Large controlled studies on intercessory prayer, most notably the STEP study (2006) on cardiac surgery patients, found prayer had no beneficial effect on recovery — and one arm even showed slightly worse outcomes for patients who knew they were being prayed for — which is exactly what we'd expect if prayer does nothing.

Why people find it persuasive

This was a rigorous, well-funded, peer-reviewed, double-blind randomized controlled trial — the gold standard of scientific evidence — and it found no statistically significant benefit from intercessory prayer, directly testing the very claim believers make about prayer's efficacy.

Important distinctions

• • Prayer as a mechanistic, on-demand intervention testable like a drug vs. prayer as personal communication with a free agent who responds according to his own will and wisdom.
• • Third-party, stranger-led, scripted intercessory prayer for unknown patients vs. prayer within relationship, communal worship, and ongoing spiritual life.
• • Statistically undetectable physical effect vs. no effect on any dimension (psychological, spiritual, relational) that such a study wasn't designed to measure.
• • One study's null result vs. the full, genuinely mixed body of prayer-outcome research.

Christian response

• • The theological model of prayer assumed by such studies is arguably a category error: Christian prayer is not a mechanistic lever that reliably produces a specified physical outcome when 'applied' by a sufficiently large sample size, but a personal request to a free, wise agent who may have good reasons, unknown to us, to answer, delay, or decline any particular request — a well-designed RCT can rigorously test whether a drug produces an effect regardless of the drug's 'wishes,' but cannot, even in principle, force a free personal agent to respond to a research protocol's stimulus in a uniform, reliable, statistically detectable way, any more than you could run an RCT proving a friend doesn't love you because a controlled trial of asking-for-favors showed inconsistent responses.
• • Even taken on its own terms as an empirical claim, the study design has been criticized on methodological grounds by researchers across the spectrum: the intercessors were strangers given only first names and a first initial, offering brief scripted prayers for a short, defined period for people they had no relationship with, which is a specific and unusual practice not representative of how most religious traditions understand efficacious prayer, which typically emphasizes persistent, relational, communal prayer, often over long periods, from people who know and love the one prayed for — closer to the reported context of most testimonies of answered prayer than a one-off scripted intervention by strangers.
• • The finding that the group who knew they were being prayed for had slightly worse outcomes is itself suggestive of a plausible psychological mechanism unrelated to God's existence or responsiveness — being told 'you are the subject of a study because your case is being watched by researchers who thought you needed extra prayer' plausibly induced performance anxiety ('am I so sick they think I need this?'), a well-documented phenomenon in clinical research entirely explicable without any claim about prayer's spiritual efficacy or lack thereof.
• • The wider empirical literature on religion, prayer, and health outcomes is considerably more mixed and complicated than the STEP study alone suggests — some studies (e.g., on personal, relational prayer's correlation with reduced anxiety, and broader religious-involvement studies like those summarized by Harold Koenig's health research) show measurable psychological and even some physiological correlates, while direct intercessory-prayer-at-a-distance studies (a much narrower and more contested research question) show genuinely inconsistent results across different trials (some earlier smaller studies, like the controversial Cha/Wirth study, reported positive results but faced serious methodological criticism) — the state of the evidence is mixed and contested, not a clean, settled null result.
• • Theologically, many Christian traditions explicitly teach that prayer's primary purpose is not to function as a guaranteed outcome-generating mechanism but to align the one praying (and, in intercession, the community) with God's will, to build relationship and dependence on God, and sometimes simply to be heard and to comfort — meaning 'no detectable physical effect in a specific narrow trial' does not touch the primary claimed purposes and effects of prayer even if it is relevant to a narrower, and theologically somewhat naive, claim that prayer functions like a medical treatment.

Christian counterresponse

There is real force to the unfalsifiability worry, and honest theology should acknowledge that if prayer is modeled as a reliable, on-demand causal mechanism, that specific model is straightforwardly disconfirmed by controlled evidence, and Christians who speak in those terms should recalibrate; but the mainstream historic Christian model of prayer (petition to a free, personal, wise God who may say yes, no, or wait, for reasons pertaining to a larger providential plan not reducible to any single request) was never meant to predict uniform statistical outcomes across a randomized population in the first place, so its content is falsifiable in other ways (e.g., if prayer never correlated with any measurable good outcome, ever, across any studied dimension, that would be some evidence against even the relational model) without being falsifiable by this particular kind of trial, which tested a much narrower and more mechanistic claim than the mainstream theological one.