Google Shopping Feed Optimisation
Google ranks and matches products based entirely on feed data. A dirty feed does not just underperform — it produces policy violations, disapprovals and wasted ad spend. CatyAI fixes the data itself, at scale, before it reaches Google.
- Google-approved Comparison Shopping Service
- Anthropic Claude Partner Network member
- OpenAI Partner Network — onboarding in progress
The problem with merchant feeds
Most product feeds are broken at the source. Not slightly imperfect — structurally broken. Across the affiliate and merchant feeds we process, we routinely find:
Feeds that send price: 1 for unavailable products, producing nonsense listings and fake discount badges.
Titles missing diacritics in Romanian feeds, inconsistent currency and decimal formats.
Missing GTINs, MPNs used as GTINs, brand fields containing legal entity names instead of brands.
Volume and size values extracted as product codes, poisoning deduplication.
Products listed as in stock days after the merchant removed them.
Google Shopping ranks and matches products based entirely on feed data. A dirty feed does not just underperform — it produces policy violations, disapprovals, and wasted ad spend.
What CatyAI actually does
CatyAI is a data-quality pipeline for product feeds. Not a labeling tool on top of campaign metrics — a system that fixes the data itself, at scale, before it reaches Google.
AI enrichment at scale
We process raw merchant feeds through an AI enrichment pipeline that rewrites and completes product data: clean titles with correct regional characters, structured attributes, accurate categorization, normalized brands. Over 3.5 million products processed to date, with a measured acceptance rate of ~95% against our validation rules.
Validation, not trust
Every enriched product passes through a versioned rules engine. Products that fail — hallucinated attributes, invalid structure, regional mismatches — are rejected or flagged, never silently published. Rules are versioned, so re-eligibility is deterministic: when a rule changes, affected products are automatically re-evaluated.
Golden records and deduplication
Products from multiple merchants are matched into single golden records. This is what makes real price comparison possible — and it requires the cleaning steps above, because matching on dirty identifiers produces garbage.
Feed hygiene at ingest
Placeholder prices, missing images, out-of-stock listings and non-creditable offers are filtered at ingestion, not patched later. A product that cannot be honestly listed does not enter the feed.
Continuous measurement
Enrichment quality is not assumed — it is measured on sampled cohorts with automated abort thresholds. If a processing batch degrades below the quality gate, the system stops itself before it burns budget on bad output.
Why this matters for Google Shopping
Complete, accurate attributes are how Google matches your products to queries.
Misrepresentation and data-quality disapprovals almost always trace back to feed defects — placeholders, stale availability, wrong identifiers.
Clean golden records are the prerequisite for any real price-comparison experience.
Who builds this
CatyAI is operated by the team behind PayAI-x, operating a Google-approved Comparison Shopping Service serving a network of 176 merchants. The pipeline was built to solve our own feed-quality problems first — the tooling is the byproduct of operating a CSS, not the other way around.
The clean source of truth that produces Google-ready feeds also produces Meta-ready catalogs — Facebook dynamic ads, Advantage+ and Instagram Shopping, from the same validated data.
Facebook & Instagram Shopping Feed Optimisation →Book a partnership call
Write to us at contact@catyai.io or call +40 756 730 193 — we'll discuss your catalog and the right package.