Chosen by AI. Bought from you.

The right product, price and stock in ChatGPT, Google and Copilot. An assistant on your own site that helps people choose, and orders that complete with bundles, accounts and service intact. Built on the platform you have, then run with you or handed over.

Led by operators who’ve led commerce platforms supporting $2B+ in annual revenue, and built a commerce-intelligence product of their own.

Shoppers ask an assistant. Your facts decide the answer.

Assistants answer from whatever facts they can find. We make your product facts right at the source, so the answer, the price and the stock are right everywhere, and the order that follows holds together.

Shoppers ask ChatGPT, Google AI Mode, Copilot, Perplexity, your site and business buyers. Every question draws on your product facts, ends in an order, and comes back as the next question.ChatGPTGoogle AI ModeCopilotPerplexityYour siteBusiness buyersYour product factsThe orderNext question

The facts

Catalog & product information

The answer and the order

AI shopping & agentic commerce · this page

The delivery

Supply chain & fulfillment

How shoppers find and buy a product, then and now

Every channel had its own listing.Each judged on clicks.

Each channel held its own copy of your product: a shopping feed, a marketplace page, a social catalog. Every report counted impressions and clicks, and nobody checked what a shopper was told.

Then the assistants started answering.From whatever facts they could find.

ChatGPT, Google AI Mode, Copilot and Perplexity answer in the shopper’s own words, and Shopify’s ChatGPT route is on by default for eligible US stores. A feed being accepted, an answer being right and a checkout being allowed are three separate questions.

So the edge moved to being the right answer.And an order that actually completes.

The facts behind the answer, the rules only you know, like which liner fits which jacket, and proof of what it earned decide the sale now. A payment credential is only one dependency in a completed purchase.

Skip the walkthrough

Look inside.Six views, built around your products.

  1. Your products.Variants, prices, stock and what’s true.

    Every variant, its price, its stock and the claims you stand behind. One set of facts that every assistant, feed and page reads from.

  2. AI answers.Right in every assistant.

    The same shopper questions go to every assistant each week, and every answer is checked against your facts. A wrong price, an old stock level or a claim you never made goes to an owner with the fix.

  3. Your assistant.Asks before it suggests.

    On your own site, the assistant asks one question, then shows a few products with the reason for each and a delivery date. When it doesn’t know, it says so and hands off with the context.

  4. Checkout.A checkout or a handoff, never a lost basket.

    Each route gets a decision: native checkout where it carries the whole basket, a handoff to your own checkout where it doesn’t. Every order reconciles from the first intent to the last shipment.

  5. Accounts.Business buyers, on their own terms.

    A verified buyer, their account and ship-to, their units and their contract price. When an order is over the limit, it waits for a person, however it’s split.

  6. Proof.What it earned, with a holdout.

    Shoppers are assigned before they see the assistant, and everyone assigned is counted, including those who never opened it. The rule for keeping it is written first.

Built in your accounts.Yours to keep, whoever runs it.

Your product facts, checkout rules and tests stay in accounts you control. If we ever part ways, the engine stays.

One question, every assistant. All the way to the order.

In week 5 of Fernway’s launch, shoppers started asking assistants about the canvas field jacket. Here is the system at work, from the first answer to the order, and then the proof.

  1. Day 1

    Four assistants. Three right answers.

    A shopper asks for “a washable field jacket under $200 that arrives by Friday” in ChatGPT, Google AI Mode, Copilot and Perplexity. Perplexity quotes an old $158 price and calls the jacket waterproof, which Fernway has never claimed.

    AI
    Asks every assistant, every week
    Rules
    Each answer checked against your facts
    Your team
    Owns the facts
    Use caseFind and fix what AI assistants say about your products
    Washable under $200, by FridayDay 1
    AssistantAnswerVerdict
    ChatGPTCanvas field jacket, $168Right
    Google AI ModeCanvas field jacket, $168Right
    CopilotCanvas field jacket, $168Right
    Perplexity$158, “waterproof”Wrong price and claim

    Fernway’s price is $168. It has never claimed the jacket is waterproof.

  2. Day 2

    Fixed once, at the source.

    The price, the “not waterproof” fact and the variant mapping are corrected once, in Fernway’s product facts. Each destination then shows what was sent, what it processed and what it now says.

    AI
    Finds the wrong fact and where it came from
    Rules
    One record feeds every destination
    Your team
    Approves the fix
    Use caseGet found by AI shopping assistants
    Price correction · $168Day 2
    DestinationSentProcessedObserved
    ChatGPT$1689:10 am$168
    Google AI Mode$1689:12 am$168
    Copilot$1689:15 am$168
    Perplexity$1689:20 am$158 · re-read due

    Observed is read back from the assistant’s own answer.

  3. Day 4

    Olive M sold out. The answers followed.

    Within the route’s refresh window, answers stop offering Olive M and show Navy M, arriving Thursday. Price and stock refresh about every 15 minutes where the route supports it.

    AI
    Picks the closest variant in stock
    Rules
    Sold-out variants leave the answer
    Your team
    Decides what counts as a substitute
    Use caseRecommend only what can arrive in time
    Olive M sells outDay 4 · 10:02 am
    AssistantNow offersUpdated
    ChatGPTNavy M, Thursday10:15 am
    Google AI ModeNavy M, Thursday10:12 am
    CopilotNavy M, Thursday10:17 am
    PerplexityOlive M until its next readTonight

Accepted. Not the same as accurate.

Two variants every feed accepted. Switch views to see what the assistants told shoppers.

Whose view

Canvas field jacket · Navy M

Fernway Outfitters · $168

Feed statusAcceptedUnchanged

Shopper questions asked100 audited shopper questions

  • Wrong priceNot shared3
  • Wrong variantNot shared2
  • Wrong factsNot shared3
  • Right answerNot shared92
DestinationsRight answers
4 of 492 of 100
Last syncWorst error
TodayAn old price, 3 times

LiveKeep

Canvas field jacket · Olive M

Fernway Outfitters · $168 · during the stockout

Feed statusAcceptedUnchanged

Shopper questions asked100 audited shopper questions

  • Wrong priceNot shared27
  • Wrong variantNot shared21
  • Wrong factsNot shared14
  • Right answerNot shared38
DestinationsRight answers
4 of 438 of 100
Last syncWorst error
TodaySold-out variant offered

LiveFix at the source

What are assistants telling shoppers about your products right now?

An AI shopping review finds out across ChatGPT, Google, Copilot and your own site, in two to four weeks.

  1. Day 5

    On your own site, one question first.

    On fernway.com the assistant asks one thing, “Rain, or mostly dry?”, then shows three jackets with a reason and a delivery date each. It only offers what can arrive by Friday.

    AI
    Asks one question, then explains the three
    Rules
    Only jackets in stock, with a date
    Your team
    Owns the facts and what it won’t say
    Use caseBuild a shopping assistant that helps customers choose

    Mostly dry. Weekend hikes.

    Three that fit, all under $200 and here by Friday. Canvas field jacket, Navy M, $168, Thursday: washable on cold, no rain rating. Canvas field jacket, Sand M, $168, Thursday: same fabric, lighter color. Waxed field jacket, M, $168, Friday: heavier and spot-clean only.

    After one question: “Rain, or mostly dry?” · From Fernway’s product facts

  2. Day 6

    The basket held together.

    A ChatGPT shopper adds the Care Kit bundle to the jacket. The native route can’t carry a bundle, so the basket hands off to Fernway’s own checkout with everything still in it.

    AI
    Explains the handoff in plain words
    Rules
    A parity matrix decides checkout or handoff
    Your team
    Chooses where native checkout is worth it
    Use caseKeep AI-referred sales whole, from answer to order
    Basket · INT-104

    2 of 2basket lines kept at Fernway’s checkout

    • Canvas field jacket · Navy M · $168The native route can carry it
    • Care Kit bundle · $28Not supported on the native route
    • Handed off to Fernway’s checkoutBasket and promise intact
    • Delivery promise · ThursdayCarried to the order
  3. Day 7

    The order reconciled, and service knew the promise.

    Intent, checkout and order sit on one record: INT-104, CK-104 and ORD-8512. When the shopper writes in, service sees the Thursday date the assistant gave.

    AI
    Explains where the order stands
    Rules
    One shared id, one order per purchase
    Your team
    Operations resolves the exceptions
    Use caseReconcile AI-placed orders from payment to fulfillment
    One purchase · INT-104Day 7
    StepRecordResult
    Shopper intentINT-104Jacket and Care Kit, $196
    CheckoutCK-104Authorized · promise: Thursday
    OrderORD-8512Created, sent to ShipBob
    ServiceCase viewSees the Thursday promise
  4. Day 60

    Then we checked what the assistant earned.

    Visitors offered the site assistant bought 2.4% more, about $22,000 a month in contribution. Engaged users converted three times as often, but most were already buying. Fernway kept it.

    AI
    Reads chats for what it couldn’t answer
    Rules
    Half held out, and everyone counted
    Your team
    Decides whether to keep it
    Use caseProve your shopping assistant earns its cost
    Site assistant · 60-day holdout

    +2.4%orders, likely between 0.2% and 4.6%

    15,360Offered the assistant
    15,000Held out

    Keep it · check again in 60 days

Built around your team. Whatever it looks like today.

Your setup

Catalog syndication is on, and nobody has checked the answers.

What we usually find

  • Shopify Plus
  • ChatGPT and Copilot channels
  • Merchant Center
  • Gorgias

Where it falls back on people

The feeds are accepted, so everyone assumes the answers are right. Nobody owns the facts the assistants repeat, or the checkout when a basket doesn’t fit.

What we’d do

We audit the answers first, fix the facts at the source and decide where checkout stays with you. Anything new gets a holdout.

  1. Design
  2. Build
  3. Run

Search and a chatbot, and they don’t share an answer.

What we usually find

  • Constructor or Algolia
  • A chatbot
  • A feed agency
  • Merchant Center

Where it falls back on people

Search, the chatbot and the feeds each hold their own copy of the facts, and shoppers who describe a need still land on a filter page.

What we’d do

We use the search and assistant you already own where they fit, give them one set of facts and test the result with a holdout.

  1. Design
  2. Build
  3. Equip

Compatibility and accounts live in your reps’ heads.

What we usually find

  • An ERP
  • A portal or CPQ
  • A shared inbox
  • A spreadsheet of part relationships

Where it falls back on people

A buyer asks what fits, or emails “same as last time”, and one person answers. A public assistant can’t see accounts, and a guess about compatibility is expensive.

What we’d do

We write the compatibility and account rules down, verify the buyer before anything else and put the buying route over the systems you have.

  1. Design
  2. Build
  3. Run

An AI visibility retainer, reporting mentions.

What we usually find

  • A monthly mentions report
  • A prompt tracker
  • Agency-run feeds

Where it falls back on people

The report counts how often your brand appears. It doesn’t say whether the answer was right, or whether anything sold.

What we’d do

We replace mentions with audits against your facts and a holdout on what moves orders, keep what’s worth keeping and run it with you until your team takes over.

  1. Design
  2. Build
  3. Run

Yours to keep. Whoever runs it.

If your partner disappeared tomorrow, would your team still know what the assistants are telling shoppers?

When the engagement ends, with
What leaves
The prompts, the audits and the know-how.
What stays
A mentions report.
What leaves
The configuration and its definition of success.
What stays
Chat logs.
What leaves
Nothing.
What stays
All of it: in your accounts, documented, with your team trained to run it.
  1. Design

    The engine, mapped to your catalog, your channels and your checkout.

  2. Build

    In your accounts, connected to your stack, tested and documented.

  3. Run it, or hand it over

    We run it with you, or train your team and step back.

Fixed fees to design and build. Monthly to run, with part of the fee tied to agreed targets where the numbers support it.

We’ve built commerce products, and the data behind them.

Our founder built Shoplist, a commerce-intelligence product that crawls and enriches product data at scale, and led commerce platforms in prior roles. Shoplist is a separate, founder-owned product, not something we offer or a source of data for your work.

  • Shoplist a founder-owned commerce-intelligence product, with recurring crawling and enrichment of product data
  • 15+ years building commerce products, teams and systems
  • $2B+ in annual revenue supported by platforms led in prior roles
About the team

Start with an AI shopping review. See what assistants tell shoppers.

2,000 shopper questions about Fernway, sorted by the answer

2,000

Right answer

1,180

Price, stock and claims match Fernway’s facts

  • ChatGPT360
  • Google AI Mode330
  • Copilot290
  • Perplexity200

Right product, wrong detail

520

The right jacket, with a detail off or missing

  • Google AI Mode160
  • Copilot130
  • ChatGPT120
  • Perplexity110

Wrong answer

300

Sold out, outdated or never claimed

  • Perplexity140
  • Copilot80
  • Google AI Mode50
  • ChatGPT30

300 answers, 15% of shopper questions answered wrong: a sold-out variant, an old price, or a claim Fernway never made.

Two to four weeks. A plan that’s yours.

We ask your shoppers’ questions across ChatGPT, Google, Copilot and your own site, check every answer against your facts and size each fix. Every opportunity is given a verdict: do first, validate, defer or leave alone.

  • Two to four weeks
  • Fixed fee, quoted after a short call, credited in full if we build it
  • The plan is yours, whoever builds it
  • The findings are yours whether or not we build

Be the right answer, and the order. Start with an AI shopping review.

We reply within one business day.