AI recycling contamination detection uses cameras on the collection truck to spot non-recyclable items as each cart is tipped into the hopper, then ties every event to the address it came from. That turns contamination from a load-level problem you discover at the MRF into a household-level problem you can fix with targeted education, warnings or fees. Cities and haulers stop paying for the same few carts on every route.
Key takeaways
- The MRF can tell you a load was dirty, but not which households caused it. Contamination surcharges often apply to the whole load.
- Cart-level feedback works: a six-week cart tagging program in Smyrna, Delaware cut contamination from 34% to 5%.
- AI detection at the hopper delivers that feedback on every route, every day, with photo evidence tied to each address and no extra staff.
Why is the MRF the wrong place to find contamination?
By the time a load reaches the materials recovery facility, it's too late to know who caused the problem. The MRF can tell you the load was dirty. It can't tell you which 30 houses on a 900-stop route put bagged trash, hoses or propane bottles in the recycling cart.
That matters because contamination is expensive and it's getting attention. In June 2026, Lawrence, Kansas reported a 17.5% contamination rate in its spring recycling audit, up from 15.1% the previous fall. Auditors pulled more than 3,700 pounds of trash out before sorting. Under the city's processing contract, a rate above 10% triggers a per-ton surcharge, and a rate above 15% adds a second one (City of Lawrence).
Contamination also undercuts the whole program. The Recycling Partnership's State of Recycling report found that only 21% of residential recyclables in the U.S. are actually recycled, and only 43% of households participate (Waste360). Every contaminated load makes the good material in it harder to sell.
“Contamination fees at MRFs can destroy a hauler's bottom line. Typically MRFs charge $50 to $100 per ton, and in many cases they charge on the entire load. Some MRFs have a tier system with escalating fees as the percentage of contamination increases. It's not uncommon for a truck to be turned away to the landfill if contamination levels are too high.”
Barry Saunders, President, WasteVision AI
Does cart-level feedback actually change behavior?
Yes, and the evidence is strong. Cart tagging programs, where staff check carts at the curb and leave a tag explaining what was wrong, have been running for years with The Recycling Partnership's support:
- Smyrna, Delaware: a six-week “Feet on the Street” program run by the Delaware Solid Waste Authority, the town and The Recycling Partnership took contamination from 34% on the first scan to 5% on the last (DSWA).
- Washington, D.C.: a 2018 tagging pilot cut plastic bags and film in the curbside stream by 20% (The Recycling Partnership).
The lesson is simple: residents respond when feedback is specific to their cart. The problem is scale. Manual cart tagging needs people walking routes ahead of the truck, so most cities can only afford it for a few weeks on a few routes.
How does AI contamination detection work on the truck?
WasteVision AI's Contamination Detection gives you cart-level feedback on every route, every collection day, without extra staff. It runs on cameras already on the truck:
- Capture at the hopper. Cameras photograph the recycling as it falls into the hopper during the lift.
- Detect. AI models flag prohibited items such as plastic bags, styrofoam, construction debris and yard waste.
- Attribute. Each event is tied to the cart's address using WasteVision AI's service verification, with GPS, timestamp and photos.
- Act. The hauler or city uses the evidence for education, warnings or fees.
Every event includes two images: one of the cart at the curb before pickup and one of the contamination at the hopper. That evidence chain is what makes it fair to send a resident a notice, and hard for anyone to argue it was a neighbor's cart. It's the same lift-by-lift record we covered in our earlier post on ending “you missed my pickup” disputes.
“Contamination detection can be performed on rear loaders, front loaders and side loaders. Placing the camera in the right position in the hopper is crucial for contamination detection.”
Barry Saunders, President, WasteVision AI
What do you do with household-level contamination data?
The data is only useful if it leads to action. Most programs work best with a step-by-step approach, and WasteVision AI supports a three-strike rule out of the box:
| Step | What happens | Goal |
|---|---|---|
| First event | Resident gets a postcard or email showing the photo of their cart, the contaminant and the address | Education |
| Repeat event | Second notice with a clear warning and a reminder of what belongs in the cart | Behavior change |
| Third strike | Fee, a service change or another step set by your contract or ordinance | Accountability |
| Ongoing | Reports by route, household, contaminant type and frequency | Target outreach where it's needed |
Residents can also get notices automatically through Service Notifications, which attach the photo and the reason when the AI flags contamination. Reports and repeat-offender counts sit in the WasteVision AI cloud platform, and the open API can send events to billing or ERP systems.
“In my opinion, it's best for cities to lead with educating their residents on proper recycling rules. WasteVision's system keeps track of the number of contamination violations at each residence, so if a city decides to fine residents, it can identify the highest violators.”
Barry Saunders, President, WasteVision AI
What results have cities seen?
WasteVision AI's published case study with a mid-sized Texas municipality combined real-time curb detection with targeted resident outreach on its residential program. The city significantly reduced contamination rates and now has a platform it can scale across more routes.
Is AI detection a replacement for cart audits?
No. Audits and AI detection answer different questions. A periodic audit tells you how dirty the stream is overall and what's in it. AI detection tells you which carts are causing the problem, every day, on every route. Used together, the audit sets the baseline and the AI shows whether your outreach is moving it.
Frequently asked questions
What is AI recycling contamination detection?
It's software that analyzes truck camera images at the hopper to identify non-recyclable items in each cart and link each event to the household's address with a photo, timestamp and GPS location.
Do drivers have to inspect carts or take photos?
No. Cameras capture each lift automatically and the AI flags contamination, so drivers stay on schedule.
What happens to residents who contaminate their carts?
That's up to the city or hauler. WasteVision AI supports education notices with photos, warnings and a three-strike rule that can lead to fees or service changes.
Does cart-level feedback really reduce contamination?
Yes. In Smyrna, Delaware, a six-week cart tagging program cut contamination from 34% to 5%. AI detection makes that kind of feedback possible on every route, every day.
See which carts are contaminating your recycling
WasteVision AI helps cities and haulers find contamination at the cart, reach the right households and track results over time. Request a demo to see contamination detection on your own routes.
Sources
- City of Lawrence, Kansas, Latest audit finds 3,700 pounds of trash in Lawrence recycling carts (June 16, 2026)
- Waste360, State of Recycling Report Highlights Issues in U.S. Residential Collection (Jan. 10, 2024), reporting The Recycling Partnership's State of Recycling data
- Delaware Solid Waste Authority, Town of Smyrna's “Feet on the Street” Cart Tagging Program Yields Remarkable Results (Dec. 7, 2023)
- The Recycling Partnership, D.C. DPW Launches Cart Tagging Campaign and Contest to Reduce Recycling Contamination (March 23, 2021)
- WasteVision AI, Contamination Detection
- WasteVision AI, Case Studies