Google Merchant Center
Google publishes which attributes, in writing
Not a guess. A specification, attribute by attribute, of exactly what a product feed should carry.
waxed canvas messenger bag with a padded laptop sleeve
Field Messenger
✓
material
waxed canvas
“12oz waxed canvas”
✓
compartment
padded laptop sleeve
“fits a 15in laptop”
a merino crew in a US medium
Ribbed Merino Crew
✓
material
merino wool
“100% Merino Wool”
✕
size system
US
the copy names no system
fine ribbed knit in charcoal
Ribbed Merino Crew
✓
pattern
ribbed
“a fine ribbed knit”
✓
colour
charcoal
“Crew Neck, Charcoal”
Shoppers stopped browsing.
They started asking.
Someone is looking for something inparticular. Make sure the answer is you.
waterproof jacket that fits tall and packs down small
Storm Shell
✓
waterproof rating
20,000mm
“20k/20k membrane”
✓
fit
tall
“also cut in Tall”
?
packed size
unknown
not in your copy
grain-free food for a senior small breed
Senior Small Breed
✓
life stage
senior
“for dogs 7+”
✓
diet
grain-free
“Grain Free”
satin brass door handle for a 44mm door
Lever on Rose
✓
finish
satin brass
“in Satin Brass”
✕
handing
left/right
the copy does not say
The offer
Your first 100 products, enriched and cited. Free.
Give us your store URL. We read your catalogue, complete and source the first 100, and hand you the file, every value linked to where it came from. No account, no integration, nothing written to your store.
How you get found
We make your product the answer.
- 1
The real question
Not a keyword. A full sentence with four conditions stacked in it, typed by someone who has already decided to buy.
- 2
We make your product the answer
The answers might be written in your copy, or we source and cite them. Never inferred, never guessed.
- 3
Because the data is there now
Every condition met by a value we matched to your copy or sourced and cited. That is what makes the match real rather than lucky.
bags & accessories
waxed canvas messenger bag with a padded laptop sleeve and a suede base
✓materialwaxed canvas✓compartmentpadded laptop sleeve✓basesuede
beauty
cruelty-free fragrance-free retinol serum for sensitive skin
✓activeretinol✓free fromfragrance✓suitable forsensitive skin✓certificationcruelty-free
DIY & hardware
replacement door handle, 72mm centres, satin brass, fits a 44mm door
✓centres72mm✓finishsatin brass✓door thickness44mm✕handingleft/right
pet
grain-free senior dog food for a small breed with a sensitive stomach
✓life stagesenior✓breed sizesmall✓dietgrain-free?sensitive stomachunknown
What you get
Matched to your copy, or sourced and cited. Never generated.
Recover
The value is already in your copy. In the title, in a tag, halfway down the description. We pull it into the structured field a platform looks at, and cite the line it came from.
Source
Where your copy is silent, we find the fact from a primary authority. Google's product taxonomy, GS1, the brand's own specification. It goes into the field with a link straight to the authority, sourced, not written.
Prove
Every value walks back to its origin in the graph, your words or the source. You, Google, or the shopper's AI can check any one of them.
Where neither your copy nor an authority carries the value, the field stays empty and we tell you it is empty.
Conversational commerce
The shopping layer is changing.
Google Merchant Center
Conversational attributes arrive for AI shopping
A 2026 set built so AI surfaces can answer a shopper's real question rather than match a keyword.
Google Merchant Center
GTIN is the strongest match signal there is
Google reports limited performance when it is missing, and defines the check digit as a mathematical check. Populated is not the same as valid.
Limited performance due to missing value: GTIN ↗ About unique product identifiers ↗
Google Merchant Center
Bad data doesn't underperform. It gets removed.
Data-quality violations disapprove products outright.
Fixing Merchant Center disapprovals for product data quality violations ↗
OpenAI Developers
OpenAI runs its own product feed spec
It asks for the same clean, identified data the Shopping Graph rewards. One asset, every assistant.
Shopify Developers
Shopify's catalog now feeds shopping agents
Commerce agents discover products through Shopify's Catalog and complete checkout for the buyer, discovery to order. What an agent reads is the structured product data underneath.
Every link goes to the platform's own documentation, for the same reason every value we ship carries its citation. A claim you cannot check is not evidence.
Why we created this
The market is preparing itself for AI shopping.
The way people shop is changing fast. Shopify says AI-driven traffic to its stores tripled in a year, and the orders coming through those AI channels tripled with it. The retailers who've prepared are the ones those AIs can read.
3x
There's a gap, and we measured it. The largest retailers have structured their product data, the colour, the size, the fields a machine reads, at more than four times the rate of independent stores. Nine in ten independents haven't done it. Their answers are written in their descriptions, where a shopper's AI can't match on them.
Largest retailersIndependent stores
Clear the 80% completeness bar38%9%
Variants with a machine-readable colour88%44%
Variants with a machine-readable size91%49%
Variants with a machine-readable barcode87%17%
Read from each store's own public product feed, the same endpoint and the same per-variant unit on both sides. The largest retailers also file to private feeds we cannot read, so their side is understated here.
That gap exists for one reason. Structuring a catalogue used to mean enterprise tools and enterprise budgets. The big players could afford it. Independents couldn't. So we built inparticular to read the words you already wrote, put them in the fields the machines read, prove every value back to its source, and do it at a price a small store can actually say yes to. So you have the same shot they do, before the season the whole market is watching.
The market is preparing itself for AI shopping. When a shopper asks an AI for something inparticular, only complete data is in the running. If yours is one of the stores that has not, your answers are already written in your copy, just not in the fields the machine reads.
A live example
This is what you get back.
Click a product to open its values. Click a value to walk one hop to what proved it. A value found in the copy walks back to the words that carried it. A sourced value walks to the authority that published it.
product
found in the copy
sourced
refused
unknown
source
Live sample, a real run of 40 products
the graph loads when you reach it
A real run of 40 products, 722 values, each one walkable to its source. Your report does this for your whole catalogue, on your exact words.
How this compares
Four ways to fill the gaps. One of them shows its working.
| inparticular | Manual or freelance | A tool you operate | A self-serve AI app | |
|---|---|---|---|---|
| Evidence attached to every shipped value | Yes | Varies | Tool-dependent | Typically not |
| Unsupported values explicitly refused | Yes | Varies | Tool-dependent | Typically not |
| Source is inspectable | Yes | Varies | Tool-dependent | Typically not |
| No write-back without approval | Yes | Varies | Varies | Typically not |
| Done for you | Yes | Manual | Typically not | Typically not |
| Priced by product | Yes | Varies | Typically not | Varies |
What we measured
Every run on the ledger, added up.
100%of the values shipped can be traced to the merchant's own words. Traceability, not a score
42,359values shipped globally, each one cited
2,263products enriched and cited globally
What it costs
100 free, twenty cents after, at any catalogue size.
| How many products | Rate |
|---|---|
| First 100 products | Free |
| Everything after | $0.20 per product |
| Minimum charge | $10.00 |
The first 100 are free and yours to keep, with no obligation to buy the rest. We do them as a working example of our quality, so you can see it before you decide anything.
All prices USD, inc. VAT. Per product, never per SKU or variant. Every variant of every product is included.
Enrich my first 100 free Your store URL and an address for the file. Nothing else asked.
heavyweight brushed-back sweat, made in Portugal
Yard Crew
✓
fabric weight
260gsm
“260gsm brushed back”
✕
country of origin
Portugal
“designed in Portugal”
machine washable wool blanket
Croft Wool Throw
✓
material
lambswool
“100% lambswool”
?
care
unknown
not in your copy