Scientists submit their experimental work to journals in their field for review and publication. Only a fraction of all papers submitted to journals are actually published. Some journals are extremely influential and broadly read, such as Science, Nature, the New England Journal of Medicine, and Cell, and are considered important sources of scientific information. They have much lower acceptance rates, probably well under 1%. Because they are so selective, the peer review of papers submitted to those journals is considered more rigorous than the average. Two or more scientists who are familiar with the subject area review papers submitted to a journal. They do not repeat the work, but make a reasoned judgment on the data's validity, and whether the data support the conclusions. The reviewers may recommend acceptance, modification, major revisions, even more experiments, or outright rejection to the journal's editor.
There are variations in the way peer review is done. Some journals conceal the authors' identity from the reviewers, and almost all journals conceal the reviewers' identity from the authors. The purpose of this is to "blind" people so that prejudices, scientific competition, and personal likes and dislikes do not come into play. It is not a perfect system, but it has served the scientific community well for many decades.
Peer review does not guarantee that the published work is correct or that the data are real. There have been some spectacular instances of fabrication, falsification, and plagiarism in major scientific literature, and numerous instances of it in less-publicized places. Major journals have had to retract papers as a result. Retraction of a paper is an editor's only option after the paper has been published, but then shown to be defective. In science, there is an assumption of integrity—we believe one another until we have a reason not to. Integrity is at the heart of the process, and people who are caught violating it cannot remain in a scientific or research career.