It’s 23:40 and you’re ordering blue-light glasses on the same screen the glasses are meant to protect you from. The ceiling light is on. There’s a decent chance it’s brighter than the laptop, and by the end of this you’ll know why that’s the part that matters.
You’re buying them for two reasons. Your eyes ache by six o’clock, and you sleep badly, and both of those follow you into the next day.
I went looking for what the trials found. Seventeen of them exist, Cochrane reviewed all of them in 2023 — and could not meta-analyse them, because too few published the numbers. That failure is the first appearance of the pattern this article is about.
Two products, one name
One thing to fix before going further, because it decides what the trials apply to.
Two different products are sold under the same phrase. Clear or lightly tinted lenses worn through the working day, sold on eye strain and retinal protection. And amber or red lenses worn for two or three hours before bed, sold on sleep. The seventeen trials cover both, but not on the same outcomes: the eye-strain evidence is about the day lenses, and the six sleep trials below are mostly about the evening ones.
Worth knowing the scale of the second one before we start. The entire double-blind, actigraphy-based evidence base for evening blue-blocking lenses is three trials and forty-nine people.
Six trials, one pattern
Six of the seventeen measured subjective sleep quality — 148 people between them. Three reported a benefit. Three found nothing.
Here’s how they sort.
The three that found nothing — Esaki 2017, Esaki 2020, Janku 2020 — published their numbers. No difference (MD 19.90 units, P = 0.07). No difference (MD 8.90, P = 0.25). No difference (MD 0.03, P = 0.98).
The three that found a benefit — Burkhart 2009, Knufinke 2019, Shechter 2018 — published no numbers at all. One reported results only as a figure. Two reported daily average change with no baseline or endpoint data, so no effect size can be calculated from any of them.
Now cross-reference that against Cochrane’s own risk-of-bias assessment. Esaki 2017 and Esaki 2020 are the trials that masked participants and study personnel — people didn’t know which lenses they’d been given. All three of the trials reporting a benefit were judged at high risk for both performance and detection bias: participants weren’t masked, and neither were the people scoring the results. Burkhart 2009 additionally had a significant baseline imbalance between groups.
Every trial that reported a benefit failed to mask both the participants and the people scoring the results, and none of the three published numbers you could calculate an effect size from. Every trial that masked its participants published its numbers and found nothing.
Cochrane graded the certainty of the sleep evidence as very low, downgrading three levels for risk of bias, imprecision and inconsistency.
I want to be exact about what this shows, because it’s easy to overclaim. It does not prove the effect is placebo. It’s equally consistent with reporting bias, or with a real effect that only unblinded designs happened to detect. What it shows is that the benefit and the knowledge of treatment arrive together and can’t be separated in this literature. That’s a different statement, and it’s the one the evidence supports.
It’s also the most portable thing in this article. Any product with a subjective endpoint — better sleep, less fatigue, sharper focus — is vulnerable to exactly this. When you read that something improved sleep quality, the question isn’t how big the effect was. It’s whether the person filling in the sleep diary knew which arm they were in.
The reviews that disagree — and what they found instead
Cochrane is not the only systematic review here, and two others reached friendlier conclusions. Both belong in this article, because one of them turns out to say something more interesting than either side of the argument.
A 2021 review from the University of Oklahoma covered 24 publications on sleep and reported substantial evidence that blue-blocking glasses reduce sleep onset latency in patients with sleep disorders, jet lag, or variable shift work schedules, concluding they are a viable intervention to recommend to patients with insomnia or a delayed sleep phase. Cochrane’s objection to it is specific and worth stating: that review was not prospectively registered and did not undertake risk-of-bias assessments — which is precisely the analysis that produced the pattern in the section above.
The second is a 2020 meta-analysis by Shechter and colleagues, and here is where it gets interesting. As characterised by a 2025 meta-analysis that reviewed all three: it suggested modest improvements, particularly in subjective outcomes, highlighting the gap between perceived and objectively measured sleep.
Read that again. The analysis most favourable to the product located its benefits mainly in what people reported, and flagged the divergence between reported and measured sleep as a finding in its own right. That is the same distinction the blinding pattern produces, reached from the opposite direction — and by a group whose own trial is one of the three unblinded positives above.
The 2025 synthesis states the resolution plainly: blue-blocking glasses may be most promising in clinical subgroups with circadian misalignment or evening hyperarousal, rather than in the general population.
Which gives the honest shape of this evidence. The case is weakest exactly where the product is marketed — healthy people who look at screens — and least weak in insomnia and delayed-sleep-phase populations, who are not who is buying them. Cochrane noticed the same asymmetry and remarked on it: of the six trials it found on sleep, aside from one recruiting recreational athletes, none included healthy individuals. They enrolled people with major depression, bipolar disorder and diagnosed insomnia.
A regulator got there in 2015
While the trials were accumulating, an advertising authority was reaching a conclusion by a different route.
In January 2015 Boots Opticians ran an advertisement in The Times for Boots Protect Plus Blue lenses at £70. It said modern gadgets, LED TVs, smartphones, sunlight and energy-saving bulbs give off a kind of blue light that can cause retinal cells to deteriorate over time, and that the lenses filtered out the harmful blue light and eased eye strain and fatigue.
Two people complained — one a pharmacist. In October 2015 the Advertising Standards Authority ruled against the advertisement, finding breaches of the CAP Code rules on misleading advertising, substantiation, and health-related products. The ruling’s operative sentence: they told Boots Opticians not to make claims that blue light caused retinal damage, or that the lenses filtered out a meaningful amount of harmful blue light, in the absence of adequate substantiation.
Two details are worth carrying.
On the evidence submitted: only full trials conducted on humans were potentially sufficient to support the claims, and what Boots supplied was a single epidemiological study which suggested that sunlight — not blue-violet light in particular — might be a risk factor for early age-related macular degeneration.
And on the product itself: the documentation included one illustrative graph, and the ASA noted that graph indicated 20% of harmful blue light was filtered by the coating, meaning 80% would still enter the eye.
Twenty per cent. Sold on protecting your retina.
Two years later the General Optical Council fined the company and made findings of misconduct and impaired fitness to practise; the company admitted the particulars, said its internal approval process hadn’t been followed, and stopped selling the lenses. A BBC Watchdog investigation had found some practices still making the claims through in-store leaflets after the ASA ruling. Reports of the fine’s size disagree — most sources say £40,000, one says £400,000 — so take the fact of the sanction rather than the number.
The one positive objective result
Two trials measured critical flicker-fusion frequency, an objective marker of visual fatigue.
Singh 2021 — double-masked, prospectively registered, allocation concealed, 120 symptomatic computer users — found no difference. MD −1.13 Hz, 95% CI −3.00 to 0.74, P = 0.24. Disclosure, because it cuts the way disclosures rarely do: two of the review’s authors ran that trial. Its risk-of-bias assessment was done by two other review authors, and the conflict, had it operated, would have pushed towards a positive result for their own work. It came out null.
Lin 2017, with 36 participants, found a significant benefit for high-filtering lenses. Cochrane’s description of that trial, in its own words: it received industry funding, did not register the clinical trial, and did not mask outcome assessors. It also published no numeric data — the reviewers estimated roughly a 2 Hz change by looking at the plots, and noted that although described as statistically significant, the clinical significance of a change that small remains unclear.
And the measure itself may not mean anything. Cochrane cites two recent studies that looked for an association between critical flicker-fusion frequency and actual visual fatigue symptoms and found none, concluding that using it as a surrogate for symptoms may not be appropriate.
So: the single positive objective finding in this literature comes from an unregistered, unmasked, industry-funded trial that didn’t publish its numbers, using an instrument that doesn’t track the symptom it stands in for. Everything in that sentence is a fact stated by Cochrane’s own review. None of it is an accusation about anyone’s intent.
What the trials found, and what nobody measured
On the symptom people actually buy these for, the review’s answer is flat: blue-light filtering lenses may make no difference to eye strain over the short term compared with ordinary lenses, and probably little or no difference to best-corrected visual acuity. That is the primary outcome of the largest synthesis in the field, and it is the finding the marketing is quietest about.
Then there’s the list of what the 17 trials, 619 people between them, did not evaluate at all: contrast sensitivity, colour discrimination, discomfort glare, macular health, serum melatonin levels, and overall patient satisfaction.
The retinal protection claim — the one the ASA ruled on, the one that sells the product — has no trials. Not weak trials. None.
The epidemiology doesn’t help it either: 10 of the 12 major population-based studies looking for a relationship between light exposure and age-related macular degeneration did not report a positive association.
There’s also a mechanism nobody mentions. Your crystalline lens yellows with age and becomes its own blue-light filter, exponentially, such that by 50 only about 20% of short-wavelength visible light still reaches the retina. Cochrane’s dry observation is that it is unclear how these lenses might benefit anyone whose lens has already yellowed.
On daytime alertness — the closest thing here to a cognitive outcome — the review found the effects unknown: two trials, 42 participants between them, evidence of very low certainty. One study in insomnia patients has reported improved neurocognitive performance with blue-blocking glasses despite minimal change on actigraphy, but that is a single study in a clinical population, and I know it only through another paper’s description of it. Nobody has tested whether these lenses change how well a healthy person thinks in the way you’d want tested. If you’re hoping for that, there is no usable evidence in either direction.
Night Shift was tested. It failed.
The intervention most people actually use isn’t glasses, it’s the setting on the phone. It has been tested properly, and the design was better than the glasses trials.
167 undergraduates aged 18 to 24 were randomised to one of three conditions for the hour before bed, seven consecutive nights: iPhone with Night Shift enabled, iPhone with Night Shift disabled, and no phone at all. Wrist actigraphy, the Sadeh algorithm, outcomes of sleep onset latency, duration, efficiency and wake after sleep onset. Phone users had an app installed to verify they actually used it. Everyone was asked to spend at least eight hours in bed (Duraccio et al., Sleep Health 2021;7(4):478–484).
Across the full sample there were no differences attributable to Night Shift. The senior author, quoted in press coverage of the trial: Night Shift is not superior to using your phone without Night Shift, or even to using no phone at all.
That third arm is what makes this worth reporting and it’s what coverage drops. Abstinence didn’t beat the phone either. The honest reading isn’t that phones are harmless before bed — it’s that one hour of phone use, in young adults given eight hours in bed, wasn’t detectable by actigraphy in any direction. A trial that can’t find the effect everyone assumes is large is telling you something about the trial as much as about the phone.
One objection deserves answering before you raise it: Night Shift shifts colour far less aggressively than dedicated software. One developer’s estimate puts f.lux at roughly seven times stronger at default settings — which means a null for Night Shift isn’t automatically a null for heavier filtering. That estimate is commentary rather than measurement, and I can’t do better than flag it.
And there is a study of f.lux, in night-shift workers, reporting improvements in subjective alertness (P < 0.001), objective alertness (P < 0.05), working memory (P = 0.008) and sleep quality (P = 0.008). It was a before-and-after design with no control group. The weakest design in this entire body of evidence produced its strongest reported results — which is the same pattern as the sleep trials, arriving from a different direction.
The mechanism is real. The product targets the wrong term.
Here’s what makes this more interesting than a debunk: blue light genuinely does suppress melatonin. That part isn’t marketing.
But what predicts the suppression isn’t “how blue.” It’s melanopic equivalent daylight illuminance — how much light arrives, weighted by how much of it your melanopsin system can actually see. In a study that controlled melanopic irradiance independently of display luminance and colour in 72 men, low-melanopic light shortened time to fall asleep, attenuated melatonin suppression and advanced melatonin onset, with dose-dependent effects.
A filter changes one term in that. Room brightness changes the other, usually by more — and there is a trial that demonstrates it directly.
Sixteen participants were exposed, in a randomised crossover, to standard fluorescent light and blue-depleted LED light at matched illuminance. At 50 lux, removing the blue mattered enormously: melatonin suppression fell from 54.3 ± 5.9% to 32.5 ± 7.6%, P = 0.001. At 90 lux — roughly a bright computer screen — it made no difference whatsoever: 71.4 ± 5.7% against 70.1 ± 7.2%, P = 0.73.
Above a certain brightness, the colour stops mattering. That is the whole argument of this section, measured rather than modelled.
How bright is your evening: in a study that built individual dose–response curves across seven light levels for each participant, nearly half of homes had light bright enough to suppress melatonin by 50%. Measured separately, screens run at roughly 80 lux for a computer, 40 for a tablet, 20 for a phone.
Two more things from that home-lighting study. The range of individual responses to the average home ran from 0% to 87% suppression — the same room does wildly different things to different people. And the circadian system isn’t a simple photon counter: duration, timing, and your light history earlier that day all change the answer. Sensitivity to evening light is lower when the preceding daytime exposure was higher.
Which means the honest version of “what to do about screens at night” is not a product. It’s a set of variables, most of which are free to change, and one of which — how bright your rooms are after dark — nobody is selling you anything for.
What actually to do
Turn down the room before you turn down the screen. Half-maximal melatonin suppression sets in below 30 lux. A phone runs about 20, a computer about 80 — and nearly half of homes are already bright enough on their own to halve melatonin before anyone opens a laptop. The screen is not nothing. The room is bigger, and it’s the one nobody sells you a filter for.
Get daylight during the day. Prior light exposure reduces evening sensitivity, so a bright morning buys tolerance for a lit evening. Mechanistically supported; nobody has tested it as a packaged intervention, so treat it as a good bet rather than a finding.
For eye strain, get your eyes examined rather than filtered. Rosenfield, who ran the test of the 20-20-20 rule described below, puts it as a clinical position rather than a trial result: the first-line treatment for digital eye strain should be a full eye examination to detect and correct uncorrected refractive error or a binocular vision anomaly.
About the 20-20-20 rule itself. It was tested — 30 subjects, four sessions, with and without scheduled 20-second breaks — and had no significant effect on symptoms (P = .70), reading speed (P = .93) or task accuracy (P = .55). There is a study pointing the other way and it should be said: 29 symptomatic computer users given personalised on-screen reminders took more breaks and reported less eye strain and dry eye after two weeks — but there was no control group, and the improvement was gone a week after the reminders stopped. The one properly controlled test found nothing, and the study that found something had no control group. Make of it what you like; just don’t treat it as established.
If you already own the evening pair, keep wearing them if you like them — with one condition. Nine trials (333 people) assessed adverse events and five reported none at all. Where they were reported: headache in one trial at 8%, lowered mood at 5%, increased depressive symptoms in one trial at 17%, discomfort wearing the glasses at 22% across two — and two trials reported more adverse events in the control arm than the intervention arm. Certainty: low. One trial concluded the discomfort was probably the frames, which came in a single size.
The condition is the only genuine harm signal in this literature, and it’s about timing. In the matched-illuminance trial above, auditory reaction times were significantly slower under the blue-depleted condition half an hour before bed; in the melanopic study, low-melanopic light reduced alertness as well as advancing melatonin. If you put heavy blockers on at eight and keep working until eleven, you’re trading alertness you wanted for melatonin you weren’t going to use for three hours. Wear them when you’ve stopped working, not while you’re still at it. And since amber lenses blocking most of the blue distort colour, don’t wear them for anything where colour matters — that one is caution rather than a finding, since nobody has studied it.
As for the day pair: the trials say they don’t reduce eye strain, and nothing has ever tested the retinal protection. Don’t buy the next pair expecting either.
The thing worth keeping from all this isn’t about glasses.
Six trials asked the same question about the same product. Which three found a benefit was determined not by the lenses but by whether the people in the study knew what they were wearing — and even the review most favourable to the glasses found the benefit sitting mainly in what people reported rather than in what was measured. When you next read that something improved sleep, or energy, or focus, ask who knew.
About this article
Written by Drew Anton. Drew covers behaviour, sleep, and the compounds people take for both — doses, timing, and what the label does and does not tell you. Not a physician or research scientist — reads the primary literature closely and refuses to round up.
Medical review: None. NeuriFuel does not currently have a licensed clinician on the editorial team, and this article has not been medically reviewed. We state this rather than implying an authority we do not have. See our About page for our full methodology.
Sources: Fourteen, comprising one Cochrane systematic review of 17 randomised trials, two further systematic reviews reaching different conclusions, one randomised three-arm trial of a phone display setting, one uncontrolled before-and-after study of screen filtering software, three studies of evening light and melatonin, one randomised test of scheduled screen breaks, one uncontrolled study of screen-break reminders, one advertising regulator’s ruling and one professional regulator’s determination, and two modelling papers on circadian photoreception. The Cochrane review was read in full, including its risk-of-bias assessments, which is where the pattern described in the opening section comes from. Funding and competing interests were retrievable for two studies and both are reported inline, including for the trial with the positive result and for the trial run by two of the review’s own authors; for the remainder — including the melanopic irradiance study, the home lighting study, the Night Shift trial and both dissenting reviews — they were not available in the sources I could reach, and that absence is marked rather than assumed clean. The 2020 meta-analysis and one cognitive finding in insomnia patients are reported as described in other papers rather than read directly, and are identified as such in the text. The regulator’s ruling is quoted from four independent secondary reports that agree with one another; the ruling document itself was not retrieved.
Corrections: Found an error? Write to hello@neurifuel.com with a source and we will fix it and log the correction.
Last updated: 7 August 2026
References
- Singh S, Keller PR, Busija L, McMillan P, Makrai E, Lawrenson JG, Hull CC, Downie LE (2023). Blue-light filtering spectacle lenses for visual performance, sleep, and macular health in adults. Cochrane Database of Systematic Reviews 2023, Issue 8, Art. No. CD013244. DOI: 10.1002/14651858.CD013244.pub2. PMID: 37593770. Two review authors were also authors of one included trial; risk-of-bias assessment for that trial was performed by two independent review authors. Stated inline in the text.
- Hester L, Dang D, Barker CJ, Heath M, Mesiya S, Tienabeso T, Watson K (2021). Evening wear of blue-blocking glasses for sleep and mood disorders: a systematic review. Chronobiology International 38(10):1375–1383. DOI: 10.1080/07420528.2021.1930029. PMID: 34030534.
- Luna-Rangel FA et al. (2025). Efficacy of blue-light blocking glasses on actigraphic sleep outcomes: a systematic review and meta-analysis of randomized controlled crossover trials. PMC12668929. PROSPERO CRD420251034611. Source for the three-trial, 49-participant figure, for the characterisation of Shechter et al. 2020, and for the clinical-subgroup synthesis.
- Duraccio KM, Zaugg KK, Blackburn RC, Jensen CD (2021). Does iPhone Night Shift mitigate negative effects of smartphone use on sleep outcomes in emerging adults? Sleep Health 7(4):478–484.
- Advertising Standards Authority (2015). Ruling on Boots Professional Services Ltd t/a Boots Opticians Ltd, published 28 October 2015. Quoted here from four independent secondary reports; the ruling document was not retrieved directly.
- General Optical Council (2017). Determination in the case of Boots Opticians, published 26 May 2017.
- Höhn C et al. (2023). Melanopic irradiance defines the impact of evening display light on sleep latency, melatonin and alertness. Communications Biology. DOI: 10.1038/s42003-023-04598-4. PMC9974389. Also the source for the device illuminance figures.
- Blue-depleted versus standard light at matched illuminance: melatonin suppression and reaction time. PMC5536841. ClinicalTrials.gov NCT01586039.
- Evening home lighting adversely impacts the circadian system and sleep (2020). Scientific Reports. DOI: 10.1038/s41598-020-75622-4.
- Phillips AJK et al. (2019). High sensitivity and interindividual variability in the response of the human circadian system to evening light. PNAS.
- Johnson S, Rosenfield M (2023). 20-20-20 Rule: Are These Numbers Justified? Optometry and Vision Science. PMID: 36473088. Authors reported no financial conflict of interest.
- Personalised screen-break reminders and digital eye strain (2023). Contact Lens and Anterior Eye. Uncontrolled design.
- Effect of f.lux screen filtering software on alertness, working memory and sleep quality in night-shift workers. PMC6717920. Pre-post design, no control group.
- Giménez MC et al. (2022). Predicting melatonin suppression by light in humans. Journal of Pineal Research. DOI: 10.1111/jpi.12786.