[AI shopping & agentic commerce]() / [Use cases]() / [AI answer accuracy]() [How it works]() [What you get]() [Scope]() [Start a review]() [AI shopping & agentic commerce]()[Found & right]() # Find and fix what AI assistants say about your products. A weekly audit of what ChatGPT, Google, Copilot and Perplexity tell shoppers about your products, with every wrong answer traced to its source and fixed there. [Start with an AI shopping review]() [How it works]() Answer audit · 2,000 shopper questions Five weeks Assistant Right Wrong detail Wrong ChatGPT 360 120 30 Google AI Mode 330 160 50 Copilot 290 130 80 Perplexity 200 110 140 Right means price, stock and claims match Fernway’s approved facts. The audit your ecommerce lead reads each week: every question sorted by whether the answer matched your facts. ## Sound familiar? - Ask ChatGPT about your bestseller and it quotes a price you changed a month ago. - Your feeds say “accepted”, and nobody has read what the assistants actually say. - A shopper repeats a claim you never made, and support has to correct it. A wrong answer sends a ready buyer to the wrong variant, an old price or a promise you never made. None of it shows up in a report. Fernway’s launch week, one week of audited questions 1. ## Four hundred questions, asked this week. Written from your support and search questions, and asked on a schedule so every change shows up. 2. ## Three hundred eighty got an answer. Some assistants don’t answer at all. Those are logged, never counted as right. 3. ## Three hundred named the right product. The right jacket in the right variant. A sold-out one, or a lookalike, drops out here. 4. ## Two hundred thirty-six had the right facts. Price, stock and claims match your approved record. Everything else goes to a ledger with an owner. 5. ## The audit became the week’s fix list. Each wrong answer is traced to its source and fixed there, then asked again. Assistant · question **400**Asked **380**Answered **300**Right product **236**Right facts Where it landed CG ChatGPT · washable under $200 Reached right facts Right jacket, right price GO Google · rain jacket for hiking Reached right facts Says not waterproof CG ChatGPT · gift for a hiker Reached right facts Right, with delivery date GO Google · machine washable? Reached right facts Right, cites care label CO Copilot · olive medium Reached right product Navy offered while sold out PX Perplexity · Fernway jacket price Reached right product Old $158, fixed at source PX Perplexity · waterproof? Reached answered Wrong claim, corrected ME Meta · jacket sale Reached asked No answer Answer audit · 2,000 shopper questions Five weeks Assistant Right Wrong detail Wrong ChatGPT 360 120 30 Google AI Mode 330 160 50 Copilot 290 130 80 Perplexity 200 110 140 Right means price, stock and claims match Fernway’s approved facts. [Read the answer audit]() ## What changes. ShowTodayWith the engine Checking A screenshot of one good answer The same shopper questions, asked every week Feeds Accepted means done Sent, processed and observed in every assistant Fixes Argued with each assistant Corrected once, at the source Reporting Mentions and share of voice Right answers out of questions asked ## How it works. 6 steps, 4 stages. - AI**2** - Rules and automation**3** - Your team**1** - Approval 1. 1 Ask 1. AI ### Ask every assistant A fixed list of shopper questions runs against ChatGPT, Google AI Mode, Copilot and Perplexity, and runs again when a price, variant or claim changes. - ChatGPT - Microsoft Copilot - Perplexity 2. 2 Check 1. Rules and automation ### Check against your facts Each answer is compared with the approved record, by variant, price, stock and claim. - Shopify 2. AI ### Trace the cause Groups wrong answers by where the fact came from: a feed, an old page, a reseller listing or the assistant’s own guess. 3. 3 Fix 1. Your team Approval ### Approve the fix Your catalog owner approves any change to a fact, and says what you don’t claim. 2. Rules and automation ### Correct at the source Price, stock, variant and claim fixes publish to every destination from one record. - Shopify - Merchant Center 4. 4 Watch 1. Rules and automation ### Ask again, and log it The same questions re-run. Each fix is marked sent, processed and seen. Each audit writes the next question list ## What you get. Real work, every week. Defect ledger This week - Perplexity · jacket priceQuoted $158, from an old page **$168**Fixed - Perplexity · “waterproof”A claim Fernway has never made **Corrected**Watching - Copilot · olive mediumOffered while sold out **Navy M**Fixed Every wrong answer, where it came from and what fixed it, with an owner. Claims · CAT-042 Owner: merchandiser Machine washable on cold. Not waterproof. Source Supplier spec, revision 2 Says Machine washable on cold Doesn’t claim Waterproof, or any rain rating Checked Every Monday, in four assistants The claims register: what you say, what you don’t, and where it comes from. Your merchandiser owns it. ## How you’ll know it’s working. 1. Right answers out of questions asked Counted on a fixed question set, by assistant, with price, stock and claims matched to your facts. 2. Time from a wrong answer to a corrected source How long a defect lives before the next shopper meets it. 3. Repeat defects Whether a fix at the source stays fixed. Baselines agreed before we build. ## Works with what you have. Commerce - Shopify Search & AI - Merchant Center - ChatGPT - Microsoft Copilot - Perplexity - Search Console We configure first Shopify’s free Knowledge Base and Merchant Center’s optional attributes already take product facts. We set those up properly, then check what each destination actually shows. We build when Wrong answers keep coming from facts the native tools can’t hold, like variant mappings, exclusions or claims you don’t make. ## What a first build includes. Built in your accounts, around your team. Scope, fee and baseline agreed before we start. 1. 1 Design 2. 2 Build 3. 3 Run it, or hand it over [Start with a review]() [Scope this build]() ### We build - A shopper question set, written from your own support and search questions - A weekly audit across four assistants, scored against your facts - A defect ledger with an owner and a fix for every wrong answer - A claims register that says what you do and don’t claim ### We start from - Product facts, variants, prices and stock - Support and site-search questions - Shopify, Merchant Center and feed access ### From your team Owner Head of ecommerce, with the catalog owner Involves - Ecommerce - Catalog - Merchandising - Engineering ## Questions teams ask. Can you make assistants recommend us more often? No, and nobody can promise it. We make what they say about your products right, then track referrals where the data allows. Shopify describes its Knowledge Base the same way: it improves the answers and makes no promise about how often you appear. How do you ask the assistants? The same questions, from a fixed list, on a schedule. We record the date, the assistant and the version of your facts, because answers vary from one ask to the next. What if the assistant is wrong and our feed is right? Then we trace where it read the wrong fact. Often it’s an old page, a reseller listing or a variant mapping, and those we can fix. Who owns it once it’s running? You do. It’s built in your accounts, documented, with training for the people who run it. Ownership usually sits with your head of ecommerce, with the catalog owner. We can keep running it with you, or step back. How do we start? Usually with an AI shopping review: two to four weeks, a fixed fee credited to the build, and a plan that’s yours. If the job is already clear, we can scope the build directly. ## Part of your AI shopping engine. One of 13 use cases, in 5 parts. [See the whole AI shopping engine]() 1. [Found & right **3**]() 2. [Help choosing **3**]() 3. [Orders that complete **3**]() 4. [Buying on account **1**]() 5. [Proof & route **3**]() In found & right - [**AI shopping readiness**Get found by AI shopping assistants]() - **AI answer accuracy**You’re here - [**Buying questions**Answer the buying questions that decide a sale]() Works with - [**Delivery-aware answers**Recommend only what can arrive in time]() See it at work - [**In Fernway’s launch week, through AI**Day 1: four assistants, three right answers]() - [**Also in**Growth & acquisition]() ## Find out what assistants say about you. Start with an AI shopping review. [Start with an AI shopping review]() [**Talk it through first**15 minutes]() We reply within one business day. --- Source: [https://www.thekeating.co/use-cases/ai-answer-accuracy/]() [Contact Keating & Co](mailto:nelson@thekeating.co) · [Agent guidance](https://www.thekeating.co/llms.txt)