Google Shopping
Feed and landing page are matched. A price or availability mismatch is one of the most common reasons an item is disapproved.
AI shopping diagnostic for ecommerce
Is AI recommending your competitors instead of your products?
KoreLens asks AI the questions your shoppers ask, shows who each answer names and where you are missing, compares what their page exposes with what yours does, and asks the same question again once you have fixed it.
Free · No sign-up · Nothing stored unless you ask · What happens to my data? →
Free · No sign-up · Real answers, word for word — nothing simulated · Every number is measured — or it isn’t shown.
See an example ↓The gap diagnosis
When the answer names a competitor, see what their cited page has.
Most AI-visibility tools stop at “you weren’t mentioned.” KoreLens shows who was cited instead — and the exact product facts their cited page exposes that yours doesn’t.
Illustration with fictional stores — not a live model answer.
I run a store
Connect Shopify or WooCommerce. KoreLens reads the real product records and ranks what to fix first, with the values and dates behind every finding.
See what to fix first →
I run an agency
Audit a prospect before the pitch, check a whole client list in one table, and send a branded before/after report of what you fixed.
See the agency workflow →
How it works
From “AI isn’t mentioning us” to dated proof of what changed.
Ask
Test the questions shoppers actually ask AI — written from what your store sells, asked for real, stored word for word.
Find
See who each answer names and where your brand is missing — question by question, with the exact answer behind it.
Fix
See what their cited page exposes that yours doesn’t, and which product facts disagree across your own sources. Get the fix as a paste-ready block.
Verify
Re-read the live site, ask the same question again, and keep the dated before and after — including when nothing moved.
Everything else measures one half. Tools that watch AI answers never see the catalogue underneath; tools that check product data never re-ask the question. KoreLens runs a closed loop — we don’t just find the problem and call it done, we re-ask the question afterwards and report the verdict, even when it’s inconclusive.
The actual overview, with sample data from a fictional store — not a live measurement.
We measured it · 249 verified UK Shopify and WooCommerce storefronts, 16 July 2026
The UK Catalogue Truth Index →2%
of verified stores (5 of 247)
expose machine-readable product data on the homepage
2.8%
of stores (7 of 248)
block any named AI crawler — the rest are open, and illegible
33.6%
of stores (of 247)
give AI a readable returns-policy signal from the homepage
60 of 389
seeded sites (15.4%)
couldn’t be read by an identified crawler at all
Why it matters
A page your customers can read is not always a page a machine can.
Your store can be perfectly clear to a person while important product facts are hard for a machine to verify — a price drawn by a script, a returns policy written as a paragraph, a size that exists only in a dropdown.
The facts a shopping answer leans on
- Price
- Availability
- Size
- Product attributes
- Returns
- Delivery
- Identifiers
- Brand
- Category
An assistant answering a shopping question works from what it can read. A fact it cannot read on your pages is a fact it has to find somewhere else — or leave out. KoreLens shows which facts those are, next to the answers that named someone else. It reports both; it does not claim one explains the other.
What’s actually broken
This was broken long before AI arrived.
For years these contradictions quietly cost Shopping impressions and Merchant Center approvals, and nothing in the stack was watching. AI made them impossible to ignore: an assistant needs facts it can verify, or it recommends a store whose facts it can.
A disapproved item earns zero Shopping impressions — you keep paying for traffic that can’t see the product.
Merchant Center
Missing GTINs, brand, or condition hold items in review — a data problem, not a policy one.
Marketplace feeds
The same identifier fields feed every marketplace listing. One wrong identifier breaks all of them.
AI assistants
An assistant needs facts it can verify. Contradictions read as unverifiable, so it recommends someone else.
One fix clears all four, because all four are reading the same fields. What KoreLens proves is that the sources agree afterwards — never that a particular listing, impression, or recommendation follows.
For agencies · The client deliverable
The monthly answer to “what am I paying you for?”
Led by the change record: the tracked question, the answer before and after with dates, and the fix logged in between — word for word, with your agency’s name on top. Nothing to edit. One link, or the PDF.
A small “checked with KoreLens” credit line stays on reports by default — your clients should always know what ran the checks. Enterprise agencies can remove it for fully white-labelled deliverables.
Prospecting
Paste up to 25 prospect stores and get every real check in one table, top blockers named, downloadable as CSV. A site we can’t reach says so — never a made-up number.
Sample report with fictional businesses — not a live model answer.
See who AI names when your shoppers ask.
What this is — and isn’t
What we won’t tell your clients: that we can guarantee AI citations, rankings, or sales. No honest tool can — the platforms decide. What we prove is the part you control: whether AI systems can read, understand, and verify your client’s data — and exactly what changed after you fixed it.
Readiness, not placement. Listing on any AI shopping surface is approval-dependent and decided by the platform — we check whether product data is ready, and we never say a store is live or approved when it is not.
Examples marked as illustrations use fictional stores.