What does an AI-ready product feed cost to build and keep accurate?
ChatGPT Shopping now leans on product feeds. The three ways to get your catalogue in, the people time that keeps prices and stock right, and what to avoid.
The short answer
For most D2C brands the build is the cheap part. A store on a platform with a native integration can connect in days; a feed partner or a direct feed takes longer. The real cost is the weekly people time that keeps prices, stock, variants and shipping right, because a stale feed now decides whether AI shopping answers show you.
Key takeaways
- In Profound's data from 687 customers (1 July–24 August 2026), feed-integrated retrieval rose from 8.26% to 61.54% of ChatGPT Shopping picks on 10 July.
- Comparing 7–9 July with 10–12 July, the top ten merchants' share of picks rose from 22.5% to 41.8%, and the number of unique merchants fell from 13,524 to 10,607.
- There are three routes into AI shopping feeds: a platform's native integration, a feed partner, or a direct feed; the first is usually cheapest.
- Ongoing accuracy costs more than the build: price, stock, variants and shipping must match your site every time the feed refreshes.
- Be wary of retainers for switching on a platform integration or for keyword-stuffing product titles.
In this article
Why the feed became the cost question
ChatGPT Shopping moved its product picks to feeds almost overnight, so the feed is now the cost that decides visibility. Profound, which sells AI visibility software, tracked ChatGPT Shopping for 687 customers between 1 July and 24 August 2026. On 10 July, the share of picks retrieved through integrated product feeds jumped from 8.26% to 61.54%.1
The shift concentrated the shelf. Comparing 7–9 July with 10–12 July, the top ten merchants' share of picks rose from 22.5% to 41.8%, and the number of unique merchants appearing fell from 13,524 to 10,607.1 Profound links that concentration to the switch towards feed-integrated merchants.
We flagged the switch in our note the week it was reported. This post answers the follow-up question D2C teams ask us: what does it cost to get in, and to stay accurate once you are in?
ChatGPT Shopping, 1 Jul–24 Aug 2026
Feeds took over the shelf
Three routes into AI shopping feeds
You can get a catalogue into ChatGPT Shopping through your commerce platform, through a feed partner or with a direct feed. Search Engine Journal's report on Profound's data summarises the routes: Shopify and Etsy sellers are integrated automatically, and other merchants join through a waitlist or through feed partners.2 The same report quotes OpenAI saying that product results are not ads and are not influenced by any OpenAI partnerships.
| Route | What you pay for | Main risk |
|---|---|---|
| Platform integration | Set-up time and catalogue clean-up | Feed only as good as your product data |
| Feed partner | A monthly tool fee, plus mapping work | Another system that can fall out of sync |
| Direct feed | Developer time to build and maintain it | Waitlist; you own every failure |
Our summary of the routes reported by Search Engine Journal; costs depend on catalogue size and platform.
Our view: if your store runs on a platform with a native integration, start there and spend the money on data quality instead. A feed partner earns its fee when you sell across several marketplaces from one catalogue, or when your platform has no integration. A direct feed only makes sense for large catalogues with an engineering team already maintaining other feeds.
What the build actually involves
The build is mostly a catalogue clean-up, not a technical project. Engines need the same facts a careful shopper wants: an accurate title, price, availability, variants, identifiers such as GTINs, images, shipping costs and return terms. Most D2C catalogues have gaps in at least two of these.
The common gaps are predictable. Variants listed as separate products, so a colour that is out of stock still shows as available. Missing identifiers on own-brand products. Shipping and returns written only in a policy page, not in the feed. Product descriptions that are pure brand voice, with no material, size or compatibility facts an answer can repeat.
The hard part is keeping it accurate
A feed that was right at launch drifts within weeks. Prices change for promotions, stock sells out, new variants arrive and shipping thresholds move. Every mismatch between the feed and your site is a chance for an answer to show a wrong price or a product that cannot be bought.
Keeping it right takes three habits. Sync on every change, not on a nightly schedule, for price and stock. Check a sample of products each week against the live site and the AI answers that show them. Give one person ownership of the feed, with the right to stop a promotion going live if the feed cannot carry it.
Our view: budget the upkeep as a standing weekly task, not a project. For a catalogue the size of Tailspin's in the example above, a few hours a week is a reasonable assumption; for a fast-moving catalogue with daily promotions, it can be a large part of someone's job. Wrong prices are also the error buyers notice most, as our note on where AI pricing errors come from sets out.
Where do agencies overcharge?
Agencies overcharge most where the work is a switch, not a service. Be wary of four patterns.
- A retainer to turn on a native integration. Connecting a platform's built-in channel is a one-off task. Pay for the clean-up, not for a monthly fee to keep a switch on.
- "AI feed optimisation" that stuffs titles. Titles crammed with keywords read badly in an answer and can breach the platform's listing rules. Factual titles and complete attributes do more.
- Per-SKU pricing for one-off mapping. Mapping categories and attributes is mostly done once per category, not once per product.
- Monitoring you cannot inspect. If a monthly fee covers "AI shelf monitoring", ask for the prompts, engines, dates and raw answers behind the report.
Pricing transparency helps here. Our survey of what GEO agencies and tools publish as prices shows how few list them at all. Ask any provider to split one-off clean-up from monthly upkeep, and to say who does the weekly checks.
Is the effort worth it for a small brand?
For most D2C brands, yes, because the cost is mostly clean-up you should do anyway. Shopify told investors in its second-quarter results that AI-driven traffic and orders to its merchants had grown sharply while search traffic was still growing, as TechCrunch reported in August 2026.3 We looked at what that means in our note on the Shopify results.
Clean product data helps Google Shopping, marketplaces and your own site search as much as it helps AI answers. If the feed was the only reason, the case would be weaker. It rarely is.
Where the feed is not the problem, the sources an engine trusts may be. Our playbook for e-commerce brands covers the other causes, such as reviews and third-party roundups that never mention you.
