K2-18b's Life Signal: What Webb Saw and What It Didn't

In April 2025, a team led by Nikku Madhusudhan at the University of Cambridge announced what it called the strongest hints yet of biological activity beyond the solar system. The planet was K2-18b, about 124 light years away. The molecule was dimethyl sulfide, a gas that on Earth is made almost entirely by marine life. Within months, other teams had reanalyzed the same telescope data and reached very different conclusions. The episode is the clearest case study so far of how a claim of alien life should be read, and it is not over.

The planet and the claim

K2-18b is about 8.6 times the mass of Earth and 2.6 times its radius, according to Cambridge. It orbits a red dwarf star in the zone where liquid water could exist. Earlier observations with the James Webb Space Telescope had detected methane and carbon dioxide in its atmosphere, which fits the idea of a hydrogen rich atmosphere over a water ocean, a so called Hycean world.

The 2025 paper, published in The Astrophysical Journal Letters, used Webb's mid infrared instrument, MIRI. It reported dimethyl sulfide, or a related molecule called dimethyl disulfide, or both, at a significance of three sigma, with concentrations of at least 10 parts per million by volume.

Cambridge was careful about what three sigma means. It corresponds to about a 0.3% probability that the signal arose by chance. The usual threshold for a discovery in physics is five sigma. The team said it would need 16 to 24 more hours of Webb time to get there.

Three sigma sounds like certainty. In physics it is a reason to keep looking.

What happened when others looked

The reanalyses arrived fast. The most thorough came from Kevin Stevenson and colleagues, published in The Astronomical Journal in October 2025 under a Creative Commons license. Running the data through many combinations of choices about how to process the MIRI spectrum, they found that 87.5% of the retrievals using their processing did not favor dimethyl sulfide. Only one configuration out of eight did.

They also pointed to correlated "red noise" in the data, the kind of structured instrument noise that can mimic a real feature, and showed that other molecules such as ethylene or chloroethane could explain the bumps in the spectrum about as well. Their paper summarizes several other independent reanalyses that reached similarly skeptical conclusions.

A separate preprint by Renyu Hu and colleagues in July 2025 analyzed four transits observed with Webb's near infrared instrument, NIRSpec. It found methane and carbon dioxide robustly, consistent with a water rich interior, but dimethyl sulfide only marginally, below three sigma and below about two sigma unless a haze was forced into the model. It also discussed ways dimethyl sulfide could form without life.

The biosignature problem

Even a clean detection would not settle the question of life. Dimethyl sulfide has been found on comet 67P and in interstellar space, places with no biology at all, as Stevenson and colleagues note. A molecule that life makes on Earth is not automatically a molecule only life can make on a planet with completely different chemistry.

That is why Stevenson's paper goes beyond the reanalysis and proposes a five question framework of standards of evidence for claims like this one, from whether the signal is real to whether anything other than life could produce it. It is the closest thing the field has to a checklist for the claim everyone wants to make first.

How to read the next announcement

None of this makes the original team careless. They published their significance level, said it fell short of a discovery, and asked for more telescope time, which is exactly how the process is supposed to work. The problem was the gap between "strongest hints" in a press release and the much more cautious statistics underneath, a gap that headlines then widened further.

The same pattern shows up in medicine, where a striking first result often shrinks on replication; the checks in our guide to reading a study apply surprisingly well to an exoplanet spectrum. For K2-18b, the useful questions are simple: how many sigma, from how many independent teams, with which instrument, and what else could make the same signal.

The most likely outcome is not a dramatic reversal but a slow narrowing, with more transits, more instruments and a signal that either sharpens or fades. If life is ever found on another world, it will almost certainly arrive this way, as a fight over noise in a spectrum, long before it arrives as a headline. The first claim will not be the one that counts. The fifth independent confirmation will.