized activities like wilderness hiking.
Phase 1: Building the Packing List
For both agents, we started the conversation with a basic prompt that slightly simplified the task at hand: “Help me create a detailed packing list for a family of 4 hiking the Laugavegur Trek in July. We’re staying in huts and won’t need camping gear.”
How did Rufus fare? Not very well.
Despite several attempts with the initial prompt and several variations, Rufus was not able to create anything resembling a packing list. While it did provide links to appropriate product categories (thermal layers, hiking poles and backpacks, etc.), it failed with the big picture.
Here’s the response to the first prompt:
“I don’t have access to detailed hiking resources, but I can help you pack for your family’s Laugavegur trek.”
Rufus then provided links to relevant items but was unable to sustain a dialog around our original question, which we attempted re-stating multiple times in more general terms to tie it back to the specific context, for example:
I am an experienced hiker planning a 4-day trek of Iceland’s Laugavegur Trek. I’ll be staying in huts along the trail and not camping. Can you help me make a list of all the gear I need?
Rufus seemed to understand the question this time, and replied encouragingly, with seeming contextual awareness of the trek’s available luggage service (handy baggage transport between huts) as well with an acceptable recommendation for a daypack:
I can help you pack for your Laugavegur trek. Will you be carrying a backpack or using a luggage transfer service? Knowing this will help me recommend the right gear. For example, “I’ll be carrying a 40L backpack.”
Yet when I replied to the direct question and told Rufus that we would indeed be using the luggage transfer service, we reached a dead end in the conversation.
I don’t have an answer for that. Do you want to try another shopping question? You can also head over to the Customer Service Home page for more information.
Rufus responded to repeated requests for help creating a packing list by presenting a collection of coherent product recommendations, but failed the test of helping us understand the totality of the gear we needed, and why.
Under the hood, we know that Rufus depends heavily on product information from product sellers and struggles with concepts and attributes that are not directly related to that data. This turns out to be a curious limitation on context-awareness.
No Doubts About Perplexity’s List-Building
Turning to Perplexity, we began again with the same straightforward prompt that flummoxed Rufus, receiving almost instantly a well-organized, thoughtful Laugavegur trek packing list that entirely exceeded our expectations.
“Help me create a detailed packing list for a family of 4 hiking the Laugavegur Trek in July. We’re staying in huts and won’t need camping gear.”
Indeed, given Perplexity’s ability to search the internet, we were not surprised by the curation of this extensive list, which was coupled with numerous links to expert blogs, travel agencies and personal travel logs.
We were impressed how good it was, however.
Perplexity’s list was comprehensive and well-organized, smartly balancing preparedness, weather adaptability and weight efficiency, all key variables for an extended back-country adventure. The list was also heavily referenced with expert sources, allowing the shopper to check all recommendations.
Some things we liked a lot:
- Layering Strategy: Excellent coverage of base, mid and outer layers to handle cold, rain and wind.
- Weather Preparedness: Includes waterproof gear, insulated layers and river crossing footwear.
- Hut Comfort: Smart additions like a sleep mask, earplugs and hut clothes enhance rest and comfort.
- Trekking Poles for river crossings: Many people regret not bringing these along.
- Family-Focused: Tips for involving kids and extra snacks show foresight.
Perplexity missed only a handful of items experience has taught us to have on hand, including nail clippers, a flask of evening beverage, a small repair kit (always including duct tape), 100 kr coins for showers and a notebook and pen.
In a separate test, we constructed a highly detailed prompt, asking for the list, product recommendations in each category, and then recommendations at different price points. This list was also very impressive and well-annotated.
One of our assumptions in choosing this well-known journey through the Icelandic Highlands was that LLM-powered packing list creation would be turbocharged by the voluminous available content on this topic.
On the list-building task, Perplexity delivered.
Phase 2: Product Recommendations
Moving on to product recommendations, we wondered how well the agents could handle the tricky process of purchasing outdoor gear, where myriad tradeoffs between price, quality, features and brand reputation can easily confound an experienced shopper.
As before we started with Rufus, then Perplexity.
Rufus knows he’s not an expert.
Moving beyond struggles building a packing list, Rufus did an overall good job making product recommendations and hand-offs to Amazon’s platform, while continuing to display a lack of contextual knowledge or user nuance to guide decision-making.
For example, we asked Rufus if we would need trekking poles for this hike, receiving a lengthy reply about the benefits of poles – such as stability and energy conservation – all without mentioning the most important contextual reason to have them for Laugavegur, namely, numerous stream crossings (as noted above).
Hiking poles can be very beneficial for a 4-day trek, providing stability, support, and reducing strain on your legs and joints…
Yet if Rufus seemed unaware of the reasons for packing poles for this Iceland trip, the agent nevertheless succeeded in putting together a detailed product page of credible recommendations, with filters such as price, lock mechanisms and brand.
Since we’d be using our poles again in the future, we searched in the $150-$250 range, and found a good selection of options from top brands such as Black Diamond, LEKI and MSR, with choices of size and material, and of course, delivery speed.
We gave Rufus another challenge with choosing a rain shell, possibly our most important decision after footwear. The responses along this line of inquiry revealed some clear limits in this agent’s ability to provide expert advice.
Thank you for the great suggestions for trekking poles. Let’s move on to the rain shell decision. What do you recommend for a 4-day hike in the Icelandic highlands?
In response, Rufus told us what he doesn’t know:
I don’t have access to expert resources like Wirecutter or Outside Magazine, but here are some potential rain jacket options for your 4-day Icelandic hiking trip:
With this, Rufus provided a set of links, including low-cost, no-name brands as well as mid- and high-end choices from reputable manufacturers such as Outdoor Research, The North Face and even a pricey top-tier choice from Arc’teryx.
Dumped into Amazon’s vast product catalog, the buyer’s anxiety about this make-or-break decision would hardly be softened amid pages and pages of choices.
In the end, Rufus let us down.
Perplexity shines with recommendations.
Having absolutely crushed the list-building task, we were excited to see how Perplexity would handle recommendations. We continued to be impressed.
The segue to recommendations happened without us needing to ask, in a “People also ask” section, depicting a set of related or follow-up topics Perplexity adds at the end of a response.
Perplexity’s suggestions for related questions for our Laugavegur shopping exercise were contextually aware and on-point, the added content was spot-on and demonstrated deep contextual awareness as well as the nuance of our initial prompt.
For example, tapping on “Which waterproof hiking jacket offers the best value for money?” elicited a detailed comparison table of 12 leading outdoor brands that included a brief summary and multiple links to gear review sites and blogs.
The table was backed up with further detail on leading brands and an expert recommendation for the Mountain Equipment Makalu, which earned a 2024 Gear of the Year award from Trail Magazine. Nice work, guys!
Curious to push Perplexity on the topic, and mindful of the Makalu’s non-trivial price tag, we ask Perplexity if the recommendation was overkill for our Laugevegur test case. The response was emphatic: no.
The Makalu is not overkill for the Laugavegur Trek in July – it is a highly suitable choice, especially if you value maximum protection and peace of mind in unpredictable, wet, and windy conditions.
Satisfied we had the upper range in sight, we put Perplexity on a budget and asked for the best choice under $250 – which qualifies as reasonable in this pricey category.
Again, we were treated to a detailed comparison table of 6 options as well as a clear winner, Patagonia’s Patagonia Torrentshell 3L, a quality option from a highly trusted brand within our price range.
We also liked that this recommendation was backed up with a link to a Reddit discussion specifically focused on choosing a rain shell for Iceland hiking! Talk about social proof!
We continued to dig into recommendations for rain pants (Marmot or Outdoor Research), hiking socks (Darn Tough merino wool) and an inflatable travel pillow (Nemo Fillo), impressed by Perplexity’s thoughtful rankings.
While Perplexity’s deep research may be more suited to gearheads than casual hikers, the experience provided a robust deep-dive that consistently referenced our context during the testing.
We trusted that Perplexity had our back out on the trail.
Phase 3: Making a Purchase
Our final test of the agents was completing a purchase. For a family undertaking such a “walking holiday,” the time needed to do all the needed shopping would be considerable, so any added convenience would be welcome.
Following our pattern, first Rufus and then Perplexity.
Rufus Makes a Swift Handoff
When it came time to make a purchase via Rufus, we were hardly surprised by the ease of the experience: this chatbot sits atop the formidable Amazon buying engine, with its host of optimizations for purchase and delivery.
The journey is actually more of a handoff, in fact, with Rufus leading the shopping into Amazon’s well-oiled workflow of aggregation pages, detail pages and simplified buying experience. That is also to say that Rufus disappears from view, curiously missing the chance to continue our buying conversation and lead me to more products.
Rufus sits on top of Amazon’s powerful e-commerce backbone, but struggles with providing complex contextual product recommendations.
We chose to make a simple purchase with Rufus: a hiking map for the trek, a critical, need-to-have even with such a popular, well-blazed route.
Not surprisingly, Rufus led us to two popular, well-known choices for maps: National Geographic’s Iceland Map, as well as the definitive Trekking Map: Iceland’s Laugavegur Trail & Fimmvörðuháls Trail, which we already knew to be go-to resources.
And yet, interestingly, when asked if it is a good idea to carry a hiking map on the trail (which should have been an obvious “yes”), Rufus came back with a response which hinted at a key limitation in the bot’s reliance on Amazon’s treasure trove of customer reviews.
Unfortunately I am not able to access website links or external information. However, based on the product details, customers find the map extremely useful for planning and navigating the Laugavegur Trek, noting its detailed information on trails, huts, campsites, and terrain features.
To put it mildly, Rufus’s failure to insist that we take a map with us into the Icelandic Highlands hardly increased our confidence using Amazon for our travel preparations.
That said, we bought the map, delighted by free shipping and next-day delivery, and said goodbye to Rufus for the moment, turning our attention to Amazon’s recommendations for additional purchase.
Those included a bevy of related maps and guidebooks, but notably, no compass.
Perplexity Almost Lets You Buy Like a Pro
When it came time to test the buying experience on Perplexity, we continued to be impressed, even if we found a small hiccup in the workflow.
Unlike Rufus, we found Perplexity’s transition to the shopping workflow is less than obvious. Some users might not instantly recognize the “Buy with Pro” button adjacent to the product card, which looks more like a chunk of branding and less like a big, obvious buy now button.
We found that prompting Perplexity to initiate an e-commerce transaction also worked, a somewhat disjoined user journey that could be improved with clearer interface cues. Where Rufus makes a swift handoff into the buy flow, Perplexity is overshy.
That said, once we figured out the trick, the rest was a snap.
Perplexity Buy with Pro (available with subscription) excels at expert-level, contextually aware product recommendations but the buy flow is a tad glitchy.
We chose to test the experience with the purchase of an inflatable camping pillow – a handy accessory that straddles the line between must- and nice-to-have.
Our conversation with the LLM identified 5 options for the travel pillow, ranging from budget, airplane-style U-shaped neck supports to a full king-size luxe model. We chose a mid-range option, which was shown with several clear product shots and a reassuring 4.7 stars.
To initiate the purchase, we prompted: “Can you help me buy the NEMO Fillo?”
Perplexity responded by opening its “Buy with Pro” workflow, prefacing the next steps with additional content reassuring us that we’d made a good choice.
The NEMO Fillo Pillow is a top pick for hikers and travelers seeking comfort, support, and portability. Here are the best options for purchasing the NEMO Fillo Pillow, along with a brief overview of its features and why it stands out for your needs.
Tapping the button opens a side-panel workflow (which requires adding some shipping and payment details) and is then seamlessly resolved with a few additional clicks.
In making our decision, we were further reassured to see a selection of reputable retailers fulfilling the product, including REI (a favorite outdoor retailer) and Dick’s Sporting Goods.
Yet while seeing these names increased trust, we didn’t see any way to enter member or loyalty numbers, which would likely lead us to bypass “Buy with Pro” for any sizable purchases, so as not to lose out on those valuable points. A planned integration with Firmly will hopefully solve that problem.
That said, Perplexity is, at least for now, offering free shipping, which is always nice to see before tapping that final “buy now” button.
AI Shopping Assistants in the Wild
Our test comparing Amazon Rufus and Perplexity Pro revealed sharp differences in how AI shopping agents handle specialized purchasing decisions.
Using Iceland’s challenging Laugavegur Trek as our test case, we evaluated each system on its ability to build comprehensive packing lists, make informed product recommendations and facilitate seamless purchases.
We’ve summarized a few key takeaways:
While Rufus struggled with contextual understanding and failed to create a complete packing list, it did provide decent product recommendations within Amazon’s ecosystem. For one-off purchases, Rufus enhances the already near-frictionless Amazon experience.
Perplexity Pro, however, demonstrated far superior contextual awareness, delivering an expertly curated packing list and nuanced gear recommendations built atop verifiable expert knowledge and backed by a flawless (once configured) checkout experience.
While we are both experienced hikers and perhaps would more likely indulge in-depth gear review of niche products, we felt that an LLM-driven experience would also be better for novice hikers. A simple adjustment of the prompt can tune the conversation to the expertise level.
In the end, Perplexity easily won the day with the first two parts of our testing, making the $20 monthly upgrade feel acceptable, at least when you need shopping lists and expert-level recommendations for complex purchases.
The final category was a tie: when you know what you want, Amazon remains an unrivalled channel for get-it-tomorrow shopping. We will be curious to see how Amazon improves Rufus’s handling of context, which our test found unsatisfactory.
Where next for AI shopping agents?
Our experience testing Rufus and Perplexity Buy with Pro points to an evolutionary path from glorified search engines to trusted advisors for complex purchase decisions. Ultimately, we think users will expect agents to do their shopping for them.
We’re not there yet, but we do see clear milestones along that journey.
- First, we expect increasing specialization, agents with deep expertise in domains such as outdoor activities, home renovation or cooking, with rich mental models based on activity profiles as well as purchase history.
- Second, hybrid systems that combine the strengths of both approaches seem inevitable. Imagine Rufus’s Amazon integration with Perplexity’s contextual understanding and multi-retailer reach.
- Third, shopping agents will need to develop stronger reasoning about tradeoffs – price against quality or reputation against innovation – while handling the diversity of consumer preferences.
- Fourth, we expect to see agent-to-agent conversations become much more common, where a generalist can interact with a specialist when a deep dive is needed in an unfamiliar category.
- Finally, robust user experience will require more persistent memory and better interfaces so that users can manage long shopping lists, working step-by-step to complete all necessary purchases.
For complex tasks such as preparing for hiking in Iceland, we’re sticking with a human-in-the-loop approach, trusting travel experts with the last word. Feel free to use our Laugavegur shopping list (linked also above) but please, check it yourself.
That said, as our experience with Perplexity shows, we’re getting remarkably close to AI systems that can prepare you for success in rugged outdoor adventures.
Agentic Foundry + NewEcom.ai :
Purpose-built Capabilities for Agentic Ecommerce
NewEcom and Agentic Foundry have joined forces to help brands seize competitive advantage with AI. Are you ready to transform online sales with a conversion engine that recognizes purchase context and anticipates customer needs?
#Agentic #AICommerce #Rufus #Perplexity #NewEcom #AgenticFoundry #FutureCommmerce #AISearch #GenAISearch #AIShopping #VirtualAssistant
