SendTilly asks AI assistants the questions your shoppers ask, on your own website, and compares their answers with what your site shows a person. You see where they get it wrong, who is at fault, and what to check.
Base layers / TopsMerino Base Layer€89Choose a storeChoose a store to see stockAdd to cart
SAMPLE In a sample check, store stock appeared only after someone picked a store, so 12 of 19 quick looks found no answer.
HOW IT WORKS
From a shopper's question to a clear report.
01 ASK
Ask what your shoppers ask.
Is this product in stock at this store, and can I collect it today? SendTilly fills in your products and stores, and puts the question to AI assistants from two vendors.
02 ANSWER
Watch how the AI finds an answer.
In agent mode the AI clicks through your site, like a shopper. In fetch mode it takes one quick look at the page, the way a chat answer does. Every step is recorded.
03 CHECK
Grade it against your own page.
A fixed recipe reads your store page, with no AI involved, at the start and at the end of each check. If stock changes in between, the result is set aside, not blamed on you.
04 ATTRIBUTE
Know who got it wrong.
Every failure has one owner: your site, your bot protection or the AI. A store failure counts only when AI assistants from two vendors fail the same way, so an AI's own mistake never counts against you.
05 REPORT
Hand your developer what to check.
Each finding comes with screenshots, steps to reproduce and what to check on your site. A person reviews the report before it reaches you. After a change, ask us to run the same missions again.
Following Anthropic · Bergen, size M
01ASKED
Missions
20 missions · 2 vendors
QUESTION SENT TO EVERY AI
Is Ridge Shell Jacket, size M, in stock at Fjellsport Demo Bergen, and can I collect it today?
ANAnthropic Agent modeopened the product page from searchclicks48s
OPOpenAI Agent modeopened the product page from the menuclicks1m 02s
ANAnthropic Fetch modetook one quick look at the pagelooks2s
OPOpenAI Fetch modetook one quick look at the pagelooks3s
GTReference reading A fixed recipe · no AIpicked Bergen in the store picker and read the stock lineno AI6s
02ANSWERED
Journey
Anthropic, agent mode · start to answer
goto /product/ridge-shell-jacket
read page data: Offer availability InStockINSTOCK
read button "Collect in store"
scroll to "Stores", store picker not opened
read store list: Bergen open until 20:00
finish: in stock, collect todayWRONG ANSWER
ANSWER
In stock
COLLECT TODAY
Yes
SAME ANSWER FROM
2 vendors
03CHECKED
Checked against the store page
Read twice, at the start and the end
AN
WHAT THE AI ANSWEREDAnthropic and OpenAI
STOCK
In stock
COLLECT TODAY
Yes
STORE
Bergen
GT
WHAT THE STORE'S PAGE SHOWSReference reading
STOCK
Out of stock
COLLECT TODAY
No
STORE
Bergen
The page shows "Out of stock at this store" at 09:14:07 and again at 09:21:40.
04ATTRIBUTED
Verdict
Bergen · Ridge Shell Jacket, size M
WRONG ANSWER2 vendors agree2 readings agree
Page data says in stock for every store
WHO IS AT FAULT
STOREThe page shows one thing and its data says another.
ACCESS LAYERThe site refused or challenged the AI.
AGENTThe AI failed on its own. Not a finding.
Observed: 3 of 18 missions got this wrong answer. A shopper who asks an AI assistant is told to come in for a size that is not on the shelf.
05REPORTED
What to check
The store's site
Page data says in stock for every store
WHAT TO CHECK
Give each store its own offer in the page data, with that store's availability, and keep the online offer separate.
REPRODUCE
Open the product page for the jacket and choose size M.
Read the page data without choosing a store. It says InStock and names no store.
Choose "Fjellsport Demo Bergen" under "Choose a store".
The stock line reads "Out of stock at this store".
FJELLSPORT DEMO · SAMPLE CHECKEXAMPLE DATA
THE REPORT
Evidence your team can act on.
One private report: the overview for the person who decides, and the detail for the developer who checks.
RESULTS BY PRODUCT AND STORE
Every product at every store, at a glance
EXAMPLE DATA
PRODUCT
OSLO
BERGEN
TRONDHEIM
STAVANGER
TROMSØ
PASSED
Ridge Shell Jacket size M
OsloPASS
BergenWRONG ANSWER
TrondheimWRONG ANSWER
StavangerPASS
TromsøNOT GRADED
2 of 4
Merino Base Layer size L
OsloPASS
BergenPASS
TrondheimPASS (stock only)
StavangerPASS
TromsøPASS
5 of 5
Trail Hiking Boot size 43
OsloPASS
BergenAGENT ERROR
TrondheimWRONG ANSWER
StavangerPARTIAL
TromsøPASS
2 of 4
Frost Beanie one size
OsloPASS
BergenPASS
TrondheimPASS
StavangerPASS
TromsøPASS
5 of 5
PASS
WRONG ANSWER
PARTIAL
AGENT ERROR
NOT GRADED
WHO THE FAILURES BELONG TO
Every failure has an owner
EXAMPLE DATA
THE STORE'S SITE
16
failures in 3 findings
THE ACCESS LAYER
2
failures in 1 finding
THE AI
1
its own error, not counted against the store
ANSWER VS PAGE
What the AI said, beside what your page shows
EXAMPLE DATA
AN
THE AI2 vendors
STOCK
In stock
COLLECT
Today
GT
YOUR PAGEread twice
STOCK
Out of stock
COLLECT
No
FINDINGS
How far each problem reaches
EXAMPLE DATA
WRONG ANSWERAgent modeAttributed to THE STORE'S SITE
Page data says in stock for every store
3 of 18Same failure on both first attempts in 2 missions
AI SESSION
The steps the AI took
EXAMPLE DATA
goto /product/ridge-shell-jacket
read page data: InStock
store picker not opened
finish: in stock, collect todayWRONG ANSWER
WHAT TO CHECK
Where your developer starts
EXAMPLE DATA
Give each store its own offer in the page data, with that store's availability, and keep the online offer separate.
REPRODUCE
Open the product page, size M.
Read the page data without choosing a store.
Choose "Fjellsport Demo Bergen". The stock line reads "Out of stock at this store".
Local stock check: is it in stock, and can I collect it today?
Products and stores agreed with you
Both modes, with AI assistants from two vendors
Private report with findings, evidence and what to check
Reviewed by a person before it reaches you
FAQ
Questions, answered.
Do I need to install anything?
No. SendTilly visits your public website the way a shopper would. There is nothing to add to your site and no access to give.
Will it slow down or harm my site?
It makes one visit at a time per site, at times agreed with you, and it stops before any payment step. Nothing is bought.
Which AI assistants do you use?
Models from Anthropic and OpenAI, called through their APIs. In agent mode they work in SendTilly's own browser. They behave like AI shoppers. They are not the ChatGPT or Claude apps, and the report says so.
What if the AI makes a mistake?
Then it is the AI's mistake. A store failure counts only when AI assistants from two vendors fail the same way, and an AI's own errors are listed apart and never counted against you.
What if my bot protection blocks it?
That is a finding too. The report shows when your site turned an AI assistant away, and which AI crawlers your robots.txt has rules for.
Who sees my report?
Only the people you send the private link to. Nothing is published.
Your next shopper may ask an AI first.
Find out what it tells them about your stores, before they make the trip.