ChatGPT is still Google Shopping's biggest fan

When ChatGPT launched its search features, the narrative was set: Google was in trouble. Every tech headline predicted the end of traditional search. Sam Altman even declared publicly, "I don't use Google anymore. I legitimately cannot tell you the last time I did a Google search."

But behind the scenes, a very different story has been playing out. ChatGPT has been quietly relying on Google's own search results to power its answers. And a new experiment from Semrush confirms it goes even deeper than anyone thought: ChatGPT is running encoded queries through Google Shopping to populate its product carousels. The tool that's supposed to replace Google? It's built on top of Google.

I've been following this story for a while now, and the irony just keeps stacking up. Here's what's actually happening, why it matters for ecommerce brands, and what you should be doing about it.

ChatGPT has been caught using Google search results

In July 2025, SEO consultant Aleyda Solis ran a controlled experiment that caught ChatGPT red-handed. She published a new page on her site without submitting it to any search engine, then tracked what happened as each engine found it. When Google indexed the page (but Bing still hadn't), ChatGPT suddenly started answering questions about it. The response used the exact text from Google's search snippet. Word for word.

Even after Bing eventually indexed the page, ChatGPT continued pulling from Google's snippets rather than Bing's. This matters because OpenAI's official position is that ChatGPT Search relies on its own crawler, Microsoft Bing, and licensed publisher data. Google isn't on that list. As Solis put it: "It's confirmed: ChatGPT is somehow relying on Google Search results snippets for their answers."

SerpAPI is the middleman Google is now suing

So how is ChatGPT accessing Google's data? In August 2025, Search Engine Land reported (based on findings from The Information) that OpenAI had been using SerpAPI, an 8-year-old scraping firm in Austin, Texas, to pull Google Search results for real-time topics like news, sports, and finance. SerpAPI had even listed OpenAI as a customer on its website as recently as May 2024 before quietly removing the reference.

Google responded in December 2025 by filing a DMCA lawsuit against SerpAPI. The allegations are something else. Google claims SerpAPI sends hundreds of millions of artificial search requests daily, with volume increasing by 25,000% over two years. According to Google, SerpAPI uses cloaking, fake browser fingerprints, and automated CAPTCHA bypassing to get around SearchGuard, a security system that cost Google "tens of thousands of person hours and millions of dollars of investment" to build. SerpAPI's other reported customers include Meta, Apple, and Perplexity. The lawsuit is ongoing.

The email that makes all of this ironic

Here's where the story gets genuinely funny. Before the scraping controversy, before the SerpAPI revelations, OpenAI simply asked Google for access. Politely.

This email between OpenAI and Google was entered as DOJ exhibit PXR0181 during the antitrust remedies trial:

On August 15, 2024, Michelle Fradin, OpenAI's Head of Business Operations, emailed Philipp Schindler, Google's Chief Business Officer, directly requesting access to Google's Search API. OpenAI stated it "believe[s] having multiple [search API] partners, and in particular Google's API, would enable [it] to provide a better product to users." The email argued this would "promote more choice in search experiences" and benefit "innovation and evolution in search."

Google said no. During testimony on April 22, 2025, Nick Turley, OpenAI's head of product for ChatGPT, confirmed that Google declined because the integration "would involve too many competitors." Turley also admitted under oath that ChatGPT was "years away" from answering most queries using its own search technology, and that its existing search partner (Bing) had "significant quality issues."

Here's what I find so ironic about this whole thing. OpenAI surfaced this email as evidence in the DOJ's antitrust case against Google, essentially arguing: "Look, Google's search monopoly is so dominant that even we need access to it, and they won't let us in." Fair point. But fast forward to today, roughly a year and a half later, and the picture looks very different. OpenAI asked for permission, Google declined, and then OpenAI found a way to tap into Google's results anyway through SerpAPI and scraping. Google is now suing over exactly that. You really can't make this stuff up.

75% of ChatGPT's top product picks come from Google Shopping

Despite all this legal drama, ChatGPT is still deeply dependent on Google for product recommendations. A January 2026 experiment by Semrush reveals how deep the dependency goes.

Researchers discovered that ChatGPT runs two separate sets of "fan-out" queries when you ask for product recommendations. The first set gathers contextual information for the written response. The second, hidden behind Base64 encoding in the network traffic, sends queries directly to Google Shopping to build the product carousel. These shopping queries are separate from the ones that power the conversational answer, meaning ChatGPT has a dedicated pipeline for pulling product data from Google.

After running the experiment 100 times (each prompt repeated 5 times to account for randomness), ChatGPT's top product recommendation appeared in Google Shopping's first 3 results 75% of the time. Product titles, retailer names, and prices matched exactly. ChatGPT's shopping carousels aren't generated by the AI model. They're pulled straight from Google's product index, including live pricing, reviews, and merchant data that Google has spent years collecting.

OpenAI is building, but Google has a 27-year head start

OpenAI is not standing still. They've launched a merchant portal where brands can apply to submit product feeds to ChatGPT. They've published a product feed specification. They've rolled out Instant Checkout co-developed with Stripe, with Etsy sellers already live and over a million Shopify merchants reportedly joining. Their Shopping Research feature, powered by a GPT-5 mini model, processes roughly 50 million shopping queries daily.

But the gap remains enormous. Google's Shopping Graph now contains over 50 billion product listings (per Sundar Pichai's January 2026 NRF keynote). It draws from merchant feeds, organic crawling, manufacturer specs, reviews, real-time pricing, and user-generated content. Google StoreBot, a specialized crawler built solely for product data, helps keep this index current, with over 2 billion product listings updating every single hour. Google first unveiled the Shopping Graph in 2021, and it sits on top of more than 27 years of search infrastructure. Replicating that is a multi-year effort, and Nick Turley essentially said as much under oath.

Google's organic shelf is your best ChatGPT strategy

When 75% of ChatGPT's top product recommendations come directly from Google Shopping, optimizing for Google IS optimizing for ChatGPT.

That means your product feeds, structured data, pricing accuracy, and review signals matter more than ever. Not just for Google visibility, but for showing up in AI shopping experiences across the board. The brands winning on Google's organic shopping shelf space will also be the ones winning in ChatGPT, Gemini, and whatever comes next.

Should you apply to OpenAI's merchant portal and submit feeds there too? Yes, it's worth getting in early. Should you start preparing for the Universal Commerce Protocol and agentic shopping? Absolutely. But the data right now is unambiguous: your Google Shopping presence comes first.

If you want to see exactly where your products stand in Google Shopping and start optimizing for both Google and AI visibility, create a free Productrise account and start tracking your organic product positions today.