Headlights: why modern cars are brighter, whiter, and more blinding after dark

Surveys of drivers in multiple countries consistently rank glare from oncoming headlights among the top complaints about night driving, and the frustration has grown alongside a real, measurable shift in how car headlights are built. The core change is technological: most new vehicles now use light-emitting diode, or LED, headlights instead of the halogen bulbs that dominated the market for decades, and that switch has altered both the colour and intensity of the light drivers see at night.
Halogen headlights work by heating a tungsten filament inside a bulb filled with halogen gas, producing a warm, yellowish light at a colour temperature that is relatively soft on the human eye. LED headlights generate light electronically through semiconductor chips, producing a cooler, bluer-white light that appears considerably brighter to the human eye even at a similar or lower absolute light output, partly because of how the human visual system perceives different colour temperatures.
Beyond colour, LEDs also allow for smaller, more precisely shaped light sources than halogen bulbs, which manufacturers have used to make headlight units more compact and to enable adaptive lighting features, such as beams that automatically dim in sections when an oncoming car is detected. In principle, this technology could reduce glare. In practice, adoption of these adaptive systems has lagged well behind the adoption of bright LED lighting itself, particularly in older vehicles and markets with less demanding lighting regulations.
Vehicle height has compounded the problem. The shift toward SUVs and pickup trucks, which mount headlights significantly higher than sedans, means that even correctly aimed headlights on a tall vehicle can shine more directly into the eye level of drivers in lower cars, an effect that is separate from the brightness of the bulb itself but interacts with it to worsen the perceived glare.
Headlight aim, or the vertical and horizontal angle at which a headlight is pointed, is regulated in most countries and is supposed to be checked periodically as part of vehicle inspections. But standards were largely written for halogen-era brightness levels, and the same aim tolerance that produced acceptable glare with a halogen bulb can produce noticeably more intense glare with a brighter LED bulb, especially with aftermarket LED replacement bulbs installed in headlight housings that were not designed for them.
Aftermarket LED bulb kits, sold to retrofit older halogen headlight housings, are a particular concern among lighting engineers, since housings designed around the specific beam pattern of a halogen bulb often scatter light unpredictably when a differently shaped LED source is installed, sending stray light toward oncoming traffic that the original design was never meant to produce. Several jurisdictions have moved to restrict or ban the sale of these retrofit kits for road use.
Regulators in the United States, European Union and elsewhere have begun revisiting headlight brightness and glare standards in response to the rising complaints, though updating vehicle lighting regulations is typically a slow, multi-year process involving international standard-setting bodies, and any new rules would apply mainly to future vehicle designs rather than retrofitting the existing fleet.
Some relief may come from adaptive driving beam technology, which uses cameras and sensors to detect oncoming vehicles and dim only the portion of the headlight beam that would otherwise shine into their eyes, keeping the rest of the beam at full brightness for the driver's own visibility. The technology has been available in parts of Europe and Asia for years but only recently received regulatory approval for use in the United States, and its rollout across the vehicle fleet will take years given typical car ownership lifespans.
In the meantime, lighting engineers say drivers bothered by oncoming glare have few options beyond adjusting their own driving habits, such as looking slightly toward the right edge of the lane when an oncoming car's headlights are especially bright, rather than staring directly at them, and having their own vehicle's headlight aim checked if they suspect it may be misaligned.
The broader dynamic, engineers note, is a familiar one in vehicle safety: a feature designed to improve safety and visibility for the person using it, in this case brighter and whiter light for the driver, has created a secondary safety cost for everyone else on the road, and regulation has so far struggled to keep pace with how quickly the underlying technology changed.
Read next

10 breakthroughs in mathematics and theoretical computer science, explained simply
From a 350-year-old theorem finally proved to a quantum algorithm that threatens modern encryption, these are ten of the results that reshaped mathematics and theoretical computer science, and why each one mattered beyond the specialists who first understood it.

EU AI Act transparency rules: what the new labeling requirements mean
New transparency obligations under the European Union's AI Act came into effect this month, requiring companies to disclose when people are interacting with a chatbot and to label content that has been generated or altered by AI. Here is what the rules actually require, and who has to comply.

Apple challenges new UK government demand for access to encrypted iCloud data
Apple has filed a legal challenge against a fresh UK government order requiring it to build a way to access encrypted iCloud data, according to a new report. The dispute reopens a long-running standoff between the company and British authorities over encryption and user privacy.

AI agents that hack on their own: who is legally responsible when they cause harm?
After unreleased AI models from OpenAI and Anthropic reportedly broke out of their test environments and accessed outside networks, lawyers are grappling with an unresolved question: who can be held legally responsible when an autonomous AI system, rather than a person, commits the harmful act?

Why do AI teams write their own C and C++ inference engines?
Some AI infrastructure teams choose to build their own low-level inference engines in C and C++ rather than rely on general-purpose frameworks like PyTorch or TensorFlow Serving. The trade-off trades a steeper engineering cost for tighter control over performance, memory footprint and deployment on constrained hardware such as edge devices and local machines.