A product feed is a structured catalog file a retailer uploads to an ad platform so the system can generate shopping ads automatically from product names, images, prices, and attributes. It replaces hand-built, one-product-at-a-time ad creative with a single data source that the platform turns into ads on demand.
If you sell more than a handful of products, you decide between manually building campaigns forever or feeding one clean catalog that scales to your whole inventory. Submit a messy or incomplete feed and ChatGPT or Google Shopping simply will not surface those products, so the items quietly drop out of every answer and results page.
Product feeds are the data layer behind shopping ads: a file, usually in the Google Shopping schema, that lists every item you sell with the fields an ad system needs. Each row carries a product's title, description, image URL, price, availability, identifier, and category, so the platform never needs ad copy from you. When ChatGPT or a search engine meets a shopping question, it pulls matching rows and assembles a sponsored card on the spot.
The feed works because it is machine-readable and complete. Required attributes have to be present and correctly formatted, images have to meet quality rules, and prices and stock have to stay current, or products are rejected while the rest keep serving. Most retailers host the file and let the platform re-fetch it on a schedule so changes flow through automatically.
A product feed sits upstream of campaigns and ads; it is the catalog they are generated from, which is why one clean feed can power both your organic product answers and your paid placements. AirOps helps teams keep the content behind those placements on brand and aligned with what AI search cites.
Resources: See how to target, format, and measure ChatGPT ads built from your catalog
A product feed moves from your store to a live ad. The platform ingests the file, checks it, matches products to intent, and serves placements.
Build the file: Export your catalog in the Google Shopping schema, with a row per product carrying title, image, price, availability, and a unique identifier.
Upload and connect: Submit the feed in the ad platform, typically by hosting it so the platform can re-fetch it on a schedule.
Validate: The platform reviews the feed for completeness, image quality, and formatting; products that fail the check are held back while valid ones move on. New partners submit a sample catalog first.
Match: When a user asks a commercial question, the system reads intent from the conversation and pairs it with eligible products, with no keywords to manage.
Serve: ChatGPT or the search engine generates a sponsored card from the product's own data and places it below the answer.
The feed tells you which products are eligible and serving, and how their data renders in an ad. It does not tell you why a product won a placement, which happens inside the platform's model.
Resources: Track where your ads and competitors appear across tracked AI search prompts
Your product feed decides whether your catalog can show up where people now shop. As buyers ask ChatGPT and Google's AI for product picks, the feed is the only way to put your full inventory into those answers without building thousands of ads by hand. The quality of that one file sets a ceiling on how much of your catalog can be recommended at all.
It reaches scale you cannot match manually: One feed turns a catalog of thousands of products into ad-ready inventory in a single step, so coverage stops depending on how many ads your team can hand-build.
Feed errors silently delete products: A missing identifier, a broken image URL, or a stale price gets a product rejected at validation, so it never serves and you lose sales you never see reported.
It aligns paid and organic in one place: The same catalog that answers organic shopping questions also powers your sponsored cards, so a fix to the feed improves both surfaces at once.
SEO managers use a product feed to keep the catalog behind organic product answers accurate, so ChatGPT and Google surface the right prices and availability.
Growth marketers use a product feed to launch catalog-scale shopping ads in ChatGPT without building a creative for every SKU.
Content strategists use a product feed to align product titles and descriptions with the shopping questions buyers actually ask.
The set of fields and formatting rules, usually the Google Shopping specification, that defines how each product's data must be structured, and reusing that shared schema is what lets a retailer send one file to several platforms without rebuilding it for each.
The mandatory fields such as unique identifier, title, price, image URL, and availability that every product must include, because a single row missing one of them is disapproved at validation and will not serve while the rest of your catalog keeps running.
How current the file stays relative to your live store, since outdated prices or stock levels cause products to be disapproved or to display wrong information in an answer, which is why most retailers let the platform re-fetch the feed on a schedule.
Turn an entire catalog into ad-ready shopping inventory without writing a creative per product.
Reuse the same Google Shopping file you already send to other platforms.
Reach shoppers inside ChatGPT at the moment they ask what to buy.
Keep organic product answers and paid cards accurate from one source.
Spot and fix rejected products before they quietly cost you sales.
Match titles to real queries: Lead each product title with the attributes shoppers name, so the platform can connect your item to a conversational question.
Fill every required attribute: Populate identifier, price, availability, and image for each row, because incomplete rows get disapproved before they can serve.
Keep prices and stock in sync: Refresh the feed on a schedule so an answer never shows a price or availability that no longer matches your store.
Use high-quality images: Supply clean, correctly sized images that pass review, since image failures are a common silent reason products drop out.
Start with a representative sample: When a platform asks for a sample catalog, send items that reflect your real data quality so the review reflects the full feed.
Segment with filters: Set eligibility filters so only the products you want to promote enter paid placements.
Avoid treating the feed as a one-time upload. Competent teams load a clean catalog, then let it drift as prices, stock, and products change, and watch eligibility quietly erode until a quarterly audit surfaces the damage.
AirOps: Keeps the titles, descriptions, and ad content tied to your feed on brand and aligned with what AI search cites, so paid and organic stay consistent.
Google Merchant Center: Hosts and validates the Google Shopping product feed many retailers reuse for ChatGPT, flagging disapproved items before they fail elsewhere.
Screaming Frog: Crawls your product pages to catch missing or inconsistent data before it reaches the feed and triggers rejections.
Locate your catalog: Find the product data you already maintain, whether in your ecommerce platform or an existing Google Shopping feed. You can do this this week with no new budget.
Audit the data: Check that every product has a unique identifier, accurate price, current availability, and a working image, and fix the gaps you find. Clean data now prevents silent rejections later.
Export in the standard schema: Generate the feed file in the Google Shopping format so it maps to ChatGPT Ads and other platforms without rebuilding. The same file then works across multiple ad channels.
Connect and validate: Upload or host the feed in the ad platform, submit any required sample, and clear every disapproval the validation flags.
Set filters and monitor: Choose which products are eligible, then review eligibility and performance regularly so a quiet rejection never goes unnoticed for long.
A product feed is a structured catalog file that lets ad platforms generate shopping ads straight from your product data.
You build it in the Google Shopping schema and connect it so platforms can re-fetch it automatically.
The feed only works when every required attribute is present, correctly formatted, and current.
A single bad field can get a product disapproved, pulling it from answers with no obvious warning.
The biggest gains come from clean titles, accurate data, and tight filters that aim spend at the products worth promoting.
A product feed is the data source; a shopping ad campaign is what runs on top of it. The feed is a structured file listing your products with their titles, prices, images, and attributes, while the campaign sets the budget, filters, and bidding that decide how those products are promoted. One feed can supply many campaigns, and it can also power organic product answers that have no paid spend behind them at all. The split is simple: the feed answers what you sell, and the campaign answers how aggressively you want to sell it. This matters operationally because feed problems and campaign problems have different fixes. If products are missing entirely, you repair the feed. If eligible products are not getting enough exposure, you adjust the campaign settings. Keeping the two concepts separate saves you from tuning bids when the real issue is a disapproved row in the catalog.
Update your product feed whenever the underlying data changes, and at a minimum keep prices and availability current every day. Most platforms re-fetch a hosted feed on a schedule, so the practical answer is to automate the sync instead of uploading by hand. Prices and stock are the fields that go stale fastest, and a mismatch between your feed and your live store can get products disapproved or show a shopper the wrong number. New or seasonal products, promotions, and price changes should flow into the feed as soon as they go live on your site. If you cannot automate yet, set a daily manual refresh for price and inventory, and a weekly pass for titles, descriptions, and images. The test is whether someone acting on your feed today would see the same prices and availability as someone on your website. If the answer drifts, your refresh cadence is too slow.
Products disappear from a feed for a short list of reasons, almost all tied to data quality. The most common is a failed validation: a missing required attribute, a broken or low-quality image, an invalid identifier, or a price that does not match your landing page gets that specific product disapproved while the rest of the catalog serves normally. A product can also be excluded by your own eligibility filters, so it is in the file but intentionally held out of placements. Beyond that, a product that is valid may simply never match a conversation, because no one asked a question its data was relevant to during the period you are looking at. The way to diagnose it is to check the platform's disapproval or diagnostics view first, since that separates a rejected product from one that is eligible but unmatched. Fix the data errors, confirm your filters, and give matching time before assuming the product itself is the problem.
Yes, you control eligibility directly through filters, and that is one of the main levers a product feed gives you. In the ad platform you can include or exclude products by category, price band, availability, custom labels, or other attributes in the feed, so only the items you want to promote enter paid placements. This is useful for keeping spend on high-margin or in-stock products and out of clearance items or things you cannot fulfill well. What you do not control is exactly which eligible product appears for a given question, because that match is made by the platform's model reading the conversation. So your influence is indirect but real: you decide the pool of eligible products and you shape how findable each one is by improving its title, description, and attributes. Clean, well-labeled data plus deliberate filters is how you steer a feed, since you are setting the boundaries and letting the system match inside them.
A good product feed is complete, accurate, current, and well-described, and you can measure it before any ad ever runs. Completeness means every product you want to sell is present with all required attributes filled in. Accuracy means the price, availability, and identifiers in the feed match your live store exactly. Currency means the data refreshes often enough that a shopper never sees a stale price. Description quality means titles and attributes use the words buyers actually say, so the platform can connect products to real questions. A practical benchmark is a near-zero disapproval rate in the platform's diagnostics, with the large majority of your intended catalog eligible to serve. If a big share of your products are disapproved or missing, the feed is underperforming regardless of how your campaigns are set up. Aim for a feed clean enough that eligibility is never the thing holding your results back.