Feeding fish, spotting disease early, and knowing exactly how much fish is swimming in a pen used to depend on human eyes and educated guesses. Today, AI in aquaculture is turning those guesses into measured decisions. Underwater cameras, sensors, and machine-learning models now watch fish around the clock, cutting feed waste, catching outbreaks before they spread, and helping farmers plan harvests with far more confidence.
Aquaculture is the fastest-growing food-production sector in the world, and it already supplies more than half of the seafood people eat. As farms scale up, manual monitoring simply cannot keep pace. This guide explains, in plain terms, how fish farms use artificial intelligence, where it delivers the biggest returns, and what its limits are.
How do fish farms use AI?
Fish farms use AI mainly through underwater cameras and sensors that feed images and water data into machine-learning models. These models estimate fish weight, count parasites, detect disease symptoms, and decide when and how much to feed. The goal is fewer losses, less wasted feed, and healthier fish.
In practice, AI in aquaculture clusters around four jobs: feeding, health monitoring, biomass estimation, and water-quality prediction. Each one replaces a slow, error-prone manual task with a continuous, data-driven one.
Smarter feeding
Feed is the single largest cost on most fish farms, often 40 to 60 percent of the total. Overfeeding wastes money and pollutes the water; underfeeding slows growth. AI-driven feeders use cameras to watch how actively fish are eating and adjust the amount of feed in real time, stopping when appetite drops.
Farmed salmon already have one of the lowest feed-conversion ratios in animal agriculture — roughly 1.1 to 1.5 pounds of feed per pound of body mass. AI helps push that efficiency further by matching feed delivery to real appetite instead of a fixed schedule.
Biomass and growth estimation
Knowing the average weight and total biomass in a pen is essential for planning treatments, feed, and harvest dates. Traditionally this meant netting and weighing a sample of fish — stressful for the animals and slow for the farmer. Computer-vision systems now estimate weight from stereo camera images without touching a single fish.
Commercial platforms such as Aquabyte report biomass-estimation accuracy above 90 percent and can forecast growth up to two weeks ahead, letting farms time harvests to market demand.
How does AI detect sea lice and disease in salmon farming?
AI detects sea lice and disease by running underwater or microscope images through deep-learning models trained on thousands of labelled examples. The models flag parasites, lesions, and abnormal swimming far faster than manual inspection, so farmers can treat problems before they spread across a pen.
Sea lice are the costliest problem in salmon farming. The global salmon industry spends an estimated $1 billion a year managing and treating them, and heavy infestations can cut biomass growth by up to 16.55 percent per production cycle. Counting lice by hand is slow and easy to get wrong.
A July 2026 study led by the Norwegian University of Science and Technology (NTNU) and Wageningen University in the Netherlands showed how much AI can help. Researchers built a video microscope, captured more than 120,000 images of louse larvae in seawater, and trained models to tell parasites apart from other particles. In one large, complex sample, trained biologists needed over 30 hours across several days to identify 82 percent of the larvae — the AI model found 97.5 percent of them in about 30 minutes.
Detecting lice larvae before they attach to fish gives farmers a head start that manual counting can rarely provide.
Beyond parasites, AI watches for behavioural warning signs. Machine-learning models trained on video can spot erratic swimming, crowding, or reduced feeding — early symptoms of stress or disease — and alert staff before losses mount. This kind of monitoring builds on the same computer-vision advances now used in AI health monitoring on dairy farms.
How does AI keep fish-farm water healthy?
AI keeps farm water healthy by continuously analysing sensor data on oxygen, temperature, pH, and salinity, then predicting dangerous shifts before they occur. Instead of reacting to a fish kill, farmers get an early alert and can adjust aeration, feeding, or stocking in time.
Low dissolved oxygen is one of the deadliest risks in intensive fish farming, and it can develop within hours. Predictive models learn the patterns that precede an oxygen crash — often tied to temperature, feeding, and weather — and warn operators early. The same predictive approach underpins precision farming in agriculture, where sensors and AI guide decisions field by field.
What are the benefits and limits of AI in aquaculture?
The upside is clear: less wasted feed, earlier disease detection, more accurate harvest planning, and better fish welfare. Automated counting also frees skilled staff from tedious sampling to focus on judgement calls.
But AI in aquaculture has real limits. Underwater cameras struggle in murky water, low light, and strong currents. Models trained on one species, cage design, or region often perform worse elsewhere, so they need local data and regular retraining. Cameras, sensors, and software also carry upfront costs that can be hard for small farms to justify. According to industry researchers, AI works best as a decision-support tool for skilled farmers — not a replacement for them.
Key use cases at a glance
- Feeding: cameras gauge appetite and trigger feeders, cutting waste and pollution.
- Health: vision models flag sea lice, lesions, and abnormal behaviour early.
- Biomass: stereo cameras estimate weight and total stock without handling fish.
- Water quality: sensors plus AI predict oxygen and temperature risks in advance.
- Planning: growth forecasts help time treatments and harvests to demand.
The road ahead for smart fish farming
As camera hardware gets cheaper and models get better at handling murky, moving water, expect AI in aquaculture to spread from large salmon operations in Norway, Chile, and Scotland to shrimp, tilapia, and catfish farms across Asia, the Americas, and Africa. The direction of travel is consistent with AI adoption across other industries: start with the most expensive, error-prone manual task, then automate outward from there.
For fish farmers, the message is practical. AI will not replace the stockperson’s eye, but it can extend it — watching every pen, every hour, and catching the problems that used to be found too late.
Frequently Asked Questions
What is AI used for in aquaculture?
AI in aquaculture is used mainly for feeding optimisation, disease and sea lice detection, biomass estimation, and water-quality prediction. Cameras and sensors feed data to machine-learning models that help farmers cut waste, reduce losses, and plan harvests more accurately.
How does AI count sea lice on salmon?
AI counts sea lice by analysing underwater or microscope images with deep-learning models trained on thousands of labelled examples. A 2026 NTNU and Wageningen study found such a model identified 97.5 percent of louse larvae in about 30 minutes, versus 82 percent over 30-plus hours for trained biologists.
Does AI reduce feed costs on fish farms?
Yes. Because feed is often 40 to 60 percent of farm costs, AI-driven feeders that match feed to real appetite can cut waste significantly. They dispense feed only while fish are actively eating, reducing both cost and water pollution.
Can AI replace fish farmers?
No. AI works best as a decision-support tool that handles continuous monitoring and counting, freeing farmers for judgement calls. Cameras struggle in murky water and low light, and models need local data, so skilled human oversight remains essential.
Which fish species benefit most from AI farming?
Salmon farming has adopted AI most widely because sea lice and disease make it high-value and high-risk. As technology gets cheaper, shrimp, tilapia, and catfish operations across Asia, the Americas, and Africa are increasingly adopting AI monitoring too.
