New scientific discoveries are communicated through multiple channels. Publications remain essential for scientists, whereas general public relies on news articles. Numerous studies investigated the path from publications to news and identified press releases as the key factor, associating them with issues such as exaggerations, but less is known about the direct influence of press releases on subject choice or on the content of media reports. Here, a crosssectional sample of publications related to genetics and CRISPR is assessed in three independent datasets (n = 1362 publications, n = 461 press releases). Analysis finds 92.5% (CI = 88.5-96.5) dependence of news outlets on press releases in terms of topic choice and 39-43% explicit use of passages from press releases. Publications without press releases are described by 74x fewer news outlets. Even if they come from leading journals or universities, lack of a press release leads to 8.8x less coverage. Given the high impact of press releases, their reliability is especially relevant, but the majority of them omits interest disclosures -84.3% (CI = 80.8-87.7) of press releases did not mention existing conflicts of interest, including multiple patent applications. These results establish the major indirect and direct role of press releases in science communication via online media. In line with previous research, this dependency raises concerns about possible distortions of science coverage.