The Two Ways Cannabis and CBD

Key Takeaways

• Independent laboratory studies have found that between 46 and 83 percent of cannabis and CBD products deviate from their stated cannabinoid content by more than 10 percent.
• In hemp-derived CBD, the error usually runs upward. Of 84 products bought online from 31 companies, 43 percent held more CBD than labeled and 26 percent held less.
• In dispensary flower, the error runs downward. A small Colorado study of 23 samples from 10 dispensaries found measured THC averaging well below the printed figure.
• THC turns up in products sold as CBD often enough to matter to anyone facing workplace drug testing, including in products labeled THC free.
• A Certificate of Analysis is the best verification tool available to a buyer, and it settles less than most buyers assume.

---

A cannabis or CBD label makes a numerical claim: this much THC, this much CBD, in this package. Over the past decade independent laboratories have checked that claim across hundreds of products sold online, through national retail chains, and in licensed dispensaries. It fails often enough that accuracy has become a more useful question for a buyer than potency.

The failures are not random. They run in opposite directions depending on where the product was bought. Hemp-derived CBD tends to hold more cannabinoid than its label states. Cannabis flower sold in licensed dispensaries tends to hold less THC than its label states. A shopper who assumes one kind of error in both aisles will be wrong in one of them, and the two problems call for different responses.

Two Markets, One Missing Regulator

The two markets sit under different rules before any product reaches a laboratory. Hemp-derived CBD entered general commerce through the 2018 Farm Bill, which removed hemp containing less than 0.3 percent delta-9 THC from the federal controlled substances list. That change opened the category without building a quality regime for it, and no federal agency approves these products, sets a potency tolerance, or requires testing before sale. Cannabis sold in licensed dispensaries is tested because states require it, though what must be measured, which laboratories may measure it, and what counts as a passing result all vary by state. Federal scheduling shifted in April 2026 for cannabis dispensed under a state medical license and remains unresolved for adult-use, and none of it changed who verifies product contents [8].

Neither category is regulated the way a prescription drug is. In both, the number on the package is a claim made by the seller, not a measurement verified by anyone with authority to stop the sale.

The Hemp CBD Market: Usually More Than Advertised

The foundational study, published in 2017, tested 84 CBD extracts bought online from 31 companies. Only 31 percent came within 10 percent of their stated CBD content. Forty-three percent held more CBD than the label claimed and 26 percent held less. THC was detected in 18 of the 84 samples, at concentrations reaching 6.43 mg/mL, in a category marketed on the premise of containing none [1]. For a buyer whose employer tests, that last figure is the most consequential number in this article.

The pattern has held up. A case series of 105 topical hemp products bought online and at national retail stores examined the 89 that listed a CBD amount. Researchers describe a product as underlabeled when it holds more than the label states and overlabeled when it holds less, which runs opposite to how most shoppers would guess. On that measure, 58 percent were underlabeled, 18 percent overlabeled, and 24 percent accurate. THC appeared in 35 percent of the full set, all below the 0.3 percent hemp threshold. Of the 37 products containing THC, four carried a THC free claim and 19 made no reference to THC at all [2]. A separate analysis of 80 hemp CBD products bought online and locally in Kentucky found 37 of them off by at least 10 percent, with 25 above the label figure and 12 below [3].

An upward error is not automatically the safer one. Taking more CBD than intended carries interaction risk for anyone on other medications, and the THC free claim printed on a package is not, on this evidence, a reliable assurance.

The Dispensary Market: Usually Less Than Advertised

Licensed dispensary products are tested, and the recorded error still tends to favor the seller.

The clearest direct comparison remains small. Researchers used HPLC to test 23 flower samples bought from 10 dispensaries along the Colorado Front Range and compared results against the potency printed on each package. Measured THC averaged 14.98 percent, roughly 23 percent below the lowest label figure and 36 percent below the highest. About 70 percent of samples came in more than 15 percent under label, and three held close to half the stated maximum, a gap the authors trace in part to how potency testing is commissioned and paid for [4]. Twenty-three samples from one metropolitan corridor cannot establish a national rate, and no larger side-by-side study has been published.

The gap is not uniform, so the printed figure cannot be fixed by subtracting a set amount. What the evidence supports is narrower: two products tested by the same laboratory can be compared against each other more confidently than two tested by different ones [5].

An older analysis points the same way in a different product category, where dose control during manufacturing is the likelier driver than any testing incentive. A 2015 analysis of 75 edible products from three dispensaries in Seattle, San Francisco, and Los Angeles found 17 percent accurately labeled for THC, 60 percent overlabeled, and 23 percent underlabeled [6]. Most of that sample shortchanged the buyer. The smaller group ran the other way, which is the more consequential error in an edible, where the dose has been swallowed before anyone can tell it was wrong.

The potency figure is not the only label element that underperforms. An analysis of commercial cannabis across six states found that the strain names and category labels attached to products did not reliably track the chemical profiles actually present, though certain labels did show a biased association with particular chemotypes [7]. Two jars sharing a name can differ in what they contain.

How to Read a Certificate of Analysis

A Certificate of Analysis is a laboratory report on a specific batch, and it is the only verification tool available to you at the point of purchase. Reading one well takes about a minute.

How easily you can get one depends on where you are standing. Hemp brands typically post reports on their websites, searchable by batch. At a dispensary the package carries a state-mandated label, sometimes a QR code that resolves to the underlying report, and sometimes nothing beyond the printed figure. Asking staff for the batch report is a reasonable request, and how a store handles it tells you something on its own.

Start with the lot or batch number. Match it against the number on the package in your hand. A COA for a different batch describes a different product, and this is the single most common way a document that looks reassuring turns out to be irrelevant.

Check the test date next. Cannabinoid content shifts in storage, so a report issued long before the product reached the shelf describes an earlier state of it.

Look at who performed the testing. An in-house result from the producer is not third-party verification. Accreditation to ISO 17025 indicates the laboratory has been assessed against an external standard, which is a floor rather than a guarantee.

Read the cannabinoid figures carefully. Total THC and total CBD are calculated figures instead of direct measurements. Raw cannabis holds mostly THCA, an inactive acid that converts to THC under heat, so the laboratory measures both and applies a conversion factor to arrive at a theoretical maximum. The figure functions like an EPA mileage rating: derived from a formula, useful for comparison, and not a promise about any particular trip. Laboratories can differ slightly in the arithmetic they use, so read the individual analyte rows and not only the headline total.

If you are buying a CBD product, check whether THC was measured at all. A report that omits THC has measured nothing, which is different from measuring zero, and the studies above found undisclosed THC to be common in exactly this category. Check the same way for terpenes, the aromatic compounds that differ between two jars of the same stated potency, which many reports leave out entirely.

Treat an absent or unmatchable COA as information. A seller who cannot produce a batch-matched report for a product is telling you something about how the product is handled.

Finally, hold the document to its actual scope. A COA describes one sample, drawn from one batch, tested by one laboratory, on one date. It does not certify the batch you are buying next month, and it does not resolve whether a different laboratory would have returned the same number.

The Bottom Line

The evidence supports a narrower claim than either optimism or alarm would suggest. Printed cannabinoid figures are frequently inaccurate, the inaccuracy runs upward in unregulated hemp CBD and downward in licensed dispensary flower, and the size of the gap varies by product type and by study. What the evidence does not yet establish is how much a single product line varies from one purchase to the next, because the published studies tested many different products once each, not the same product repeatedly over time.

A buyer who has decided to use these products can still act on the evidence, by requesting the batch-matched report, reading it against the package, and treating the printed potency as an estimate that carries a known direction of error in that market.

For the research-facing account of why analytical consistency governs what clinical trials on cannabis can establish, see Cannabis Product Standardization and Quality: Why Consistency Matters for Research.

Frequently Asked Questions

Why do two laboratories report different numbers for the same product? Because testing methods are not standardized across the industry. An analysis of Washington state seed-to-sale traceability records found systematic differences in the cannabinoid content reported by different certified laboratories, and those differences persisted after the authors controlled for plausible confounding factors [5]. In markets where price tracks THC percentage, that variation creates an incentive to send samples to laboratories that return higher figures, a practice the Colorado authors cite as a probable contributor to the gaps they measured [4].

Does that mean COAs are useless? No. A batch-matched report from an accredited third-party laboratory is substantially more information than a printed package figure, and it is the only verification available to a buyer. The limitation is on precision and on how far a single result generalizes, not on whether the document is worth reading.

Can a CBD product cause a positive drug test? It can. THC was detected in 21 percent of the online CBD extracts in the 2017 analysis and in 35 percent of the topical products in the 2022 case series, including in products carrying a THC free claim [1, 2]. Whether any given exposure produces a positive result depends on the amount, the frequency of use, and the testing threshold applied.

Is a higher THC percentage a better product? Potency and quality are separate questions, and the published work above indicates the printed percentage is an unreliable guide to the first of them. Terpene content, which many reports omit, is not captured by a THC figure at all.

Does the strain name tell me what I am getting? Not dependably. Commercial labels across six states did not consistently align with the chemical profiles measured in those products, although some labels did show a biased association with specific chemotypes [7]. The name is a marketing identifier, and the COA is the closer approximation of contents.

---

References

• [^1]: Bonn-Miller MO, Loflin MJE, Thomas BF, Marcu JP, Hyke T, Vandrey R. Labeling accuracy of cannabidiol extracts sold online. JAMA. 2017;318(17):1708-1709. https://pubmed.ncbi.nlm.nih.gov/29114823/
• [^2]: Spindle TR, Sholler DJ, Cone EJ, Murphy TP, ElSohly M, Winecker RE, Flegel RR, Bonn-Miller MO, Vandrey R. Cannabinoid content and label accuracy of hemp-derived topical products available online and at national retail stores. JAMA Netw Open. 2022;5(7):e2223019. https://pubmed.ncbi.nlm.nih.gov/35857320/
• [^3]: Johnson E, Kilgore M, Babalonis S. Label accuracy of unregulated cannabidiol (CBD) products: measured concentration vs. label claim. J Cannabis Res. 2022;4(1):28. https://pubmed.ncbi.nlm.nih.gov/35658956/
• [^4]: Schwabe AL, Johnson V, Harrelson J, McGlaughlin ME. Uncomfortably high: Testing reveals inflated THC potency on retail Cannabis labels. PLoS One. 2023;18(4):e0282396. https://pubmed.ncbi.nlm.nih.gov/37043421/
• [^5]: Jikomes N, Zoorob M. The cannabinoid content of legal cannabis in Washington State varies systematically across testing facilities and popular consumer products. Sci Rep. 2018;8(1):4519. Author correction published Sci Rep. 2020;10(1):14406. https://pubmed.ncbi.nlm.nih.gov/29540728/
• [^6]: Vandrey R, Raber JC, Raber ME, Douglass B, Miller C, Bonn-Miller MO. Cannabinoid dose and label accuracy in edible medical cannabis products. JAMA. 2015;313(24):2491-2493. Correction published JAMA. 2015;314(2):188. https://pubmed.ncbi.nlm.nih.gov/26103034/
• [^7]: Smith CJ, Vergara D, Keegan B, Jikomes N. The phytochemical diversity of commercial Cannabis in the United States. PLoS One. 2022;17(5):e0267498. https://pubmed.ncbi.nlm.nih.gov/35588111/
• [^8]: U.S. Drug Enforcement Administration. Marijuana rescheduling regulatory actions. 2026. https://www.dea.gov/marijuana-rescheduling-regulatory-actions