A yogurt cup looks like one of the least technical objects in the supermarket. It’s a bit of plastic, a foil lid and a printed label, and most people throw it out without a second glance.
Behind that cup, though, a surprising amount of AI and machine vision now decides what happens to it. Cameras on the filling line check whether the lid sealed properly. Further down the line, in a sorting plant, other cameras try to work out whether that same cup is food-grade polypropylene worth recycling, or just another piece of mixed plastic headed for the incinerator.
That second step is where a lot of dairy’s sustainability promises quietly fall apart, and it’s one reason packaging decisions have become far more complicated than choosing a nice shape. Suppliers such as Greiner dairy packaging now offer brands cups, tubs, lids, pails, and cardboard-plastic combinations across a wide range of formats. The hard part is choosing among them when protection, cost, regulation and recyclability all pull in different directions.
Why “recyclable” so often isn’t
Put a cup in the recycling bin, and it may still never be recycled. If the sorting facility can’t identify it quickly and accurately, it ends up in a mixed bale or gets rejected. A “recyclable” label on the pack doesn’t change what happens at the plant.
This is where AI has started to matter. UK startup Greyparrot builds AI-powered cameras that sit above conveyor belts and identify waste in real time using object-detection models. In 2024, Bollegraaf, the Dutch group that builds more recycling plants than anyone else, agreed to put that technology into its new facilities. US firm AMP Robotics pairs similar AI vision with robots that pick materials off the line.
Then there are digital watermarks, which the dairy industry should probably pay more attention to. These are codes printed across a package, invisible to people, that a camera can read in the sorting plant. They aren’t AI in the strict sense, but they serve the same purpose of giving machines enough information to sort packaging properly. The European HolyGrail 2.0 initiative, run by the brands association AIM with the Alliance to End Plastic Waste, tested the idea at industrial scale on real household packaging between August and December 2024. Sorting efficiency reached around 90% or higher in a single pass. The watermarks could also separate packaging down to individual products, which no current industrial sorting method can do.
For dairy, that granularity matters. Food-contact recycled plastic needs clean, well-separated input. If a sorting plant can tell a yogurt cup from a detergent bottle, food-grade recycling becomes realistic instead of aspirational. The project has now moved on to market-adoption pilots under a successor programme, HolyGrail 2030.
Circular design runs into real life
None of this works if the packaging itself fights the system. Monomaterial cups, lids that come off cleanly, clear disposal instructions and lighter structures all help, but only where local collection and processing can actually handle them.
A small dairy-sector transition project in the Pyrenees showed how messy that gets on the ground. Designer Mercè Rua brought together dairy producers from a valley that straddles France and Spain to talk about packaging waste. She found small producers already trying returnable, milkman-style systems who felt, in her words, like activists, with everything working against them. Her conclusion was that new materials alone won’t shift the sector. Regulation and public infrastructure need to stop favouring the linear model and start making circular options easier.
So “recyclable” isn’t the finish line. A brand needs to know whether shoppers can find the right bin, whether parts need separating first, and whether key markets actually collect the material.
The filling line has cameras now too
At the other end of the chain, machine vision has become standard on many packaging lines. Cameras check every container for a sealed lid, a straight label, the right fill level and visible contamination, at speeds no human inspector could match.
This matters more as packs get lighter. Thinner walls and lighter lids save material, but they leave less margin for error. A lightweighted cup that cracks in transit or leaks in the fridge creates food waste, and wasted yogurt has a far bigger environmental footprint than the few grams of plastic saved. Automated inspection helps brands push material reduction without quietly raising spoilage. It catches problems at the line instead of in a customer’s shopping bag.
That said, AI vision alone won’t rescue a badly designed package. It can only reject defects; it can’t prevent them.
Regulation is setting the deadline
In Europe, the clock is already running. The EU’s Packaging and Packaging Waste Regulation entered into force in February 2025 and started applying on 12 August 2026. Some requirements took effect immediately. Limits on PFAS “forever chemicals” in food-contact packaging, for example, now apply to anything placed on the EU market.
The bigger shift comes later. From 2030, packaging must meet minimum recyclability requirements, and plastic packaging faces minimum recycled-content rules. Packs will be graded on how recyclable they are, from grade A (at least 95%) down to grade C (at least 70%), and harmonised labelling rules phase in from 2028.
For multi-market dairy brands, that means a format that looks cheap today could become a liability within a few years. It also explains the growing interest in sorting technology. Recyclability on paper won’t count for much if nobody can sort the pack in practice.
Shoppers still decide in seconds
All of that happens out of sight. At the shelf, people decide quickly and react to what they see. In one European consumer survey, 59% of respondents reacted negatively to what they considered excessive plastic packaging.
That’s pushing design toward restraint. Clear product names, fewer claims, simple nutrition cues and more white space now do more work than busy graphics. QR codes and digital watermarks can also take some text off the pack. A code that explains which bin the lid goes in, or where the milk came from, earns its space. One that just opens the brand’s homepage doesn’t.
Digital printing helps too. Short runs, regional editions and seasonal designs are easier to test without committing to huge print volumes.
Where a dairy brand should actually start
Not with a trend list. A better starting point is finding where the current pack causes the most trouble: waste, complaints, damage in transit, confusion at the recycling bin, or regulatory exposure in a major market.
From there, a few questions sort out most decisions. Will the format still protect the product if it gets lighter? Can it run on existing filling lines? Can the main markets’ sorting plants identify it, with or without a watermark? And will it still be compliant when the 2030 requirements arrive?
Test changes under real conditions too: cold chains, lunch bags, home fridges, overflowing recycling bins. A pack that looks perfect in a design review can fail in any of them.
Quick answers
Q. What is the biggest change coming to dairy packaging?
In Europe, it’s the shift from “recyclable in theory” to “recyclable in practice,” driven by PPWR recyclability requirements from 2030 and better sorting technology.
Q. Do digital watermarks use AI?
Not exactly. They’re invisible codes read by cameras in sorting plants. They do the same job as AI vision, though: giving machines enough information to sort packaging accurately.
Q. Is lighter packaging always better?
No. If lightweighting causes breakage, leaks or spoilage, the extra food waste can outweigh the plastic saved.
Final Words
Dairy packaging still has to do what it always did: keep the product safe, travel well and open easily. What’s changed is that its life now runs through cameras and algorithms at both ends, on the filling line and at the sorting plant, and brands designing new packs need to account for both.
Related: How AI Vision Is Catching Filling-Line Defects Before They Reach the Capper
