We Re-Ran Our ChatGPT Shopping Audit Two Weeks Later. Here's What Changed.

November 11, 2026

By Steve Merrill, Founder of WRKNG Digital | November 11, 2026

What did we go back and check?

Two weeks ago we audited a batch of Shopify stores to see how ChatGPT Shopping decides which products to surface. We wrote up what we found in How Does ChatGPT Decide Which Products to Recommend? What We Found Auditing Shopify Stores. Incomplete Product schema. Reviews hidden in JavaScript widgets. Feed data that didn't match the live page.

This week we went back to the same 14 stores and ran the same 20 buying prompts against each one. Same prompts, same order, same day of the week. We wanted a clean before-and-after, not a vibe check.

Six stores made at least one fix from our list. Eight did nothing. The gap between those two groups is the whole story.

Which stores fixed something, and what happened?

Here's the thing. Every store that touched its product data moved, and most moved fast.

A skincare store fixed GTIN mismatches between its feed and its product pages. Two weeks ago it showed up in 2 of our 20 prompts. This week, 9 of 20. That's the single biggest jump in the group, and it came from a feed fix, not new content.

A yoga equipment store added full Product schema, pulled its reviews out of a widget, and published them as plain readable text. It went from 0 citations to 4 out of 20. Zero to something is a bigger deal than the raw number looks, because it means the assistant went from never considering the store to actively quoting it.

A pet supplies store rewrote its top 20 product descriptions to answer real buying questions like sizing and material, and added FAQ schema to its category pages. It climbed from 3 citations to 7. I've seen this exact pattern in other audits: specific copy gets quoted, generic copy gets paraphrased into nothing.

The other three stores that made fixes saw smaller but real gains, averaging about 2 extra citations each. Not huge. Still movement, in two weeks, from changes that took an afternoon.

Which stores did nothing, and what happened?

Flat. Mostly flat.

Five of the eight stores that skipped our fix list landed within one citation of where they started. No real change either direction. The assistant kept reading the same broken schema and the same locked reviews it read two weeks ago, so it kept making the same decisions.

One store actually dropped, from 5 citations to 3. Nothing changed on their page. What changed was a competitor in the same category, who fixed their GTINs and stock status during the same two weeks. ChatGPT didn't punish the store that stood still. It just found a better answer somewhere else and used that instead.

Same story, different year, different platform. The algorithm or the assistant always rewards the page that gives it less work to do.

Did ChatGPT Shopping change anything on its own?

A little. OpenAI has kept building out product results inside the chat experience since the feature launched (OpenAI), and in this second pass we noticed the model leaning harder on live feed data and current page content, and less on whatever it remembered from training. We can't prove that from 14 stores and 20 prompts. But it lines up with what Google has said for years about structured data: machines trust what they can verify right now over what they memorized once (Google Search Central).

If that pattern holds, stale product data gets more expensive every month, not less.

What's the one change that keeps winning?

Feed accuracy. Specifically, GTINs and stock status matching between your feed and your live page.

Shopify documents exactly which identifiers your product data needs to stay trustworthy to outside systems (Shopify), and in this re-audit, the stores that cleaned up that layer moved more than the stores that only rewrote copy. Copy helps. A clean feed moves the needle faster.

We rewrote plenty of product descriptions over the past two weeks too, and those mattered. But if you only have one afternoon, spend it on your feed.

What should you take from this?

Fix something. Anything on the list from the original audit. The stores that touched their data moved within two weeks. The stores that waited stayed exactly where they were, or lost ground to a competitor who didn't wait.

We'll run this same audit again in two weeks on the same 14 stores. If the pattern holds, the gap between the fixers and the non-fixers is going to keep widening, and it gets harder to close the longer you leave it.

Frequently Asked Questions

How long does it take for ChatGPT to notice a fixed product page?

In this re-audit, stores that fixed schema and feed issues showed up in test prompts within two weeks. Some moved inside the first week. The common thread was a clean feed and complete Product schema, not store size.

Can a store lose ChatGPT Shopping visibility it already had?

Yes. One store in our sample dropped from 5 citations to 3 out of 20 prompts with no changes on their end. A competitor fixed their feed and took the spot. Visibility is relative here, not something you keep once you earn it.

Did OpenAI change how ChatGPT Shopping works in the last two weeks?

We didn't see a dramatic platform shift. Answers in our second pass leaned harder on live product feeds and less on older training data than they did two weeks earlier. That trend favors stores with current, accurate feeds.

What is the single fastest fix for ChatGPT Shopping visibility?

Matching GTINs and stock status between your product feed and your live page. Across this re-audit, that fix was tied most consistently to a citation increase, ahead of new copy or added FAQ schema.

Is it worth re-auditing AI shopping visibility this often?

Every two to four weeks is enough to see real movement without chasing noise. Checking daily just shows you random variance in how the assistant samples sources, not a trend you can act on.

If you run a Shopify or DTC store and want to know exactly where your product data is losing you ChatGPT Shopping citations, see how WRKNG Digital approaches agentic commerce here: https://wrkngdigital.com/agentic-commerce-landing-page.

Back to Blog