Many of the world’s most expensive perfumes begin with a wounded tree. When some Aquilaria trees are damaged, they produce a dark, fragrant resin that becomes agarwood, or oudh – an ingredient so valuable that wild trees have been illegally felled in its pursuit.

The result has devastated wild Aquilaria populations. But what if perfumers could create the smell without cutting down the tree?

With the help of AI, scientists are learning to predict how molecules that have never been made will smell, reconstructing scent-producing genes from flowers believed to be extinct. Others have collected microbes from clouds and used them as a starting point for new perfumes.

Now scientists and biotechnology companies say advances could soon allow the fragrance industry to reproduce scents and compounds from endangered plants without repeatedly harvesting them.

“Being able to recreate the scent of an endangered plant would be a huge win for both the environment but also the brands,” said Alex Wiltschko, the founder of Osmo, a Google Brain spin-off that has raised $130m to develop artificial intelligence capable of predicting and recreating smells.

Osmo has recreated the scent of a rare tree from a piece of bark “the size of a finger”, Wiltschko said. A spectrometer analysed the sample, its AI reconstructed the chemical formula and the scent was reproduced in its laboratory.

BioHarvest Sciences says it has created a stable cell culture from an endangered south-east Asian fragrance plant, allowing its valuable compounds to be produced without cultivating or harvesting the original plant.

The plant has not been named for commercial reasons but BioHarvest has announced an agreement anticipating production of 20 tonnes of the fragrance composition, with limited production potentially beginning next year.

“The challenge is that human demand can be effectively unlimited while biological supply is not,” said BioHarvest’s chief executive, Zaki Rakib. “We believe biotechnology can help close that gap, not by replacing nature but by allowing us to use what nature has given us without continually putting the source under pressure.”

But conservationists and perfumers say such claims risk giving advances in fragrance technology an environmental gloss they have not yet earned.

“Calling this conservation is greenwashing,” said Mandy Aftel, the pioneering natural perfumer, author and founder of the Aftel Archive of Curious Scents. “Making a synthetic version of an endangered plant’s smell doesn’t save the plant.” She said producing a substitute did not stop people buying the original – or address other causes of a species’ decline.

Rhian Smith, of the Royal Botanic Gardens, Kew, agreed. She pointed to the Aquilaria tree. AI could help design an alternative to oudh, she said, but wild-sourced material remained highly prized. “It doesn’t completely take the pressure off wild plant species,” she said.

There is also the question of motivation. Wiltschko admitted that conservation was not “an explicit priority” for Osmo’s work.

Synthetic perfume itself is nothing new: the industry has relied heavily on synthetic ingredients for more than a century, and machines have long been able to identify molecules in a scent.

“What’s changed is what happens after the data comes off the machine,” Wiltschko said. “What’s new is prediction.” He claimed Osmo had trained its AI on more than 5m human responses to smells. “That means it can predict how combinations of molecules will be perceived,” he said.

Elsewhere, AI is claimed to be changing the whole world in which perfumers work. Christina Agapakis is not a perfumer, but her work reconstructing genes from preserved flowers has produced fragrance compounds that perfumers can use to create new scents.

Another perfumer, Future Society, has gone into the clouds, making two flights in a small plane to collect living microbes. Thirteen were identified, producing 57 scent molecules, which became the starting point for Cloud Reverie, a perfume launched this year.

Even AI has its limits: a flower’s scent can contain hundreds of molecules and change according to which part of the plant they come from and the conditions in which it grows.

It is still not enough for the purists. “The smell we got was this ghostly fragment of what had been lost,” Agapakis said of her project. She could not recreate the living organism, or the soil, climate, season and ecology that produced its scent, she said.

“What we have to remember is, ultimately, however powerful AI is, a smell is not simply a formula or dataset.”