edX occupies a specific and useful spot in the online learning market, because unlike a marketplace where anyone can publish, its catalogue is built around content from actual universities and established companies, and that provenance is the main reason to use it. Founded by MIT and Harvard and now part of the wider 2U business, it hosts a deep set of AI and machine learning material that spans the well known Columbia artificial intelligence course, IBM and other industry certificates, Berkeley and MIT data science and machine learning content, and full MicroMasters programmes that bundle several courses into something closer to a graduate module. For a learner that means two real advantages. First, the teaching often comes from people who genuinely research or build in the field, so at its best the material has a depth and rigour that a lot of quick marketplace courses simply do not attempt.
Second, the certificates carry more weight than most, and the MicroMasters credentials in particular are recognised by employers and can even count toward credit at partner universities, which matters if you are learning partly to prove something on paper. The audit model is also genuinely learner friendly in principle, because you can enrol in most courses for free and work through the video lectures and readings without paying, which makes edX one of the better places to actually try before you buy. The honest problems are worth going in with eyes open. Because every course is really the product of a different institution and teaching team, quality is not consistent, and you can move from a beautifully constructed MIT course to one that is dry, dated or poorly paced without much warning, so the platform name is not a reliable guarantee on its own.
The free audit track has also been squeezed over the years, and more of the graded assignments, exams, deadlines and sometimes even the ability to finish the course properly now sit behind the paid upgrade, so the free experience is more of a preview than a full course than it used to be. On top of that, a field like AI moves quickly, and some of the headline courses were filmed several years ago and have not been meaningfully refreshed, so you occasionally get solid fundamentals wrapped in examples and tooling that feel a step behind current practice. My take is that edX is best understood as a curated shop window onto university and industry teaching rather than a single coherent course provider, and used that way it is excellent. If you want real academic depth, a credential that carries recognition, and the option to audit before committing, it is one of the strongest platforms available, particularly for the MicroMasters programmes and the better MIT, Columbia and Berkeley material.
Just treat each course on its own merits, audit before you pay, check when the content was last updated, and budget for the verified track if you actually want the graded work and the certificate rather than only the lectures.