I like the other comment regarding asking people you know or have hired for referrals. But there's a separate question about what to do when you don't have any referrals to rely on.
In the past, the answer was to rely on recruiting experts and your own experience to analyze resumes, and to hope that some fraction of your time interviewing was well spent. But as others in this thread have pointed out, the quality of resumes as a signal is rapidly decreasing with AI. So now the question is what the new signal should be.
The purpose of hiring is really just to understand three things: (1) can the person do the job, (2) will I and the rest of the team want to work with them for the foreseeable future, and (3) are there any red flags. The latter two are assessed in interviews or backchanneling. But the first, "can this person do the job," is the one that has gotten much harder to tell from the resume alone.
What does "can the person do the job" even mean in the era of AI? For SWE, a lot of teams are still leaning on outdated coding challenges that test the ability to write code by hand (quality, organization, efficiency, speed) and the direct knowledge held in someone's head (vocabulary, concepts, tools). But what we actually need to test for is much more nebulous: thought process, judgment, knowing what questions to ask, taste, context switching, and how the candidate does all of this with AI effectively. Traditional coding challenges (where you prohibit cheating as much as possible) get at some of this, but they're no longer the optimal solution. They're an old comfortable solution being applied to a more modern, complicated problem.
I said this in another comment in this thread, but the best thing to do is offer an open-ended question. One that doesn't have a concrete answer, and one they're free to answer however they wish. Like a Turing test, 5 minutes at most is enough. Want to use AI? Great. Want to curate your answer? Great. Ask them to architect something. Describe a bug and ask if it should be fixed. If they think it should, ask how they'd approach debugging. Or ask about a technical decision they've made that they're not sure was right, and have them make the case for the other choice.
The same approach works in other domains. For sales, have them sell to someone in a specific scenario. For marketing, ask about a campaign they didn't like and what they would have done differently. For accounting, ask about the most common problems they've run into with clients and internal stakeholders, and how they handled those situations.
The goal is to get them talking on video for 2 to 5 minutes about something real that matters to the role. And whenever they put in the effort to do this, give them transparency. Let them know you've personally received it and will watch it. Follow up with an actual response out of respect for their time, even if the answer is no.
To anyone worried about the time this takes: (1) it immediately weeds out low-interest candidates, (2) I'd rather watch a few-minute video than do a 30-minute phone screen based on a resume alone for someone who's obviously not a fit, and (3) multiple stakeholders can watch the same video and form a baseline for their own follow-up interviews. It makes the rest of the process better, not worse.