How does an AI summary of crappy reviews help? It doesn't make the original data any more trustworthy.

it depends on the sources that are used and above all on how much data is imagined before making an average ranking of the products... so in that case it is very useful rather than reading N sources and then deciding what to focus on

That's the thing, every freely available source about any remotely popular product may as well be entirely astroturfed, or pure SEO listicle slop. The good sources aren't openly available, they're private. The average ranking of 100 listicles doesn't tell me anything.

I don't follow your logic here... why would N+1 or N+100 sources necessarily be any better than N? If you start with 5 hand-curated, in-depth reviews picked for quality and then combine them with 100,000 shitty AI-generated ones... you're going to end up with less signal than you started. Expanding your dataset to include more poor quality sources won't make the resulting summary better.