When a customer asks AI about your prices, policies or hours, is the answer right?
Customers ask an assistant before they ask your team. When it quotes the old returns window, a withdrawn service or last year’s prices, your team inherits the complaint. The audit finds every wrong claim and the page that caused it.
Four failures. Each one fixable, once it’s found.
- Returns windowAI
30 daysYou 14 days - DeliveryAI
Always freeYou Free over £50 - CancellationAI
No feeYou £25 inside 24 hours
Illustrative
It quotes policies you’ve changed
Returns, refunds, delivery, cancellations — whichever version it read. Your team honours the old terms, or argues with the customer.
Can I cancel my appointment the day before without paying?
“Yes — [the clinic] offers free cancellation up to the day of your appointment, so you won’t be charged…”
Illustrative
It answers the customer before you do
Confident, polite and wrong. The customer arrives expecting it, and the disappointment is your team’s to handle.
- Saturday openingNot known
- Second branchNot known
- Withdrawn serviceStill offered
- Current pricesOut of date
Illustrative
It doesn’t know what’s changed
New services, branches and hours reach the assistant late, or not at all — so the customers they were for never hear of them.
- Forum thread, 2023Cited
- Review quoting an old policyCited
- Your FAQ pageNot cited
- Your terms pageBlocked from crawlers
Illustrative
Old threads answer for your help pages
An old forum thread outranks your FAQ as the source of truth, and every customer who asks gets the thread’s answer.
Your part of the plan. With the evidence behind every fix.
- The Claim Register: every wrong price, policy, service and hour, with the correction
- The page or source behind each wrong claim
- Which of your help and policy pages the engines can read, and which they ignore
- A ranked plan your team can act on — without a developer, where possible
- Policy accuracyReturns, refunds, cancellations, delivery
- Price accuracyWhat customers are told things cost
- Service accuracyWhat’s offered, and what’s been withdrawn
- Hours and placesOpening times, branches, addresses
- Source of truthWhich page the answer is taken from
- ReadabilityWhether crawlers can reach your help pages
Before you ask. Each answered in its first sentence.
Why would AI get our policies wrong if our website is right?
Because your website is one source among many. Assistants also read old pages still online, review sites, directories and forum threads — and when those disagree with your policy page, the answer can follow them.
Can we just tell ChatGPT the right information?
Not directly. There is no form to correct an answer; the fix is to make the correct version the one every source agrees on.
Which pages matter most for customer experience?
Usually the ones customers ask about before buying: prices, delivery and returns, cancellations, opening hours and locations.
Is this the same as review monitoring?
No. Reviews are what customers say after the experience; this is what an assistant tells the next customer before it.
How often should these claims be checked?
After any change to prices, policies or hours — and every month on the managed programme.
Your next buyeris askingtoday.
Find out what they’re being told now — and make sure the next one hears your name, and hears it right.
Free · two pages in two working days · one short form · no obligation