Your product feed is not just a pipeline
a guide to Google Shopping feed optimisation

When setting up a new E-commerce website or managing an existing sites relying on the automatically generated product feed connected to Google Merchant centre can be where optimisation begins and ends.

This shouldn’t be the case – a product feed that runs without supervision is not the same as a product feed that is working. And in 2026, the gap between those two things has widened considerably.

This guide covers what a Google Shopping feed actually does, which Merchant Centre attributes are worth reviewing, common issues with a default product feed, and what changed in the specification this year.

What is a Google Shopping product feed?

A product feed is a structured file that describes every product listed on your website in a way that Google can read.

It is submitted to Google Merchant Centre, and it is what makes your products eligible to appear in Google Shopping ads, free product listings, and increasingly in AI-driven shopping results. The data is structured much like a spread sheet with each row representing a product and the columns covering attributes such as: title, price, availability, image, identifier, category and so on.

Most stores generate this automatically, either through a platform app or a built-in export. An automated product feed is time saving, but also the reason so few feeds ever get looked at.

Using a standard automated feed means that there often isn’t a point of differentiation between stores providing similar products.

An automatic feed can be automatically wrong

Automation is only as good as what it is automating.

A feed that generates itself from your product database will faithfully reproduce whatever is in that database, including old titles, missing attributes, thin descriptions and identifiers that nobody has checked since the product was first added. That is a reflection of how the store was built, which is why product feed quality and e-commerce website development are harder to separate than most people expect.

This is the part that catches people out. Product feed problems rarely announce themselves. Products do not usually disappear. They show less often, in fewer places, against fewer queries, and the drop is gradual enough that it reads as market conditions rather than a data problem.

By the time it shows up in a monthly report, it looks like a bidding issue. It usually is not.

Google Merchant Centre product feed requirements

Before optimising anything, it helps to know what Google actually requires.

According to the Merchant Centre product data specification, every product needs an ID, a title, a description, a link, an image link, availability, price, and a condition value if the item is used or refurbished. Brand is required for new products, with the exception of movies, books and musical recordings.

Those are the minimums. Meeting them gets your products accepted. It does not make them competitive.

The gap between an accepted feed and an optimised feed is where nearly all the available opportunity sits, and it is a gap most catalogues never close.

Product identifiers: GTIN, brand and MPN

Google’s position on product identifiers is not ambiguous.

For products with assigned identifiers, Google’s documentation on unique product identifiers states plainly that you should “always include them to ensure best performance.” Products that have a GTIN but do not submit one may see limited visibility. The same applies to brand and MPN data that exists but is not passed through.

Google also puts a number on it. Retailers submitting correct GTINs saw a 20% increase in clicks on average. That is a rare piece of specificity from Google, and it is worth taking seriously.

The reverse matters too. Google warns: “Don’t make up, guess, or include values from similar products.” Inaccurate identifiers can lead to products being disapproved outright, which is a worse outcome than leaving the field empty. If a product genuinely has no manufacturer identifier, the correct action is to omit the attribute rather than invent something plausible.

This is where a lot of well-intentioned feed work goes wrong. Someone notices the empty GTIN column, fills it with internal SKUs to make the warnings go away, and quietly creates a disapproval risk in place of a visibility one.

If you manufacture your own products, the rule is different again. Use your store name as the brand and assign your own MPN.

Product title optimisation is where the gains are

If you only fix one attribute, fix your titles.

The title has a limit of 150 characters, but Google notes that users typically see only the first 70 characters. Everything that matters needs to be at the front.

Google’s own guidance is to “include important attributes in your title, such as strong brand names, age group, gender, size, colour, size type, or personalisation options,” and to “put key details at the front of your product title, especially if they’re not visible in the image.”

A workable structure for most retail categories:

Brand + Product Type + Key Attribute + Size or Colour

So “Patagonia Down Parka, Waterproof, 600-Fill, Hooded, Black, Medium” rather than a title assembled from every term you would like to rank for.

Things to keep out of titles: promotional text, pricing, sale dates, shipping or delivery information, your company name, all caps, symbols and HTML tags. All of these are explicitly discouraged, and some will get products disapproved.

The same discipline applies in organic search, which is why we built a free tool for checking title and description pixel widths. Character counts and visible widths are not the same thing.

Product type and Google product category are not the same field

These two get confused constantly, and the confusion is costly.

Google product category is Google’s own taxonomy. You pick one value from Google’s predefined list. It tells Google what kind of thing your product is, in Google’s language.

Product type is your taxonomy. You define it yourself, up to 750 characters, and you should submit the full category path rather than an abbreviated version. It is also what you use for campaign structure and bidding, which makes it quietly one of the most useful fields in the whole feed.

Google recommends categorising at least two to three levels deep. “Apparel > Outerwear > Jackets > Down Jackets” does considerably more work than “Jackets.”

Getting this right improves both how Google understands the product and how well you can segment campaigns, which is why it sits at the intersection of feed work and Google Ads campaign management.

Product images and the 2027 resolution change

Images do more work in Shopping than anywhere else in paid search, because the image is the ad.

Google’s advice is direct: “Provide the highest resolution image you have available. Products with high quality images are more likely to show for relevant search queries.” Placeholder images, promotional overlays, watermarks and logos are all discouraged.

There is also a deadline worth diarising. The minimum resolution for image link and additional image link is moving to 500 x 500 pixels, with warnings having started in April 2026 and enforcement from 31 January 2027. Google will auto-optimise smaller images where it can, but that is a fallback rather than a plan.

If your catalogue includes legacy product photography, that is a check worth doing now rather than discovering in January.

Availability, price and variants

These three cause more disapprovals than anything else, and almost all of them are avoidable.

Availability accepts four values only: in stock, out of stock, preorder and backorder. Price must be numeric with a valid currency code. Both must match what appears on the landing page and at the checkout. A mismatch between feed price and checkout price is one of the fastest ways to get products suspended.

Variants are grouped using item group ID, up to 50 characters, which is required for variants in most countries and for free listings. Each variant still needs its own distinguishing attributes.

Price and availability mismatches often point at something worth fixing on the site itself rather than in the feed, which is where conversion rate optimisation and feed work start to overlap.

Product feed enrichment: the attributes most stores ignore

Everything above is table stakes. This is where the separation happens.

There is a long tail of optional attributes that almost no automatic feed populates, and that Google uses to understand products more precisely:

• Product highlight. Short benefit statements, up to 150 characters each, between two and 100 per product.
• Product detail. Technical specifications structured as section name, attribute name and value.
• Additional image link. More angles, more context, more chance of matching a query.
• Custom labels. Five fields you define yourself, invaluable for campaign segmentation by margin, season or performance tier.
• Sale price and sale price effective date. Enables promotional annotations rather than just a lower number.
• Shipping and shipping weight. Accurate delivery information visibly affects click-through.

Feed enrichment is unglamorous, and that is precisely why it works. In most categories, the competitor you are up against is running a default export.

How AI has changed what your product feed needs to say

Here is what has genuinely changed.

Google’s Shopping Graph now holds more than 50 billion product listings, with 2 billion updated every hour. That catalogue increasingly feeds AI-driven surfaces rather than a straightforward ranked list of shopping ads. It is the same shift we wrote about when Google began changing how AI answers search queries, now applied to products.

When someone asks for a warm waterproof jacket for a Wellington winter under $400, the system answering that is not matching keywords. It is assembling a recommendation from structured product data.

A person can look at a photograph and reasonably conclude that a jacket looks warm. A model cannot. It needs the fill power, the shell material and the temperature rating stated explicitly, in the feed, as attributes.

Google has built for this directly. It now offers a set of what it calls conversational attributes, described as helping “customers discover information about your products across AI-driven surfaces, like AI Mode in Search.” These include question and answer, document link, related product, item group title, variant option and popularity rank.

There is also a disclosure requirement that has gone largely unnoticed. If you generate product titles using AI, Google requires them to be submitted through the structured title attribute with a digital source type of trained algorithmic media, rather than through the standard title field. Anyone running AI-generated feed content at scale should check how their tool is submitting it.

The old logic still holds, but with higher stakes. A customer may forgive a vague product title if the image is clear. A recommendation engine deciding whether to surface your product at all will not.

What changed in Google Merchant Centre in 2026

Three things worth knowing.

Conversational attributes arrived. Google added dozens of new data attributes aimed at AI Mode and conversational shopping experiences. Almost none are populated by a default feed.

New delivery and content attributes. As of 14 April 2026, the specification added handling cutoff time and minimum order value, plus loyalty programme sub-attributes. A video link attribute also arrived, with submitted videos becoming eligible to serve from 30 June 2026.

Image requirements are rising. The 500 x 500 pixel minimum, enforced from 31 January 2027, as covered above.

None of these require a rebuild. They require somebody to look.

Supplemental feeds and feed rules

A common objection to feed work is that the source data is hard to change. Often it genuinely is.

This is what supplemental feeds are for. A supplemental feed lets you add or override specific attributes without touching the primary feed or the underlying database. You can enrich titles, add custom labels, fill in missing categories or patch identifiers, all without a developer ticket.

One important limit: supplemental feeds can only update existing products. They cannot create new ones.

Feed rules sit alongside this, applying transformations inside Merchant Centre itself. Between the two, most feed problems can be solved without re-engineering the store.

That matters commercially. It means product feed optimisation is usually a marketing project rather than a development project, and it can start this week rather than next quarter.

Shopify, WooCommerce and platform feeds

Most feeds in New Zealand are generated by an app or plugin rather than built by hand.

Shopify’s Google channel, WooCommerce feed plugins and BigCommerce exports all produce a valid feed quickly. They also all make the same trade. To work for every store, they map fields generically. Product title usually becomes the store’s product name. Description becomes the full product description, HTML and all. Product type often maps to a single collection.

Valid, but generic. The defaults are a starting point, not a configuration.

The practical approach on any platform is the same: let the app generate the feed, then use supplemental feeds and feed rules to enrich the attributes that matter, rather than fighting the platform.

A product feed review checklist

If you do nothing else, work through this once a quarter.

• Titles. Key attributes in the first 70 characters. No promotional text, pricing or company name.
• Identifiers. GTIN, brand and MPN. Correct, or absent. Never guessed.
• Google product category. Assigned, and specific.
• Product type. Your own taxonomy, full path, two to three levels deep.
• Images. Above 500 x 500 pixels. No watermarks, overlays or placeholders.
• Availability and price. Matching the landing page and the checkout.
• Variants. Item group ID applied, distinguishing attributes present.
• Descriptions. Matching the landing page. No links or promotional language.
• Enrichment. Product highlights, product details and custom labels populated.
• Diagnostics. Merchant Centre warnings reviewed, not just errors.

That last one matters more than it looks. Most accounts clear errors and ignore warnings. Warnings are where the visibility is leaking.

Google Shopping feed optimisation: frequently asked questions

How often should a product feed be updated?

Price and availability should update at least daily, and more often for fast-moving catalogues. Structural attributes like titles and categories are worth a full review quarterly.

Do I need a GTIN for every product?

No. You need one for every product that has one assigned by the manufacturer. Products without assigned identifiers should omit the field rather than have a value invented for them.

Why are my products disapproved in Merchant Centre?

The most common causes are price or availability mismatches between the feed and the landing page, inaccurate identifiers, image policy breaches, and missing required attributes. Merchant Centre diagnostics will name the reason, though not always the underlying cause.

Does product feed optimisation help free listings as well as ads?

Yes. Free product listings draw on the same Merchant Centre data, which makes the feed one of the few places where organic and paid visibility share a direct dependency.

Can I optimise a feed without changing my website?

In most cases yes, using supplemental feeds and feed rules inside Merchant Centre. Some fixes are better made at source, but a great deal can be done without touching the store.

What this says about how we work

We manage paid search and Google Shopping campaigns regularly, across a range of catalogue sizes and platforms.

That experience means we tend to look at the data layer before we look at the bidding. When shopping performance softens, the instinct in most accounts is to adjust budgets and targets. Often the actual cause is sitting in the feed, in a field nobody has opened since the product was created.

It is the same diagnostic habit we apply when Google Ads clicks start dropping, and the same reason we validate unusual data before acting on it. Understand what has changed before deciding what to do about it. That principle sits at the core of any effective digital marketing strategy.

Product feed maintenance is as much a marketing decision as a technical one. It just happens to be one that lives in a spreadsheet.

Need help with Google Shopping feed optimisation?

If your Google Shopping performance has drifted and the budget has not changed, the problem may not be your campaign structure.

It may be what your product feed is telling Google about your products.

At FizzyPop, we help e-commerce businesses audit their Merchant Centre data, fix what is limiting visibility, and build product feeds that hold up as Google’s shopping surfaces continue to change. You can see examples of the work we do across retail and e-commerce clients, or meet the team who would be doing it.

If you want a Google Shopping partner who understands both the advertising and the data behind it, get in touch with our team today.