AI product recommendations are becoming a shopping channel in their own right: OpenAI says hundreds of millions of people use ChatGPT to find, understand and compare products[3]. This post covers what OpenAI says decides ChatGPT product recommendations, why AI recommendations change from one answer to the next, and what that means for how you measure it and how to get recommended.
Third-party research. SparkToro co-ran its study with Gumshoe.ai, which sells AI tracking, and its authors note they are not professional researchers[7].
How ChatGPT picks the products it recommends
OpenAI describes the inputs in its own help centre. A product appears when ChatGPT judges it relevant to your intent, considering your question and context such as Memory or custom instructions[1]. To decide, it uses three things:
- Structured product data from first-party and third-party providers, such as price and description, plus other third-party content.
- The model’s own answer, generated before it considers any new search results.
- OpenAI’s safety standards and product policies.
It is explicit that “product results are selected independently by ChatGPT and are not ads, nor influenced by any OpenAI partnerships”[1]. When several shops sell the same item, merchants are ranked on factors like availability, price, quality, and whether they make or primarily sell it[1].
Sources for the timeline, in order[5][4][3][2]. For feeds, checkout and merchant setup, see the ChatGPT Shopping guide.
Why ChatGPT’s product recommendations keep changing
Ask the same buying question twice and you will usually get a different list. SparkToro and Gumshoe had 600 volunteers run 12 prompts through ChatGPT, Claude and Google’s AI, 2,961 runs in all. There was less than a 1 in 100 chance that ChatGPT or Google’s AI would return the same list of brands in any two responses, and less than 1 in 1,000 of the same list in the same order[7].
Two more findings matter for anyone reporting on this. Brands could be present almost every time without leading: City of Hope appeared in 69 of 71 ChatGPT answers but was the first mention in only 25[7]. And consistency depended on the category: it was higher where there are few real options, like cloud providers, and lower where there are many, like novels[7].
SISTRIX saw a similar split for the sources behind AI answers. For brand queries, brand domains were anchored: 43% of them were present in all 17 weeks it tracked, while the domains cited alongside them rotated at around 70% a week[8]. Some positions are held; most are passed around.
What one check can and cannot tell you
If recommendations rotate, a single check is a coin toss reported as a finding. Move the sliders to see how wide the honest range is at different numbers of runs, and when a gap between you and a rival stops being noise.
This is why SparkToro concluded that a visibility percentage across dozens to hundreds of prompts, run several times, is a reasonable metric, and that “any tool that gives a ‘ranking position in AI’ is full of baloney”[7]. It is also why AI visibility tracking in CoreCited reports a range next to every share-of-voice number, and competitor tracking refuses to rank brands whose ranges overlap.
How to get ChatGPT to recommend your product
There is no form to fill in and no placement to buy. What you can do is make your product easy to find and easy to vouch for, in the places ChatGPT looks. Tick these off as you go; your progress is saved in this browser.
The evidence for each: Shopify Catalog and feed applications are described by OpenAI[2][1]; the OAI-SearchBot rule is in OpenAI’s crawler documentation[6]; the Reddit share is Ahrefs’ September 2026 data[9]. Google says much the same for its own AI features: they can show what is being said about products across the web, including blogs, videos and forums, and Merchant Center feeds help products appear[10].
To check the crawler part now, run your site through the AI Crawler Checker, or build the rules with the robots.txt generator. To find the reviews and lists that recommend your competitors, use citation analysis. The guide to how AI chooses citations covers the source side in more depth.
What does not work
Questions people ask
How does ChatGPT decide which products to recommend?
OpenAI says product results are selected by ChatGPT based on relevance to your intent, using your question and context such as Memory, structured product data from first-party and third-party providers, other third-party content, and the model's own response before it looks at new search results. It says product results are not ads and are not influenced by OpenAI partnerships.
Why does ChatGPT recommend different products each time?
Because answers are generated fresh each time and draw on search results that change. In SparkToro and Gumshoe's January 2026 study, there was less than a 1 in 100 chance that ChatGPT or Google's AI would give the same list of brands in any two responses to the same prompt.
How do I get my product recommended by ChatGPT?
Make sure ChatGPT can find accurate data about it and trusted third parties talk about it. That means complete product data (automatic for Shopify merchants, by application for others), allowing OAI-SearchBot in robots.txt, and being covered in the reviews, comparison lists and forums ChatGPT cites. No one can guarantee a recommendation.
Can I pay to be recommended by ChatGPT?
Not in the product results. OpenAI says they are not ads and ads are separate. ChatGPT does show ads in some markets, clearly separated from the organic answer, but buying one does not change which products ChatGPT recommends.
What are AI product recommendations?
Products an AI assistant suggests in answer to a buying question, such as ChatGPT, Gemini, Google's AI Mode or Perplexity naming the best running shoes or CRM tools. Unlike a ranked search results page, the list is written fresh for each answer and often changes from one run to the next.
Is there a ranking position in ChatGPT?
Not a stable one. Rand Fishkin's conclusion from the SparkToro study was that a visibility percentage across many prompts run many times is a reasonable metric, while a single ranking position in AI is not. Measure how often you are recommended, not where you appeared once.