Key takeaways:
- Sourdough’s biggest industrial challenge isn’t flavor, it’s managing microbial variability as production scales across sites, climates and ingredients.
- New research suggests AI-backed analysis could help bakers understand and control fermentation behavior without abandoning sourdough’s natural identity.
- While lab-designed microbial communities aren’t plug-and-play yet, better data and modeling could reduce waste, improve consistency and support global sourdough growth.
Sourdough has never been a neat fit for industrial baking. The wild fermentation, shifting microbial mix and sensitivity to environment that give it depth and character are the same traits that make it unpredictable once production scales.
That tension is no longer theoretical. Sourdough is now baked, frozen, shipped and sold at volume, expected to perform the same way week after week. For large bakeries, that brings familiar headaches: long fermentation times, uneven performance and food-safety risk when conditions drift.
A new academic review suggests artificial intelligence (AI) could help bakers get a better handle on those variables. Published in Trends in Food Science & Technology, the paper from researchers from Cardiff University, Shanxi University and Jiangnan University review how AI could be combined with advanced microbial analysis to make sourdough fermentation more predictable at scale.
“For thousands of years, sourdough fermentation has relied on complex communities of lactic acid bacteria and yeasts,” said Dr Faizan Sadiq, coauthor and assistant professor in microbial biofilms at Cardiff University. “That diversity produces distinctive breads, but it also makes industrial production inconsistent and harder to control.”
In a large bakery, those inconsistencies don’t stay theoretical for long. Swap a flour lot, tweak hydration or get a warmer-than-usual day on the line and fermentation can behave differently. When that happens across multiple plants, it can throw off planning, push up waste and leave retailers dealing with bread that isn’t quite consistent week to week.
Why sourdough won’t sit still
Traditional sourdough starters develop through spontaneous fermentation. Microbes come in with the flour and water, pick up others from the environment and shift over time as the dough is repeatedly refreshed. A handful of lactic acid bacteria and yeasts tend to dominate, but the balance is never completely fixed.
That flexibility is part of sourdough’s appeal. It’s also why it resists standardization. Small bakeries can adjust by feel. Large operations don’t have that luxury, especially when production spans different climates and raw material sources.
Industrial sourdough systems already rely on defined starter cultures to reduce variability. Even so, fermentation remains slow and sensitive. Changes in ingredient quality or process conditions can still affect both timing and finished quality.
One reason sourdough remains difficult to control is that much of its microbial activity has historically gone unseen. Bakers know what sourdough does, but not always why. Until recently, measuring microbial interactions in detail wasn’t practical outside research settings.
That’s starting to change. Multi-omics techniques now allow scientists to see which microbes are present in a starter and what they’re doing at a functional level – from gene expression to metabolite production. The drawback is scale. The result is a flood of data that’s hard to make sense of without computational help. Machine learning can sort through it, identify the microbes that matter most and show how fermentation shifts when conditions change.
“Sourdough has surged in popularity, but its natural variability makes consistent, large-scale production challenging,” Dr Sadiq said. “Integrating AI with multi-omics gives us a way to design more stable microbial communities without losing the qualities people associate with sourdough.”
A central concept in the review is synthetic microbial communities, or SynComs. These are deliberately assembled combinations of bacteria and yeasts selected for specific fermentation roles. Unlike wild starters, they’re designed to behave in predictable ways, whether that’s steady acidification, reliable leavening or controlled flavor development.
What matters for bakers now
The appeal of AI is practical rather than philosophical for larger bakery manufacturers. More predictable fermentation can shorten production cycles, reduce batch failure and tighten food-safety control. Starters that perform consistently across regions could also ease some of the operational strain that comes with globalized production.
There are sustainability implications as well. Better-controlled fermentation means less energy spent on stalled or extended processes and fewer raw materials written off. Improved stability can also support shelf life, which matters for frozen and long-distribution products.
The researchers also flag opportunities in product development. Microbial communities could be tailored for alternative grains, ancient wheats or reduced-gluten formulations – areas where fermentation performance is often less forgiving and trial-and-error is costly.
Prof Guohua Zhang, researcher at Shanxi University in Taiyuan, China, noted that the same principles extend beyond bread. “We reviewed evidence showing how AI and multi-omics could help optimize sourdough and other complex fermentations,” he said.
The authors are careful not to promise quick wins. Translating lab-designed microbial communities into open bakery environments remains difficult, especially where temperature, humidity and ingredient quality vary day to day. The paper suggests redundancy – using multiple strains capable of the same function – may be necessary to keep systems stable.
For now, the research reads less like a roadmap and more like a reframing. Sourdough’s wild side has long been treated as the price of authenticity. This work suggests that variability doesn’t have to disappear – but it may be better understood and better managed, as sourdough continues its shift from craft staple to global category.
Study:
Enhancing sourdough fermentation with AI and multi-omics: From natural diversity to synthetic microbial community. Yujuan Yu, Jiale Wang, Faizan Ahmed Sadiq, et al. Trends in Food Science & Technology, Volume 165, 2025, 105233, ISSN 0924-2244, https://doi.org/10.1016/j.tifs.2025.105233.
