Google Merchant Feed Doctor

Check your Google Merchant Center product feed against Google's product data specification. For every product you get its errors, warnings and tips, each with the offending value and how to fix it. You also get a normalized feed with only safe, mechanical fixes applied, and a BEFORE/AFTER change log.

Passing these checks does not guarantee Merchant Center approval. Google also checks landing pages, images, policies and account settings, which this Actor cannot see.

Who is it for

• Merchants who upload a feed to Google Merchant Center and want to find disapproval risks before uploading.
• Agencies and feed managers who check many clients' feeds and need a repeatable, rule-by-rule report.
• Developers who generate feeds and want an automated check in their pipeline (API, schedules).

What it checks

Every rule cites Google Merchant Center Help, "Product data specification" (https://support.google.com/merchants/answer/7052112, read 2026-10-30). The rule source is included in every finding.

ERROR — required attributes and fixed formats

• Required attributes: id (≤ 50 chars), title (≤ 150), description (≤ 5000), link, image_link, availability, price.
• Fixed formats:
  • price and sale_price must be 15.00 USD style: period as decimal separator plus an ISO 4217 currency.
  • availability must be in_stock, out_of_stock, preorder or backorder; preorder needs availability_date.
  • condition, gender, age_group, adult and identifier_exists must use their enumerated values.
• GTIN:
  • check digit;
  • no restricted (02, 04, 2) or coupon (98–99) prefixes.
• Sale price:
  • sale_price currency must match price;
  • sale_price_effective_date must be in the right format and order.
• Duplicate ids: every copy is kept and flagged. The message says whether the copies are identical or different.
• Broken XML: reported with the exact line and column. This includes an unescaped &.

WARNING — conditional rules

• Identifiers: a missing GTIN/MPN without identifier_exists=no, or identifier_exists=no while a GTIN is present.
• Brand: missing for new products, or a placeholder brand such as "N/A" or "Generic".
• Sale price: a sale_price that isn't lower than price.
• Price 0.
• Apparel (by target country): missing color/size/gender/age_group. These are required in BR, FR, DE, JP, UK and US.
• Text and images: HTML in description, unsupported image formats, more than 10 additional images, colour values that are numbers or hex codes.
• Dates: availability_date more than a year ahead, expiration_date more than 30 days ahead.

TIP — non-blocking

• All-caps titles.
• Promotional text in titles.
• Missing google_product_category.

What it fixes, and what it never does

Fixes (only lossless, deterministic ones; each is logged in the change log)

• whitespace trimming;
• repeated spaces in titles;
• In Stock → in_stock (and other unambiguous availability spellings);
• New → new, TRUE → yes for yes/no attributes;
• GTIN spaces and dashes removed (Google ignores them anyway);
• 8,50 EUR → 8.50 EUR;
• USD 12.00 → 12.00 USD.

Never done

• No invented ids, brands, GTINs, prices or stock.
• No guessed values for ambiguous input. 1,299.00 USD and sold out are reported, not changed.
• No currency conversion.
• No uploads to Merchant Center.

Start in 60 seconds

1. Press Start without a feed. The Actor validates a small synthetic demo feed (fictional products on example.invalid) and shows every kind of finding. No item or report events are charged for demo runs.
2. Enter your feed's https URL, or pick a key-value store record. Set your target country and run.

Example input

{
    "feedUrl": "https://www.example.com/feeds/google-merchant.xml",
    "targetCountry": "US",
    "onlyItemsWithIssues": false
}

Example output row (SYNTHETIC demo data)

{
    "id": "NV-101",
    "title": "Nordvik Kupa Şöğüç 350 ml",
    "status": "HAS_ERRORS",
    "errors": 1, "warnings": 0, "tips": 1,
    "issues": [
        { "severity": "ERROR", "code": "GTIN_INVALID", "field": "gtin", "value": "4006381333932",
          "message": "GTIN is not valid (CHECK_DIGIT).",
          "fix": "Use the manufacturer-assigned UPC/EAN/JAN/ISBN-13/ITF-14 with a correct check digit, or leave gtin empty.",
          "spec": "https://support.google.com/merchants/answer/7052112" }
    ],
    "fixesApplied": [
        { "field": "availability", "before": "In Stock", "after": "in_stock", "rule": "AVAILABILITY_CANONICAL_VALUE" },
        { "field": "price", "before": "8,50 EUR", "after": "8.50 EUR", "rule": "PRICE_DECIMAL_POINT" }
    ],
    "dataOrigin": "SYNTHETIC_DEMO"
}

Input formats

Format	Notes
RSS 2.0 and Atom XML	g: namespace; CDATA; nested shipping and tax.
CSV and TSV	Header names like image link, image_link and g:image_link are all accepted.
.gz files	Read directly. Decompression is capped, so a "zip bomb" is refused.

Text encodings: UTF-8 with or without BOM, the HTTP charset, or the XML encoding="…" declaration (for example ISO-8859-9 for Turkish).

Output

Where	What
Dataset	One row per product: status (OK / TIPS_ONLY / HAS_WARNINGS / HAS_ERRORS), issues (severity, code, field, offending value, message, fix, spec source), fixesApplied, dataOrigin.
ISSUES_CSV	One line per finding. UTF-8 with BOM. Text formulas are neutralised; numbers stay intact.
CHANGELOG_CSV	Every fix as BEFORE/AFTER, with its rule.
NORMALIZED_FEED_TSV	Your feed as TSV with only the logged fixes applied. TSV cannot hold tabs or line breaks inside values; those become spaces, and the report counts them.
REPORT	Counts by status, severity and rule, with the spec version.
OUTPUT	Run status: SUCCEEDED, PARTIAL_RESULT or EMPTY_FEED, or FAILED with an error code. Also counts.

Error codes on FAILED:

• INPUT_INVALID
• ACCESS_DENIED
• NOT_FOUND
• RATE_LIMITED
• UPSTREAM_UNAVAILABLE
• INPUT_TOO_LARGE
• PARSE_ERROR
• UNSUPPORTED_FORMAT

Limits and performance

These figures were measured on synthetic feeds on a VPS:

Feed	Time	Peak memory
10,000 products (6.6 MB)	0.5 s	about 300 MB
100,000 products (66.7 MB)	about 7 s	about 1.3 GB

• Default memory is 1024 MB, which covers the default limit of 10,000 products. For about 100,000 products, set 2048 MB.
• maxItems defaults to 10,000 and can go up to 200,000.
• maxFileSizeMb defaults to 50 and can go up to 200.

Pricing

The Actor is not yet published, and prices are not set yet. The code is prepared for pay-per-event: one event per validated product (feed-item-validated) and one per run report (feed-report). Failed and empty feeds and demo runs charge nothing, and the spending limit is respected. These rules are tested with a simulated platform.

Security and privacy

• Feed URLs must be HTTPS. Private, internal, loopback and cloud-metadata addresses are refused, also after redirects and at DNS resolution. URLs with embedded credentials or non-standard ports are refused too.
• Download size and decompressed size are capped.
• Offending values in the output are cut to 200 characters. Secrets in URL query strings (token=, key=, sig= …) are masked.
• Limited permissions: the Actor reads only the key-value store you pick.

Known limitations and what is not verified

• Not verified on a real merchant feed. Tests used synthetic feeds and one publicly shared community example. That example is not well-formed as published; after two mechanical repairs, 6 of its 6 items passed without false findings.
• Not compared with Google's official example file. The file could not be downloaded (HTTP 403).
• Not verified that this matches Merchant Center's own diagnostics. The Actor cannot see landing pages, images, policies or account settings.
• Rules cover the attributes listed above. Shipping, tax, loyalty, subscription, energy-label and certification attributes are read but not deeply validated.
• Product category requirements (for example mobile phones or gift cards) are not checked against the full Google taxonomy.

Support

Use the Issues tab on this Actor's page.