I am working on a web page whose visibility for Google is very important to the client. I have included a JSON-LD section on the page that contains a number of reviews and a number of ratings in an aggregateRating element. Those numbers are different: Alle reviewers have also posted a rating, but there are many who have rated but not reviewed. As a result, the ratingCount is much higher than the reviewCount. To me, this makes sense, and nothing in the documentation at schema.org or Google says the numbers need to coincide.
My client has asked ChatGPT for advice. (Well, you can do that and why not.) ChatGPT missed a number of things that are already present on my page, no idea how, but above all it insists that ratingCount and reviewCount should be the same, otherwise Google might rate our page less highly.
I have put the page through Google's own tool for rich results and it does not display any errors.
Since Google is famously silent on how it rates the pages' metadata, can anybody help me with some fact-based ideas?
LLMs (e.g. ChatGPT) can only answer from the information they have already scraped from the web. If there is a lack of accurate answers they will probably make one up.
And I think this is the case. I've heard a few talk about rating and review counts.
I think you understand the meaning behind both, and I agree with your conclusion.
Google does not report issues with them because their rich snippets only care about the ratings.
I think, behind the scenes, they do look into more details to verify what is marked up. Are the ratings on the page, are reviews visible on the page, can people add reviews etc.