Industries / Food, farming and forestry / Agriculture and farming / Seed production
Certification
Checking in the field and in the lab that seed is the variety its label names and meets set standards, then labeling and tracking it so a buyer can trust what is in the bag.
6
Milestones achieved, from 1919 to 2018.
1
To watch for: announced, not yet confirmed outside.
2
Open deliverables. Nobody has shipped them.
The road
AchievedWatch forOpen
Achieved
1919
Growers agree seed standards
In Chicago in 1919, the Canadian Seed Growers' Association and its US counterparts formed the International Crop Improvement Association to run shared standards for certified seed. It became the Association of Official Seed Certifying Agencies (AOSCA) in 1968, and in 2019 its board president, Bill Foote, gave its membership as 50 organizations in eight countries.
Credit
Canadian Seed Growers' Association and US state crop improvement associations
1939
Labels must tell the truth
The US Federal Seed Act, approved on 9 August 1939, is a truth-in-labeling law for farm and vegetable seed shipped between states. The label must give purity, germination percentage, noxious weed seeds per pound, any chemical treatment, the kind and variety, and the shipper's name and address. Each year USDA's Agricultural Marketing Service plants samples that state officials collect from stores and warehouses, to check that the plants match the variety on the label.
Credit
US Congress; USDA Agricultural Marketing Service
1958
Certified for trade abroad
The OECD Seed Schemes, set up in 1958 and run by the Organisation for Economic Co-operation and Development since 1961, certify that seed sold across borders is the variety claimed and is pure, with field inspections as part of the process. The schemes are split into eight groups of crops, and 64 countries took part in 2025, by the count of the US program manager.
Credit
OECD and participating national authorities
1966
Only certified cereal seed sold
Council Directive 66/402/EEC of 14 June 1966 ruled that cereal seed could be marketed in the European Economic Community only if it had been officially examined and certified as "basic seed" or "certified seed". It drew on the FAO's maize seed standards and the OECD scheme, set conditions for purity, germination and health, and set up Community test plots to check certified seed after sale.
Credit
Council of the European Economic Community
2017
Shop seed often not genuine
Tessa Bold, Kayuki Kaizzi, Jakob Svensson and David Yanagizawa-Drott bought hybrid maize seed and fertilizer in local shops in Uganda and tested them in the lab and in trial plots. The hybrid seed contained "less than 50% authentic seeds", fertilizer was missing 30% of its nutrient, and farmers using average shop quality would on average lose money. Their 2015 working paper was published in the Quarterly Journal of Economics in 2017. A later study, Barriga and Fiala (2020), points instead to seed spoiling along the supply chain.
LessonCertifying seed when it leaves the company does not protect the farmer if it is swapped or spoiled before it reaches the shop shelf.
Credit
Bold, Kaizzi, Svensson and Yanagizawa-Drott
2018
A scratch code per packet
In 2018 the Kenya Plant Health Inspectorate Service (KEPHIS) required every seed packet sold in Kenya to carry a scratch-off sticker with a unique code that a farmer can text to check the seed and to see whether the code was used before. Researchers from J-PAL and CEGA (Eric Hsu, Edward Miguel and Anne Wacera Wambugu) still found more than 40% of packets in retail markets problematic in one or more ways. In a randomized trial across 282 rural market areas in four Kenyan counties in 2020, showing farmers how to use the codes raised their maize yields by 55 kg per acre (6.5%), but sellers did not improve their seed; one in six left the market instead.
LessonA check that buyers can run moves buyers faster than it moves sellers.
Credit
KEPHIS (system); Hsu, Miguel and Wambugu (evaluation)
Watch for
India's QR-coded seed law
India's agriculture ministry published a draft Seeds Bill in November 2025 to replace the Seeds Act of 1966 and the Seeds (Control) Order of 1983. It would require a QR code, generated through a central traceability portal, on every seed container, require registration of every variety offered for sale, and punish supplying spurious seed with up to three years in prison and fines of up to ₹3 million (30 lakh rupees). In September 2026 the minister said there was "no deadline yet" for it.
Achieved whenParliament passes the law and an independent check of shops finds QR codes on seed packets that trace back correctly.
Claimed by
Ministry of Agriculture and Farmers Welfare, November 2025
Open
01
Test seed from online sellers
A public test of vegetable seed bought from third-party sellers on online marketplaces in one country, germinated under AOSA or ISTA rules and compared with the germination percentage and other details the label must carry.
The briefCredit · unclaimed02
Pool the states' seed tests
One public table of the official seed samples that US state seed control offices test each year, with the labeler, crop, label claims, lab results and type of violation for each, gathered from the states' annual reports, so that compliance can be compared across states, companies, crops and years.
The briefCredit · unclaimed
Open 01 · the brief
Test seed from online sellers
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A public test of vegetable seed bought from third-party sellers on online marketplaces in one country, germinated under AOSA or ISTA rules and compared with the germination percentage and other details the label must carry.
Why it is still open
Official testing covers samples that inspectors collect. In Connecticut's 2016 report, state inspectors collected 250 vegetable seed samples and 21 of them (8.4%) fell short of their label germination; USDA's federal checks also start from samples taken in stores and warehouses. In 2026 a FoodPrint reporter bought one packet of sunflower seed on Etsy that came with none of the label information US seed law requires and grew 8 of 13 seeds, but searches found no systematic test of seed from online marketplace sellers.
Built with what exists
an ISTA-accredited or AOSA-member seed lab for the official counts, or rolled paper towels in a warm room for a first pass; a spreadsheet of labels and results; only sellers that ship from inside the country, with seedlings sealed and thrown away after counting, never planted outside.
First step
Buy 10 packets of tomato seed from 10 different online sellers and run a 100-seed paper-towel germination test on each.
Done looks like
100 packets from at least 50 sellers across 10 crops, each with its germination result next to its label claim and the Federal Seed Act tolerance, and the share of packets missing a germination figure, test date or lot number.
Open 02 · the brief
Pool the states' seed tests
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
One public table of the official seed samples that US state seed control offices test each year, with the labeler, crop, label claims, lab results and type of violation for each, gathered from the states' annual reports, so that compliance can be compared across states, companies, crops and years.
Why it is still open
Each state publishes its own report in its own layout. Kentucky's 2022 Seed Inspection Report found 205 of 1,212 official samples in violation of their labels, a compliance rate of 83.13%, and lists the results by labeler; Connecticut publishes its vegetable seed results separately. A November 2026 web search found yearly reports from Indiana, Kentucky and Pennsylvania but no table that pools them.
Built with what exists
Python with pdfplumber or Camelot, libraries that pull tables out of PDF files; an AI model that reads scanned tables, with a person checking its work (see the Note "Seven more flagged"); DuckDB to hold and query the table.
First step
Download five years of Kentucky's seed inspection reports and turn the 2022 labeler table into a checked CSV file.
Done looks like
Five years of reports from at least five states loaded, every row traced to its page, 200 rows drawn at random checked by hand with at least 99% correct, and a published comparison of violation rates by crop and by labeler.