In some sense gatekeeping is essential - for example, one might want to skim all recent accepted papers from the top venues of a particular domain but skimming all submitted papers from all venues in the same domain is physically impossible, even if that would be a full time job doing nothing else. You do need the community to read all the proposed papers and throw away the majority of them, and sort others into lower-ranked publications focused on a narrow niche which you read only if that niche matters to you.

So publishing on your own site on arxiv isn't sufficient on it's own, some curation and gatekeeping is essential for academic publishing to work. However, this is not a service that Elsevier et al provides for their payments, this service is provided by uncompensated reviewers from the community.

On the other hand, the ratings are used by the evaluators of science e.g. funding agencies which won't (and can't) evaluate results directly but need some trusted proxy to separate random ramblings from acceptable science, and the prestigeous journals fulfil that function.