Imagine there's a best-selling toaster. Amazon can recommend it and many people will be happy buying it. Then, one day a new toaster comes out, and it might be even better than the old best-seller. How would it get sufficient traction to be recommended by the algorithm? To send a strong enough signal, it would probably need a level of sales that could only be achieved with some level of advertising.
Separately, I'd note that Amazon doesn't want high levels of returns (which it has to pay for in many cases), so they still try to herd people toward products that they will find acceptable. Might they not love them as much as some others? Perhaps, though this is a fine line to try to walk, and I'm not sure that Amazon is actually trying to walk it.
> Then, one day a new toaster comes out, and it might be even better than the old best-seller. How would it get sufficient traction to be recommended by the algorithm?
This is a solved problem. Recommendation algorithms aren't just "highest average rating", they take into account number of ratings, recent trends, etc.
Even if some ads improve this experience, Amazon is orders of magnitude past that threshold into hostility and toxicity. You have to scroll through pages of sponsored results to find anything worthwhile.