Who captures the sale when AI recommends your brand?
Updated: Oct 24
TL;DR: AI is changing where buyers research faster than where they buy, so a recommendation can create a sale or a lead for another business. A brand sold through a marketplace still gets paid but loses the customer relationship, and a curator who supplied the answer may get nothing. In my own scans of about 190 UK holiday-let businesses, around nine in ten did not appear when a simulated AI search asked a general category question. The fix is to make the direct route the most attractive one and to be the entity AI points to.
AI shopping research is moving faster than checkout
PYMNTS Intelligence's October 2026 Consumer AI Benchmark surveyed 2,191 US consumers in August. It found that 23% of US consumers, around 61 million people, now name AI as their main starting point for online retail research. Most of their latest AI-assisted purchases were still completed at large retailers.
Where the latest AI-assisted purchase was completed | Share |
Amazon | 59.7% |
Big-box and department stores | 19.5% |
The brand's own website or app | 6.3% |
59.4% of AI users agree they now research on AI products they used to search for on Amazon, and 57.9% say they usually still buy on Amazon even when AI helps them find the product.
AI does send some buyers somewhere new: 44.7% agreed it had led them to buy from online stores they had never used before.
These are self-reported figures from US shoppers. PYMNTS is explicit that the data shows where research and purchase happen, not that AI caused Amazon's share.
The four buying stages of an AI-assisted purchase
Each buying stage can be owned by a different business.
Buying stage | What happens | Who may own it |
Discovery | The buyer starts researching in an AI tool | The AI platform |
Recommendation | The answer names brands, products or providers | The business AI chooses to mention |
Destination | The buyer goes somewhere to verify or buy | A brand site, retailer, marketplace or directory |
Transaction | Money changes hands, or the enquiry is made | Often a different business from the one recommended |
Asked for the main reason they chose where to buy, 27.5% named price or a discount and 17.3% named free or fast shipping. Only 2.5% named a direct link from the AI tool.
When the survey held the item, price, shipping and returns equal, 40.2% said they would send an AI assistant to a large marketplace. 16.1% chose a brand's own website, and 21.8% had no preference.
Where the sale goes depends on whether the buyer already knows you
I ran five simulated AI searches for each of about 190 UK holiday-let businesses. Around nine in ten did not appear when the question was a general category search, the kind a buyer asks before they have chosen anyone. Among the roughly 180 operators, just over half appeared when the question was about booking direct, about four in ten appeared in part, and about one in twenty did not appear at all.
A buyer who already knows the operator tends to find the operator. A buyer who asks a general question gets whoever AI names, and for most of these businesses that was someone else. The scan doesn't record who AI named instead or which page the buyer would reach, so it shows where operators are missing and says nothing about where the sale went. Each prompt was run once, so treat the figures as indicative.
This fits the PYMNTS finding that 44.7% of AI users had bought from a store new to them. An operator missing from general searches doesn't reach those buyers. The full results are on the AI visibility audit page for UK holiday lets.
What it costs you when the sale lands elsewhere
The cost depends on whether you sell products, supply information, or win clients through enquiries.
1. Brands and providers whose sale lands on a marketplace or booking platform
The revenue still arrives, so this is a margin and relationship problem. Marketplace fees reduce what you keep. More importantly, you often don't get the customer's details or a route to contact them again.
A direct customer can be re-engaged through email, through an account they hold with you, or by searching for you by name. A marketplace customer often has to be won again from scratch, through ranking, advertising or, now, another AI recommendation. The cost shows up in repeat purchase and lifetime value, not on the first order.
The same applies to hotels and holiday lets: a booking through a third-party site fills the calendar, but only the guest who booked direct is someone you can invite back.
2. Curators, directories and comparison sites
In a hypothetical case, a curator publishes a detailed guide to a region. A traveller asks an AI assistant for a family-friendly place to stay, the answer draws on the guide and names a property, and the traveller books through a platform.
The curator supplied the information and received nothing from that booking. What that costs depends on how the curator earns. One paid by commission on bookings loses the revenue directly, with nothing to fall back on if AI delivers the answer without sending the buyer on. One paid by owners through an annual fee or subscription keeps its revenue for now, but owners renew because the listing sends them enquiries. If AI answers absorb those enquiries and the curator can't show owners a measurable referral, the loss arrives at renewal. That second pattern is an inference from how these businesses are funded, not something the scan measured.
3. High-touch services and B2B
The PYMNTS data covers US retail, so it doesn't show how this plays out for clinics, dental practices, law firms, finance providers or SaaS. The same question applies, with an enquiry or sign-up in place of a purchase.
A buyer asks AI for options, then checks a review site, a directory or a software marketplace, and enquires or signs up through it. You never receive their name or a booked call, and you may be paying for the lead you would have won directly. That is an inference from the pattern rather than something the survey tested, and it is worth testing in your own category.
Businesses where the only way to buy is direct are less exposed, because the recommendation and the transaction can only meet in one place.
How to check where AI-influenced sales land
Ask the buying-stage question. Put a "where should I book" or "which should I choose" prompt for your category into ChatGPT, Gemini and Perplexity. Note which entity the answer sends the buyer to.
Ask your customers. Ask recent buyers or enquirers where they checked you before choosing. Compare that with what AI named.
Work out what a leaked sale costs you. For a product or booking, compare what you keep from a direct sale with what you keep from a marketplace sale, then how often each customer comes back. For a service, compare what a directory lead costs with what a direct enquiry costs.
How to make the direct route the one buyers take
The aim is to make your own site the most attractive place to finish.
Give them a reason on the terms that decide it. Price and delivery drove the choice of destination in the survey. If you can't match a marketplace on price, compete on what it can't offer: a member rate, a bundle, an extended guarantee, flexible changes, or a direct-only extra.
Make that reason readable by AI. If your direct benefits sit in a banner or an image, an AI answer can't repeat them. State pricing, availability, delivery and direct-only terms in plain text on pages AI can read. My post on whether your website is ready to convert AI traffic covers what the landing page must confirm.
Be the clearest entity. Use one consistent name, one official booking or enquiry route, and the same details across your site and profiles. That gives AI less reason to point at a directory and more reason to point at you. For high-touch services this is the main lever.
Capture the relationship on the first sale. For marketplace orders, that means something worth registering for: warranty cover, a repeat-order discount, or a reorder reminder. Some marketplaces restrict inserts and links that steer buyers off-platform, so check the seller rules first.
Become the name people search for. Repeat buyers who ask for you by name skip the category prompt where AI and marketplaces compete.
Use marketplaces deliberately. Treat them as an acquisition channel and your own site as the retention channel, and know your margin gap so you choose that on purpose.
If you're a curator, make your value measurable. Referral or affiliate agreements, bookable experiences inside your own pages, and licensing your content are ways to be paid for the information you supply, though whether a platform accepts them is partly out of your hands. If owners pay you a fee, report the enquiries and referrals your listing produces, because that evidence is what decides whether they renew.
If AI agents begin completing bookings and purchases on a buyer's behalf, agent experience adds a further layer. My earlier post on whether AI search traffic converts covers a different gap, which is buyers who return to your own site later by another route.
Frequently asked questions
If AI recommends my brand, who gets the sale?
The recommendation and the transaction can happen at different businesses. A buyer may research with AI, then buy on a marketplace, a retailer or a booking platform they already trust.
Is a sale on Amazon or a booking platform bad for a brand?
Not necessarily. The revenue still arrives. What you give up is usually margin, customer data and the easy route to a repeat purchase, so the real cost depends on how often your customers come back.
Do AI recommendations send buyers to the brand's own website?
Often not for the purchase itself. In the PYMNTS survey, 6.3% of latest AI-assisted purchases were completed on a brand's own website or app, and 2.5% of respondents said a direct link from the AI tool was the main reason they bought where they did.
Does AI find a business more easily when buyers already know its name?
In my simulated searches of UK holiday-let businesses, just over half of the operators appeared when the question was about booking direct and about four in ten appeared in part, while around nine in ten businesses did not appear for a general category question.
How can a business get buyers to finish on its own site?
Give them a reason on price, delivery or direct-only benefits, state it clearly in text AI can read, and make your own entity the clearest option so AI doesn't send buyers to a directory instead.
Is this only a problem for retailers?
No. Any business whose buyers can verify or purchase through a third party is exposed, including clinics, law firms, finance providers and SaaS. The survey covers US retail, so test the pattern in your own category.
How can a business see where AI-influenced sales land?
Not fully, yet. Referral reports show clicks to your own site. Asking customers where they checked you, and testing which provider AI names at the buying stage, gives a practical view.
The AI Visibility Scan shows where AI recommends you, and a readiness audit checks whether the pages behind that recommendation convert.
Comments