Long COVID in 2026: What Actually Helped, and What Still Hasn't

By the CDC's most recent national survey data, 6.4 percent of US adults reported currently experiencing long COVID, with state-level prevalence ranging from 2.9 percent to 9.7 percent, and roughly a fifth of those affected reporting significant limitations in daily activity. The National Institutes of Health has spent well over a billion dollars since 2020 on its RECOVER Initiative, the largest coordinated long COVID research program in the world, built explicitly to find treatments. That is the uncontested part of the story: the condition is common, and the money has been real.

What has been much harder to establish is what actually works. RECOVER has drawn public criticism, including from within the research community, for taking years to move from observation to treatment trials, and for producing a large biomarker literature without a matching list of therapies a clinician can prescribe. 2026 is the year that gap started, unevenly, to close — with one clear success, one clear failure, and a mechanistic debate further from settled than most coverage suggests.

What actually helped

The most substantive result of the year did not come from a drug trial. On July 1, 2026, a team led by University College London published a randomized clinical trial in “JAMA Network Open” testing ten weeks of individualized, video-delivered cognitive rehabilitation against treatment as usual in 78 adults with long COVID-related cognitive impairment. The trial ran across three sites in England between February 2023 and March 2024, enrolling participants aged 30 to 60 with prior COVID-19 infection and objectively measured cognitive impairment.

The therapy borrowed techniques long used in stroke and dementia rehabilitation: hour-long video sessions in which a therapist helped each participant work toward specific, self-chosen goals — returning to a particular task at work, reengaging with a hobby, managing a social commitment — rather than treating “brain fog” as a single target to erase.

The effect was large and it lasted. Goal attainment was significantly better in the rehabilitation group at three months (Cohen's d = 1.57) and still significantly better at six (d = 0.91). At six months, 53 percent of those who received cognitive rehabilitation reported a substantial improvement — four or more points on a ten-point scale — against 15 percent of the control group.

The caveats are real and worth stating plainly. Seventy-eight people is not a large trial. The design was single-blind: outcome assessment was blinded, but participants and therapists both knew who was receiving the extra sessions, and the primary outcome was self-reported goal attainment. That combination leaves room for expectation effects, and the comparator was treatment as usual rather than an active control — so some of the benefit may reflect attention and structure rather than the specific cognitive techniques. What argues for taking it seriously is that the benefit held at six months rather than fading, and that the outcome measured something concrete: whether people actually got back to the specific things they wanted to do. That is rare enough in this literature to matter.

What still hasn't

The pharmacological side of RECOVER has had a harder year. RECOVER-AUTONOMIC, a trial testing the heart-rate-lowering drug ivabradine against postural orthostatic tachycardia syndrome in long COVID patients, reported results at the American College of Cardiology's March 2026 conference: ivabradine did not produce a significant improvement in symptoms. A companion arm of the same trial is still testing intravenous immunoglobulin (Gamunex-C) in the same population, with results not yet available.

The largest ongoing drug trial, REVERSE-LC, is testing baricitinib — an arthritis and severe-COVID drug that dampens a specific inflammatory signaling pathway — against placebo in adults with persistent neurological and cardiopulmonary symptoms. It is a well-designed, adequately powered, placebo-controlled trial, which is exactly what this field has been missing. It is also still recruiting, with full results not expected until 2027. Any claim about what baricitinib does for long COVID right now is not a finding; it is a guess dressed as an update.

The treatment that has actually worked so far is not a drug, and the drugs that might work have not, yet, been shown to.

The unsettled question underneath it all

Much of the public narrative around long COVID's cognitive symptoms rests on a specific mechanistic story: that the virus leaves behind microscopic blood clots and persistent brain inflammation, and that fixing those would fix the fog. That story is plausible and partly supported — researchers studying blood samples from long COVID patients have identified abnormal micro-clot structures tied to markers of thrombo-inflammatory activity. But it is not the whole picture. Imaging work reported in 2026 from the University of Turku looked directly for the widespread brain inflammation the theory predicts and did not find it, complicating the assumption that inflammation is doing most of the work. The honest state of the science is that several plausible mechanisms are each partially supported and none is proven sufficient on its own — which is part of why a behavioural intervention with no biological target beat the drugs to a clear result.

What a reader should take from this

Long COVID cognitive rehabilitation, delivered the way UCL's trial delivered it — structured, goal-directed, sustained over ten weeks with a trained therapist — now has randomized evidence behind it, with a large effect at three months that was still present at six. That is genuine, if modest, good news, and it is available through existing rehabilitation frameworks rather than waiting on a drug approval. It is also one trial of 78 people against usual care, and it needs replication against an active comparator before anyone calls it settled.

On the pharmacological side, ivabradine has been tested and did not help; baricitinib and IV immunoglobulin remain open questions with trials still running, not treatments with results in hand. And the underlying biology — micro-clots, brain inflammation, or some combination not yet mapped — is still being worked out, which means any single-mechanism explanation currently being sold as settled science is getting ahead of the data.

This article is general information, not medical advice. See our Editorial Standards.