“What’s out there?”
“Help me find the best sofas under $1,000.”
AI helps the shopper discover options. That is the visibility question; our test starts with the next step.
FRICTION AI BUYER TEST
Your next buyer
might be AI.
Would it choose you?
We put your product and three competitors in front of AI buyers across 30 custom situations. Find out who gets chosen, why, and what you can change.
A done-for-you testing service. Human-reviewed reports. Competitive decisions—not AI SEO or GEO.
1,800 prompt runs Leading AI models Human-reviewed reports
THE BUYER’S SIDE OF THE STORYIllustrative
Same sofas. Different buyer. Different choice.
“I have three dogs and a $1,500 budget. Compare these sofas for easy cleaning, everyday wear, and comfort. Which would you buy, and why?”
Your product$1,290
Competitor A$1,390
Competitor B$1,690
Competitor C$1,450
What tipped the decision?
Competitor A makes removable, machine-washable covers explicit. Your page says ‘easy care’ but leaves the cleaning method unclear.
Care instructions Change the buyer situation to explore
Oneproduct
Threecompetitors
Thirtycustom buyer situations
1,800AI prompt runs
Oneprioritized action plan
30 situations × 3 variations × 20 model-and-agent setups. Start with a product, a collection, or your brand.
THE BUYING JOURNEY IS CHANGING
Before AI buys for them,
it helps them choose.
The shift starts well before checkout. A shopper brings their budget, their needs, and a shortlist to AI. The answer can shape which product they buy next.
“Which one fits my life?”
“I’m choosing between sofas from Brand A, B, C, and D. I have two dogs, a small apartment, and a $1,000 budget. Which should I buy, and why?”
My budgetMy constraintsMy shortlist
All four brands are in consideration. Now the buyer’s needs decide which features matter, which trade-offs are acceptable, and which missing details leave doubt. This is the decision Friction tests.
“Take care of it for me.”
“Choose the best fit, check delivery, and make the purchase within my budget.”
As shoppers delegate more, agents can carry that comparison further into the buying journey.
You don’t have to wait for autonomous shopping to care. Your customer can still click “buy” themselves, after AI has helped decide who gets the sale.
REPORTED SHOPPER BEHAVIOR39%
AI is already part of the shopping journey.
In Adobe’s survey of more than 5,000 U.S. respondents, 39% said they had used AI for online shopping.
Adobe Digital Insights · April 20262030 FORECAST · U.S.10–20%
AI could handle more of the purchase.
Morgan Stanley estimates agentic shoppers could account for 10–20% of U.S. e-commerce spending by 2030. This is a forecast, not today’s share.
Morgan Stanley Research · December 2025This shift is already reaching Shopify: eligible stores can make products available in AI shopping channels through Agentic Storefronts. Read Shopify’s guide
AI SEO / GEO: DISCOVERY
“Can AI find my product?”
FRICTION: COMPETITIVE CHOICE
“Given these options and this buyer, which would AI choose?”
WHAT WE TEST
You’re on the shortlist.
Do you become the choice?
In controlled tests, we put your offering and three competitors into the comparison. Then we change the buyer’s needs and the AI doing the choosing.
The competitive set.
Your product, collection, or brand. Three alternatives. The customers you want to win.
The buying decision.
Which option each AI chooses for that buyer, what it cites, and where its answer changes.
The next moves.
An insight brief and detailed report, with prioritized actions for your team.
This is not a visibility service. We don’t test whether AI finds you unprompted or sell higher rankings and more mentions. We test how it evaluates you against the alternatives you care about.
A SMALL GAP. A DIFFERENT CHOICE.
Your product can be a great fit.
Your listing has to show it.
A washable cover. A clear warranty. A delivery date. The detail you leave out can be the reason an AI buyer favors a competitor.
We separate missing evidence from product weaknesses, so you know whether to improve the page, sharpen the positioning, or rethink the offer.
Try adding the missing product detail.
THE PET-OWNER TESTIllustrative
YOUR PRODUCT PAGE
The Everyday Sofa
A comfortable sofa for everyday life.
Care instructions
Made with durable, easy-care fabric.
“Easy care” leaves a question open.
For a home with three dogs, a competitor’s clear washing instructions may be more convincing.
Only publish details that are true of your product. More information helps evaluation; it does not guarantee a win.
YOUR OFFERING. YOUR COMPETITIVE SET.
Pick the decision you want to understand.
Choose one scope for your test. We build the 30 buyer situations around your customers, your offering, and three relevant alternatives.
One product. Three alternatives.
Get specific about the features, details, and trade-offs that help one product win a particular buyer.
Product-page clarityFeature trade-offsBuyer fit
Your product
Product A
Product B
Product C
THE BUYER’S QUESTION“Which sofa fits my home and the way I live?”
All three formats use 30 custom buyer situations. We agree one scope and three competitors before the test.
ONE BUYER BRIEF. MULTIPLE AI PERSPECTIVES.
The same buyer.
A different AI.
A different choice.
One model may favor your product. Another may prefer a competitor. Even the same model, working through a different agent or browsing tool, can find different evidence and reach a different decision.
THE BUYER BRIEF STAYS THE SAME
Illustrative · fictional results“I have dogs, a small living room, and a $1,500 budget. Compare your sofa with these three alternatives. Which fits my needs best?”
WHY THE CHOICE CHANGES
Model A · assistant
It gives more weight to your compact dimensions and price, treating easy cleaning as a secondary trade-off.
Compact dimensionsWithin budgetCare details uncertain
YOUR NEXT MOVE FOR THIS SETUP
Your small-space advantage is understood. Keep exact dimensions and access details easy to find.
A demonstration of the comparison, not measured behavior of any named model. Your report uses actual tested models and setups, with the responses and conditions behind each finding.
See the whole competitive picture.
We test across all leading AI models, using 20 model-and-agent setups. See the patterns that hold across tools and the disagreements an overall summary could hide.
Know which AI chooses you.
Compare choices within the same buyer situation. See which models, agents, and harnesses favor your offering, which pass it over, and the reasons they give.
Refine for the tools that matter.
Turn model-specific gaps into changes to your listings, positioning, and supporting evidence. Focus your effort on the evidence each setup needs to make an informed choice.
A harness is the software around an AI model that gives it tools to browse or act. We confirm the exact 20-setup lineup before testing and identify each setup in your report.
THE WORK HAPPENS ON OUR SIDE
You bring the context.
We do the testing.
You provide the business context. Our team designs the situations, runs the tests, and reviews the responses. You receive the findings and recommendations without operating a testing tool yourself.
01
Tell us about your buyers.
Share your offering, three competitors, ideal customers, repeat buyers, and common purchase objections.
02
We build 30 buyer situations.
Different needs, budgets, constraints, and trade-offs. Three prompt variations per situation test how sensitive the answers are to wording.
03
We run the comparisons.
We run 1,800 prompts across 20 setups spanning leading AI models, agents, tools, and harnesses. Shopper-facing answers are paired with structured choice and evidence fields.
04
We make the findings useful.
Our team reviews the outputs, checks patterns against the answers, and separates repeated findings from inconsistent results.
30buyer situations
×3prompt variations
×20model-and-agent setups
=1,800prompt runs to learn from
1,800 prompt runs across the agreed comparison scope. Twenty setups can include multiple tools or harnesses using the same underlying model. We confirm the lineup before testing. These are simulated decisions, not completed purchases.
The value is in the pattern, not just the prompt count. One answer is one observation. Thirty buyer situations reveal where context changes the choice. Three wording variations help expose fragile results. Twenty AI setups show where tools agree or diverge. Our team turns those patterns into priorities, with the supporting responses and test conditions behind them.
THE DELIVERABLE
The insight to act.
The evidence to back it up.
1,800 prompt runs, distilled into two documents. Understand where you stand overall, how individual models and agents see you, and what to work on next.
01 / THE INSIGHT BRIEF
Your next moves, in 1–2 pages.
Where you win, where you lose, and the changes to make first. A short brief your team can use to decide what gets worked on next.
02 / THE DETAILED REPORT
The full competitive picture.
Competitor comparisons by buyer situation and by model, agent, and harness. Supporting responses, areas of disagreement, and targeted actions for the tools that matter to your buyers.
Friction / Buyer test reportSample report · illustrative data
Buyer fit, situation by situation.
Select a row to explore the finding.
Scroll across to compare all three competitors →
| Buyer situation | You | Comp. A | Comp. B | Comp. C |
|---|---|---|---|---|
| Unclear | Chosen | Over budget | Not a fit | |
| Chosen | Not a fit | Unclear | Not a fit | |
| Unclear | Not a fit | Over budget | Chosen | |
| Chosen | Unclear | Not a fit | Unclear |
A small excerpt, not an aggregate score. The full report examines all 30 situations, prompt variations, and agreed AI setups.
Human interpretation, supported by test evidence.Preview only · not a live test
FROM FINDING TO ACTION
What kind of problem are you solving?
The same lost comparison can call for very different work. The report helps you tell these apart.
The product fits. The evidence is missing.
Your covers are washable, but the page doesn’t explain how. Publish the verified care details so buyers can assess the fit.
Next move: clarify the listing.Your advantage misses this buyer’s priority.
You lead with style. The buyer needs a compact sofa that fits a narrow doorway. Make the relevant dimensions and access details explicit.
Next move: sharpen the buyer story.The alternative actually meets the need better.
The buyer needs removable covers and yours are fixed. Better copy cannot change that. Consider a different buyer segment or a product change.
Next move: reconsider the fit or offer.Illustrative findings. We assess the AI’s stated reasons against available product evidence and flag uncertainty. Your team decides which recommendations to put into practice.
BUILD IT INTO YOUR WORKFLOW
Test once.
Learn again when things change.
Start with a baseline. Come back when there’s a new decision to make, a major page update, or a competitor worth watching.
Test Learn Improve
SIX REASONS TO RUN A BUYER TEST
A test for the decision on your desk.
BEFORE YOU PUT BUDGET BEHIND IT
Launch with a clearer view of the competition.
Test a new product or collection against the alternatives your target buyers already consider. Find out which buyer situations it wins and which need a stronger proposition.
Before launchA QUESTION YOU COULD BRING US
“We’re launching a compact sofa. Would AI choose it for apartment renters over three established alternatives?”
New product or collectionThree established competitorsYour intended buyer
WHAT YOU WALK AWAY WITH
A launch brief showing your strongest buyer fits, unanswered questions, and positioning to sharpen before you scale acquisition.
FOUNDING MERCHANT OFFER
A clear answer.
A one-time price.
1,800 prompt runs across leading AI models and agent setups. Two human-reviewed reports. Prioritized next steps. The first 10 merchants get the complete service at an introductory price.
First 10 merchantsSave $200
Founding merchant test
$299USD · one time
A complete competitive decision test, from your buyer brief to both reports and prioritized recommendations.
1,800
AI prompt runs30 situations × 3 variations × 20 setups
- Personally led end-to-end by Friction’s founder
- One product, collection, or brand
- Three competitors · 30 custom buyer situations
- 1,800 prompt runs · 20 model/agent setups
- Coverage across all leading AI models
- Choices and recommendations broken down by setup
- Human-reviewed insight brief + detailed report
After the introductory offer
Standard buyer test
$499USD · one time
The same end-to-end service for merchants joining after the first 10.
1,800
AI prompt runs30 situations × 3 variations × 20 setups
Same depth. Same deliverables.
All leading AI models across 20 model-and-agent setups. Findings by buyer context and by setup, both human-reviewed reports, and prioritized recommendations.
Each package covers one agreed comparison scope. We confirm competitors, AI coverage, and delivery timing before you book.
Is this an AI SEO or visibility audit?
No. Friction AI Buyer Test is competitive decision testing, not AI SEO or generative engine optimization (GEO). We put your offering and three chosen competitors into the test brief, then ask AI to evaluate them for specific buyer situations. We report who gets chosen, the stated reasons, and what you can change. We do not measure unprompted discovery, sell visibility improvements, or promise more mentions in AI answers.
Can you test a product, a collection, or an entire brand?
Yes. Choose one comparison scope for your test: one product against three competing products, one collection against three competing collections, or your brand against three competing brands. We agree the boundaries and competitors with you before starting, so the comparison stays useful.
Which AI models and agents do you test?
The test covers the leading AI model landscape through 20 model-and-agent setups, including models, shopping assistants, tools, and harnesses (the software that lets an agent browse and act). We confirm the exact lineup before starting. A setup is a model paired with a particular tool or agent environment, so 20 setups does not mean 20 distinct underlying models. Thirty buyer situations, three prompt variations, and 20 setups produce 1,800 prompt runs. Your report breaks down choices, reasons, and evidence by setup and records versions where available, dates, and test conditions.
Can the report help me improve for a specific AI model or agent?
Yes. Alongside the overall competitive picture, we show which tested models and agent setups choose your offering, which favor a competitor, and where their stated reasons differ for the same buyer context. We turn those differences into targeted recommendations for your product information, positioning, and supporting evidence. We help you refine how your product is represented; we do not train or modify the AI models themselves.
What do you need from me?
Your store and product, collection, or brand links; three competitors; and a short brief about your ideal buyers, repeat customers, common objections, and what makes your offering different. We use that context to build 30 custom buyer situations. You do not need to send customer-level personal data.
Will you change my store or guarantee more sales?
We deliver analysis and recommendations; your team decides which changes to make. AI responses vary with the model, prompt, available information, and timing. The report is evidence from controlled tests, not a prediction of every shopper’s experience, a sales forecast, or a guarantee of recommendations. Simulated purchase decisions do not place real orders.
Is this a subscription? How long does it take?
The $299 introductory package and the $499 standard package are one-time services, each including the insight brief and detailed report. We confirm a delivery date before you book. If you want a monthly or biweekly testing cadence, we can agree a separate scope and price with you. There is no automatic subscription.
FRICTION AI BUYER TEST
You built a product worth choosing.
Find out if AI chooses you.
1,800 prompt runs. Leading AI models. Your next move, made clearer.
Bring your store, your competitors, and your buyer context.