On this page
A shopper can see your price inside an AI answer before visiting your page. A price that is wrong in your feed, the list of your products you give Google for shopping, can be wrong in the answer however good the page is.
Comparing each product's price, stock and ID across the page, the product details coded into it and the feed takes an ecommerce coordinator about 19% of the week by hand, on our assumption: about $24,282 a year at the Bureau of Labor Statistics’ mean pay for marketing specialists with benefits.
At Qvery, we help online retailers stop paying people to compare product pages with the feed by hand and keep a wrong price out of shoppers' answers, with custom AI agents that check every product's price, stock and markup each morning and draft the fix for whoever owns it.
Your Shoppers See a Price Your Page Does Not Show
Answers read from the page, its coded product details and the feed, and a rebuild only touches the first two. Google's Merchant Center, where you manage that feed, reads the page's product details to correct feed prices and availability; when it cannot settle a mismatch, it can disapprove the product instead.
Which Copy Is Wrong
of the product pages the answers used that showed a price showed it before any code ran.
So most of the pages these answers used already carry their price as the server sends them, and a rebuild is the answer only when your store cannot send it that way.
Readable product data does not mean your product gets named: whether an answer names it is a separate count, so the same shoppers' questions need asking on a schedule.
What Qvery’s Agents Check Each Morning
Each morning, and after each price change, the check compares every product's price, stock and ID across the page, its coded details and the feed. Most of that is a script, a short fixed program: fetching pages and comparing fields needs no judgment.
The agents earn their place on the products that fail: they read the page's code, the template it is built from and the feed row, work out which copy is wrong against your catalog export, and why (an old saved copy of the page, an app that adds the price after loading, a feed rule that rounds) and draft the ticket for the person who owns that copy.
The tickets arrive in one list each morning: the product, the copy that disagrees, the evidence and the owner it goes to.
The calculator prices that comparison for a retailer of about 250 people with a few thousand products: one ecommerce coordinator, about 19% of the week. At about $127,800 a year, that is about $24,282 a year.
With the agents taking about 65% of it, at about $300 a month to run, it costs about $12,099, saving about $12,183 a year and 257 hours; $10,000 of build pays back in about 9.8 months. The share of the week, the share the agents take and the running cost are our assumptions, to replace with your own.
Catalog and Feed Check ROI Calculator
What Comparing Each Product's Price, Stock And ID Across The Page, Its Coded Product Details And Your Shopping Product List, And Drafting The Fix For Whoever Owns The Copy Costs You, And What An Agent Saves
Saved Per Year
$12,183
257 hours handed back, after $3,600 a year to run it
- Hours Handed Back
- 257 hours a year
- Agent Running Cost (Estimate)
- $300 a month
- Payback Per $10,000 Of Build
- 9.8 months
On these numbers, the agent saves $12,183 a year after its own running cost, and every $10,000 of the build is earned back in 9.8 months.
What Stays With Your Catalog Owner
What the agents may read and do is set in a written agreement, your data is processed where that agreement says, and every action the agents take is logged.
Bring Us a Catalog Export and a Month of Price Changes
Bring a catalog export and a month of price changes, with access to your server logs, the record of which programs read the site, and tell us how your team approves a change. We have the first agents running within two weeks once access is in place, and the owner of each copy approves every fix.
Frequently asked questions
Can ChatGPT read JavaScript on my product pages?
OpenAI's documentation does not say its reading program runs the code that fills a page in after it loads. Put the price, the stock status and the coded product details in the page as your server first sends it.
Does Google AI Mode read JavaScript?
Google renders pages after it reads them, so content that code adds later can reach Google, sometimes after a wait. Google still recommends putting the coded product details in the page as first sent, because details built by code make its shopping reads less frequent and less reliable.
Does Product schema make ChatGPT or Google recommend our products?
No engine says so. Product schema, the coded product details, states your name, price and availability in a form machines read reliably, and Merchant Center uses it to keep your Google listings accurate. Which products an answer names depends on far more than the coded details.
Should we rebuild our product pages for AI search?
Consider a rebuild only if your store cannot send prices, stock and the coded product details in the page as first sent. Check access, the page as first sent, its coded details and the feed first: the failed check may need only a setting or a template change.
If we block GPTBot, do our products disappear from ChatGPT?
No. GPTBot is the program OpenAI uses to gather training data; its search uses a separate program, OAI-SearchBot, and OpenAI treats the two as independent settings in the file that tells them what they may read. Blocking OAI-SearchBot is what keeps a site out of ChatGPT's search answers.
We sell on Shopify. Do we need to do anything for ChatGPT?
Shopify connects its merchants' catalogs to ChatGPT, so there is usually no separate listing to build. What remains is checking that the catalog your store sends is right, and that your product pages carry the same price and stock as the catalog.
How do I see my product page the way an AI crawler sees it?
View the page source, or reload with JavaScript off, to see roughly what a reading program that runs no code receives. Your logs, filtered by the addresses OpenAI publishes for its search program, show what it got; Google's URL Inspection tool shows the page Google received.
Can our firewall block AI search even if robots.txt allows it?
Yes. Bot protection can turn away a search engine's reading program even when robots.txt, the file that tells it what it may read, allows it. Check your logs against the published addresses before changing a rule; a blocked impostor is not evidence that search is blocked.
Why is the price in an AI answer different from our product page?
A shopping answer can draw on a product feed or retailer listing as well as your page. Compare the feed price with the visible page and its Product markup, then check Merchant Center for mismatch warnings or disapprovals.
Does a passing Rich Results Test prove every crawler sees our markup?
No. Google's Rich Results Test, its checker for coded product details, reads the page after code has run, so it can see details that code added later. Search the original page source for the product data and its price to see what a reading program receives first.
When are custom agents worth building for catalog checks?
A small catalog on one template may need only a weekly spreadsheet check. Agents become worth it when thousands of variants span several templates and feeds; time the manual comparison before deciding what to automate.
What should a person approve when the agents check our store?
Your team approves changes to the store's design, the feed and the firewall, including which reading programs to let in. The agents read prices and report mismatches; they never set a price, and a page the search engine's program did not visit that day stays unknown, never passed.
How should we judge whether automated checks help?
Time a manual catalog comparison before the agents start, then track hours saved each week. Open mismatches, days from detection to fix, and Merchant Center disapprovals show whether findings are being fixed, rather than merely collected.
