Why Google's newly documented auto-categorization should change how you manage your product feed — and how GoDataFeed's AI Categories puts you back in control.

Google's Merchant Center documentation states it plainly: “All products are automatically assigned a product category from Google's continuously evolving product taxonomy.” And in July 2026, Search Engine Roundtable reported that Google added the same disclosure to its Google Ads help documentation — spelling out, where advertisers live, behavior that had only been documented on the Merchant Center side for years. Every product you sell has already been filed into one of the thousands of nodes in Google's taxonomy, whether you chose that category or not.

The same documentation hands merchants the remedy: the google_product_category attribute “can be used to override Google's automatic categorization in specific cases” — Google explicitly accepts overrides to correct category-specific attribute requirements, to reassign products within your Google Ads campaign structure, and to keep compliance-sensitive categories like alcohol classified correctly. In other words, your classification can take precedence — but only if you actually submit it.

Most merchants never have. Here's why that's a problem, and how to fix it at scale.

Why Auto-Assigned Categories Quietly Cost You Money

Product categories look like metadata, but they behave like infrastructure. The category Google assigns to each product influences how that product gets matched to shopper searches, drives the category insights you see in Google Ads, and feeds the signals Smart Bidding optimizes against. When a product lands in the wrong bucket, the damage compounds in ways that rarely trigger an alert:

• Mismatched search intent. A phone case categorized as luggage doesn't surface for the shoppers actually looking for it. Impressions drop, and the impressions you do get convert poorly.
• Noisy reporting. Category insights become unreliable when the categories themselves are wrong. You end up making merchandising and budget decisions on distorted data.
• Misdirected bidding. Smart Bidding follows the signal it's given. A miscategorized product trains automated bidding on the wrong competitive set and the wrong benchmarks — quietly, continuously.
• Compliance exposure. Some categories carry specific requirements. A wrong auto-assignment can attach obligations to products that don't warrant them, or hide requirements your products actually need to meet.

And because Google's taxonomy is “continuously evolving,” this isn't a one-time audit. A mapping that was correct last quarter can drift as Google adds, splits, and renames category nodes. The problem isn't just fixing miscategorization — it's keeping thousands of SKUs correctly categorized, on every channel, indefinitely.

The Manual Approach Doesn't Scale — and Never Did

The traditional fix is a spreadsheet: export your product types, VLOOKUP them against Google's taxonomy file, paste the results into your feed. Anyone who has actually done this knows how it goes. Catalogs grow. Channels shift requirements. Google's taxonomy updates. Amazon's taxonomy is structured completely differently from Google's, so multichannel sellers maintain parallel mappings that drift out of sync. One miscategorized group can disapprove hundreds of SKUs and crater impression share before anyone notices.

Checking what Google decided your products are takes minutes. Correcting it across a 20,000-SKU catalog, per channel, and keeping it corrected — that's the part that breaks manual workflows.

How GoDataFeed's AI Categories Solves This

GoDataFeed built AI Categories to make correct, channel-specific categorization the default state of your feed rather than a recurring project. Here's what it does differently.

It reads your products the way a merchandiser would. The AI engine analyzes your product_type, titles, and descriptions and cross-references them against each channel's taxonomy to assign the most accurate node. It's contextual, not keyword matching: a category group called “Cases” gets read as electronics accessories when the SKUs mention iPhone and Samsung, and as travel gear when they mention Samsonite and luggage.

It's channel-aware by design. Google files throw pillows under Home & Garden > Decor > Decorative Pillows; Amazon wants Home & Kitchen > Bedding > Decorative Pillows, Inserts & Covers. AI Categories maps to each channel's native taxonomy simultaneously, so one catalog produces correct classifications everywhere you sell.

You stay in control. This is a human-in-the-loop system, not a black box. Bulk-confirm the suggestions that are obviously right — typically the vast majority — regenerate anything that looks broad or off, and manually override from the full taxonomy tree when you know exactly which node you want. Google's auto-categorization gives you no such review step; AI Categories is built around one.

Rules handle the exceptions. Category Rules layer if-then logic on top of AI mappings: IF title contains “Boots,” THEN set google_product_category to Apparel & Accessories > Shoes > Boots. The broad mapping stays intact while predictable exceptions land in their most specific node — exactly the granularity that sharpens search matching and bidding signals.

You can verify before you submit. Compile the feed and preview the final, SKU-level value in the channel's category field before anything ships to Google. No guessing what the channel will receive.

The Payoff You Can Measure

The clearest scoreboard is Google Merchant Center's Diagnostics tab. After implementing AI Categories, invalid google_product_category errors should drop to near zero — a direct input to product eligibility and Shopping visibility. Downstream, correct categories mean cleaner category insights, bidding signals that reflect your actual competitive set, and products that surface for the searches they belong in. Work that used to consume days of spreadsheet maintenance compresses into minutes of review.

The Bottom Line

Google's documentation didn't change how categorization works — it confirmed what has been true for years: if you don't classify your products, Google will, and its guess becomes the foundation of your Shopping performance. Google also documents the lever to take that decision back. The google_product_category attribute lets you override the automatic assignment; GoDataFeed's AI Categories makes exercising that control practical at catalog scale, across every channel, continuously.

And this extends well beyond Google. AI Categories supports the native taxonomies of Amazon, TikTok, Facebook, Instagram, eBay, and Pinterest — so the same catalog gets correctly classified, in each channel's own category structure, everywhere you sell.

Go look at what Google decided your products are. Then decide for yourself — in minutes, not days.

Ready to retire the mapping spreadsheet? Start a 14-day free trial of GoDataFeed or book a demo for a complimentary feed setup — and see your catalog correctly categorized before your competitors check theirs.