Can Dogs Smell Cancer? What New Research Reveals About Their Remarkable Noses
Can Dogs Smell Cancer? What Science Has Found So Far
Research shows that specially trained dogs can distinguish samples from people with certain cancers from comparison samples under controlled conditions. That is a remarkable finding—but it does not mean an ordinary pet can reliably diagnose cancer by sniffing its owner.
Scientists have investigated this ability for decades. In a 2004 proof-of-principle experiment, trained dogs identified urine samples from people with bladder cancer more often than expected by chance. The researchers emphasized that demonstrating this ability was an initial step, not proof that dogs were ready for clinical screening.
More recent work explores whether technology can help turn canine responses into measurable results. A research abstract published in 2026 described a system combining trained dogs, sensors and computational analysis to assess breath samples for several cancer types. It reported promising performance in an independent validation group, although it was a conference abstract reporting interim findings.
The central question is now broader than whether a dog can recognize an unusual scent. Researchers must establish whether the approach works consistently with unfamiliar samples, different patient populations and conditions that resemble cancer.
Detecting a difference in a research sample and providing a dependable medical screening test are separate achievements. A useful test must identify enough genuine cases while avoiding excessive false alarms—and demonstrate that it helps patients in practice. The studies discussed here do not establish an untrained household dog’s behavior as a diagnostic test.
What Could Cancer Smell Like to a Dog?
Scientists are investigating volatile organic compounds, often shortened to VOCs. These are chemicals that can enter the air and contribute to odor. Changes associated with cancer may alter the mixture of compounds released through breath, urine or other biological samples.
A dog may therefore be responding to a pattern of chemical signals rather than one universal “cancer smell.” The exact compounds responsible for a trained dog’s response are not necessarily known, and successfully identifying a sample does not reveal which molecules the animal detected.
One pilot study investigated this question by combining trained dogs with chemical analysis of urine and profiling of urinary microorganisms. Researchers found differences between prostate cancer samples and comparison samples, then explored whether a computer model could learn from the dogs’ responses. This illustrates how canine scent detection can help guide research into measurable biological signals.
However, disease-related odors are not automatically specific to cancer. Inflammation, infection and bleeding can also affect biological samples. Researchers must account for these competing explanations so that dogs learn a relevant disease-associated pattern rather than an unrelated difference between groups.
The scientific challenge is understanding what the dog detects—and whether that signal remains reliable outside the original experiment. That is why careful sample selection, blinded testing and independent validation matter as much as the animal’s impressive nose.
What Did the New Cancer Detection Study Investigate?
The study examined whether trained dogs combined with digital measurements could distinguish breath samples from people with cancer from control samples. Participants provided samples for testing; the dogs were not examining people directly.
The research involved six hospitals, with samples collected between October 2024 and June 2025. Its design was described as an assessor-blinded, multicenter case-control study.
Study feature | What the researchers reported |
Sample type | Exhaled breath |
Development group | 1,250 participants used to set the system’s decision thresholds |
Independent validation group | A further 1,250 participants |
Validation group composition | 250 people with biopsy-confirmed cancers and 1,000 controls |
Detection approach | Canine responses combined with sensor, camera and computational data |
The decision thresholds were fixed before validation. This helps test whether a system can perform on a separate group rather than simply fit the data used to develop it.
The publication discussed here is a conference abstract reporting interim results. It appeared in a 2026 journal supplement associated with a 2025 conference. Those dates should not be confused with the period when participants were recruited.
How Were the Dogs Trained and Tested?
Cancer scent-detection research requires dogs to learn a distinction between sample groups. The intended task is to recognize a disease-associated odor pattern in unfamiliar samples, rather than remember individual specimens.
An earlier bladder cancer experiment illustrates this process. Six dogs underwent seven months of training before blinded testing. They encountered cancer samples alongside controls that included non-cancerous conditions, helping researchers investigate whether the dogs could distinguish cancer-associated odors from changes such as bleeding or inflammation. These details describe that earlier experiment—not the newer breath study.
In the newer study, researchers recorded the dogs’ indications and behavior using an instrumented testing area. A computational model combined these measurements to produce screening classifications. The published abstract does not provide enough detail to reconstruct the complete training protocol.
Blinded testing is essential because human expectations can unintentionally influence an animal’s response. Independent samples also matter: a convincing test must show that the dog or combined system can generalize beyond the examples used during training.
Another important challenge is testing situations in which no sample is positive—or several are. Selecting one cancer sample from a lineup is a different task from determining whether an individual patient might have cancer. Researchers have recognized this distinction since the early proof-of-principle studies.
What Did the Results Show—and What Do the Numbers Really Mean?
The interim research abstract reported 93% sensitivity and 88% specificity for the combined dog-and-computer system in its independent validation group. These figures describe two different aspects of performance; they should not be compressed into the claim that “dogs detect cancer with 93% accuracy.”
Measure | What it means | How to interpret the reported result |
Sensitivity: 93% | The proportion of cancer cases correctly identified | Approximately 93 out of every 100 cancer cases would be flagged at this performance level |
Specificity: 88% | The proportion of controls correctly classified as negative | Approximately 88 out of every 100 controls would receive a negative result |
False negatives | Cancer cases the system misses | Some people with cancer would still receive a negative result |
False positives | Positive results in people without cancer | Some people without cancer would be referred for further investigation |
The “out of 100” figures explain the percentages; they are not the study’s actual participant counts. Sensitivity is not the probability that someone with a positive result has cancer. That probability also depends on how common cancer is in the population being tested.
Consider a hypothetical screening group of 10,000 people in which 1% have cancer. If the same sensitivity and specificity applied, the system would identify approximately 93 true cases but also produce 1,188 false positives. Only about 7% of positive results would represent cancer.
This calculation is an illustration, not a prediction of the system’s real-world performance. It explains why results from a study containing a relatively high proportion of cancer cases cannot automatically establish how useful a test will be in general screening.
How Did Artificial Intelligence Help Interpret the Dogs’ Responses?
The reported results came from a combined detection system, not from dogs working alone. Researchers collected several kinds of information as the animals investigated breath samples:
Whether a dog indicated a sample as positive.
How long it spent investigating different locations.
Behavioral information interpreted from camera recordings.
Each dog’s previous performance.
Additional sample-related information.
A Bayesian model combined these inputs to generate a screening classification and an estimate of uncertainty. The study’s decision thresholds were established during development and fixed before testing the independent validation group.
The computer therefore helped interpret measured responses. It did not directly identify a cancer molecule simply by watching a dog, and the reported overall performance does not reveal how much each individual input contributed.
Earlier pilot research explored a related approach: combining dogs’ responses with laboratory analysis of chemicals in urine and an artificial neural network. That work investigated whether canine detection could help researchers identify patterns useful for future machine-based tests.
The next question is whether these combined methods remain reliable when tested independently in the people who would actually use them. Adding artificial intelligence does not, by itself, resolve the challenges of false positives, missed cancers or differences between study populations.
Which Types of Cancer Have Researchers Studied?
Scientists have investigated canine scent detection in several human cancers, using different biological samples and testing methods. Evidence from one cancer type cannot automatically be applied to another, and success in a particular experiment does not establish a universal cancer-detection ability.
Cancer type | Examples of samples studied | What the research explored |
Bladder cancer | Urine | Whether trained dogs could distinguish cancer-associated odors from control samples |
Prostate cancer | Urine | Whether canine responses could identify cancer-associated patterns, including alongside chemical analysis |
Lung cancer | Exhaled breath | Whether trained dogs could distinguish samples from patients and controls |
Breast cancer | Exhaled breath | Whether cancer-associated breath patterns were detectable by scent |
Colorectal cancer | Exhaled breath and watery stool samples | Whether a trained dog could identify cancer samples in controlled lineups |
These examples come from separate studies, not a single experiment testing every cancer under identical conditions.
The newer breath-based study discussed earlier investigated detection across multiple cancer types. However, its reported overall performance should not be interpreted as proof of equally strong performance for each individual cancer. That requires detailed results and adequate participant numbers within each group.
There is also a distinction between flagging a possible cancer-associated signal and identifying where a cancer is located. A system that separates cancer samples from controls does not necessarily determine the affected organ. Its capabilities must be established through the specific tasks researchers actually tested.
Can Dogs Detect Cancer at an Early Stage?
Some studies have reported detection of samples from people with early-stage cancer. This is one reason the research attracts attention: a useful early-detection method could potentially identify disease while treatment options are more favorable.
A 2006 pilot study involving lung and breast cancers reported similar detection performance across disease stages. A separate colorectal cancer study published in 2011 also reported strong performance for early cancers. However, that colorectal study used one specially trained Labrador, making it important to establish whether the findings generalize to other dogs and testing settings.
The newer multicenter abstract reported comparable sensitivity in stage I–II disease. Its summary does not provide enough stage-specific detail to establish precisely how reliably the system detects every early-stage cancer category.
“Early stage” does not necessarily mean “before symptoms” or “before conventional tests could detect it.” Recognizing a sample from someone with an already confirmed early-stage cancer is different from discovering an unknown cancer during routine screening.
To demonstrate screening value, researchers must test the approach in its intended population, investigate positive results and follow participants over time. Detecting a signal is only part of the challenge; the eventual goal is to improve health outcomes while limiting unnecessary investigations. Established cancer screening methods are evaluated for benefits such as reducing the chance of dying from the disease.
Why Promising Research Does Not Yet Mean a Reliable Screening Test
A dog correctly identifying cancer samples in a study is an exciting finding. However, detecting a difference between samples is only one step toward developing a useful screening test.
A screening programme must work in people whose cancer status is unknown. It must also provide a clear route for investigating positive results and demonstrate that its benefits outweigh the harms. Detecting more cancers does not automatically mean that screening helps people live longer.
For canine cancer detection, several questions need convincing answers:
Can the results be reproduced? Performance should remain consistent with different dogs, handlers, laboratories and patient populations.
What is the dog responding to? Researchers must investigate whether medications, other illnesses or sample handling could explain some responses.
Does the method work with unfamiliar samples? Testing must establish that the system recognises a relevant disease pattern beyond its training samples.
What happens after a positive result? Doctors need an appropriate diagnostic pathway to confirm whether cancer is present and identify its location.
These concerns are practical research questions. For example, one lung cancer study investigated potential confounding factors and identified two medications that warranted consideration when interpreting detection results.
False results also matter. A false positive can lead to anxiety and unnecessary investigations, while a false negative can create reassurance that delays medical attention. Even an accurate screening signal requires medical follow-up; it does not establish a diagnosis by itself.
The research supports continued investigation, but it does not justify replacing recommended screening with a dog’s response.
Can an Untrained Pet Dog Detect Cancer in Its Owner?
There are reports of pet dogs repeatedly sniffing, nudging or showing unusual interest in an area where their owner was later diagnosed with cancer. Such observations helped inspire scientific research into canine scent detection. However, an individual story cannot establish how reliably an untrained dog detects cancer.
The missing information is important. We rarely know how often dogs behave similarly when their owners do not have cancer, or how often cancer develops without any noticeable reaction from a pet. Without those comparisons, there is no dependable way to interpret a particular behaviour as a cancer warning.
Research dogs undergo structured training and testing. Investigators define what counts as an indication and compare responses with confirmed medical findings. An ordinary pet’s sniffing, licking or pawing has no equivalent validated interpretation. Evidence from trained detection dogs therefore cannot simply be applied to a family dog at home.
For owners, two distinctions are essential:
Your dog’s interest in a body area does not establish that cancer is present.
Your dog’s lack of interest does not establish that cancer is absent.
If unusual attention from your dog leads you to notice a physical change, describe that change to a healthcare professional. Decisions about examination, testing and screening should be based on medical assessment, symptoms and individual risk—not on whether your dog appears concerned.
You can appreciate your dog’s remarkable sense of smell without treating its behaviour as a home cancer test.
What If Your Dog Keeps Sniffing or Licking the Same Spot?
When a dog repeatedly focuses on one part of your body, it can feel unsettling—especially after reading stories about pets detecting cancer. However, this behaviour alone is not evidence of cancer, and there is no validated way to interpret its frequency or intensity as a measure of disease.
The sensible response is to pay attention to your own health without trying to turn your dog’s behaviour into a diagnostic test. If its interest draws your attention to a new lump, a changing mole, unexplained bleeding or a sore that does not heal, arrange a medical assessment. These changes often have explanations other than cancer, but they deserve evaluation regardless of whether a dog notices them.
You can help your doctor by explaining when you first noticed the physical change and whether it has grown, changed colour, become painful or started bleeding. Describe the symptom itself rather than concluding that your dog has detected a tumour.
Avoid repeatedly encouraging your dog to sniff the area to see whether it reacts again. A repeated response would still not confirm cancer, and a loss of interest would not rule it out. Likewise, do not wait for a dog’s reaction before seeking advice about a symptom that concerns you.
If there is no visible change or symptom, the behaviour alone does not establish a need for cancer testing. Continue appropriate screening based on your age, medical history and your healthcare professional’s recommendations. Your dog’s attention may prompt you to notice something, but medical assessment determines what that finding means.
Could Dogs Help Scientists Develop New Cancer Detection Tools?
One promising direction is to use canine scent detection to help researchers understand disease-related odour patterns. The aim is to investigate whether chemical information that dogs respond to could eventually support a reproducible laboratory or sensor-based test.
A pilot study published in 2021 explored this approach using urine samples from people with aggressive prostate cancer and biopsy-negative controls. Researchers combined trained dogs’ responses with chemical analysis, microbial profiling and an artificial neural network. The computer model used the dogs’ classifications to help identify potentially informative patterns in the chemical measurements.
The study demonstrated the feasibility of bringing these methods together. However, its small sample pool limited the conclusions, and the findings required validation in larger studies. It did not establish a ready-to-use cancer screening device.
There is also an important distinction between using AI to interpret a dog’s behaviour and developing a machine that analyses odours directly. In the first approach, the dog remains part of the detection system. In the second, researchers aim to measure chemical patterns with instruments and classify them computationally.
For either approach, the essential challenge is demonstrating reliable performance beyond the original experiment. A useful system must cope with unfamiliar samples and the biological variation found in real patients.
Dogs may therefore contribute to cancer research in more than one way: as trained participants in detection studies and as a source of clues for developing new technologies. Their ability is scientifically intriguing, while the medical value of any resulting test must be established through careful clinical validation.
How Are Detection Dogs’ Welfare and Training Needs Protected?
Cancer detection research should protect the dogs’ wellbeing as carefully as it protects the people contributing samples. A dog’s ability to participate does not mean it should work continuously or be pressured to respond. Training and testing should allow for rest, individual preferences and changes in motivation.
Reward-based training gives dogs a positive reason to participate. A correct response may earn food, a toy or another reward the individual dog enjoys. Mistakes should guide adjustments to training rather than lead to punishment. Dogs should also be able to disengage when they are uncomfortable or no longer interested.
Work sessions should be balanced with breaks, exercise, social contact and opportunities for normal canine behaviour. There is no single working schedule that suits every dog. Researchers and handlers need to consider the animal’s health, age, stamina and behaviour when deciding whether to continue a session.
Good welfare also supports sound research. A dog that becomes tired, distracted or frustrated may respond differently, making results harder to interpret. Handlers should therefore monitor engagement and record relevant changes rather than treating every response as equally reliable.
Biological samples require appropriate handling procedures that protect both animals and people. Study protocols should address safe sample presentation, hygiene and veterinary care, alongside the scientific methods used to measure detection performance.
Importantly, readers should not assume that every study follows identical welfare procedures. Transparent reporting of training methods, housing, rest periods and animal oversight helps others assess how responsibly the research was conducted. Detection dogs are living research partners whose needs remain important regardless of how promising the results appear.
What Should Pet Owners Take Away from This Research?
The most useful conclusion is that trained dogs can distinguish certain cancer-associated samples under some experimental conditions. This makes canine scent detection a valuable subject for further research. It does not mean that every dog can recognise every cancer or that an owner can interpret ordinary sniffing as a medical result.
The research is especially interesting because it connects animal behaviour with chemistry and technology. Understanding what trained dogs respond to could help scientists investigate new ways of detecting disease. However, encouraging results must be followed by independent validation and evidence that any proposed test benefits patients.
For pet owners, the practical message is straightforward: appreciate your dog’s remarkable sense of smell while keeping healthcare decisions grounded in medical assessment. A dog’s unusual interest should neither cause an immediate assumption of cancer nor provide reassurance when that interest disappears.
Continue recommended screening and seek medical advice about unexplained physical changes. If your dog draws your attention to something you had not noticed, the finding itself is what needs assessment. The dog’s behaviour cannot determine its cause.
It is also worth approaching dramatic headlines carefully. Ask whether a report describes a small experiment, a conference abstract or a fully published clinical study. Check whether dogs were tested on unfamiliar samples and whether the researchers studied people with known cancer or people undergoing screening without a diagnosis.
Dogs may help researchers discover better detection methods, but their role in research should not be confused with a dependable cancer test at home. The promise is real enough to investigate; the remaining uncertainty is important enough to explain.
Frequently Asked Questions About Dogs Smelling Cancer
Can Dogs Smell Cancer in Humans?
Studies suggest that specially trained dogs can distinguish samples from people with certain cancers from comparison samples under controlled conditions. However, performance varies between studies. This does not establish that any dog can reliably detect cancer in everyday life.
What Does Cancer Smell Like to a Dog?
Researchers are investigating patterns of volatile organic compounds associated with cancer and changes in the body. These substances can occur in samples such as breath and urine. There is no single established “cancer smell” that applies to every tumour, and the exact signals dogs respond to are not fully understood.
How Accurate Are Dogs at Detecting Cancer?
There is no single accuracy figure that applies to all detection dogs. Results depend on the cancer studied, sample type, training methods, comparison groups and testing procedures. Sensitivity describes how often a method identifies cancer-positive samples, while specificity describes how often it correctly identifies cancer-negative samples. Neither figure alone tells an individual their probability of having cancer.
Can Dogs Detect Cancer Before Symptoms Appear?
Some studies have reported detection of samples from people with early-stage cancer. However, early-stage cancer and cancer without symptoms are not the same thing. Demonstrating that dogs can support screening before symptoms appear requires studies in the intended screening population, with appropriate medical follow-up.
Does My Dog Licking a Mole Mean I Have Skin Cancer?
No. Licking or sniffing a mole is not a validated sign of cancer. Nevertheless, a new or changing mole, unexplained bleeding or a sore that does not heal should be assessed by a healthcare professional. The physical change deserves attention regardless of your dog’s behaviour.
Can My Dog Reassure Me That I Do Not Have Cancer?
No. A dog showing no unusual interest in you cannot rule out cancer. Continue recommended screening and seek medical advice about concerning symptoms, even if your pet behaves normally.
Can I Train My Dog to Check Me for Cancer?
Home training should not be used to make medical decisions. Research involves carefully characterised samples, structured training, blinded testing and comparison with confirmed medical findings. Teaching a dog to react to an odour at home does not establish that it can identify cancer or distinguish it from other conditions.
Could AI and Dogs Work Together to Detect Cancer?
Researchers are exploring this possibility. Some systems analyse dogs’ responses alongside behavioural measurements, while other projects use canine findings to help investigate chemical patterns in samples. These approaches remain distinct: interpreting a dog’s behaviour is different from building an instrument that detects disease-related odours directly.
Will Cancer Detection Dogs Replace Conventional Tests?
The studies discussed here do not establish that dogs can replace recommended screening or diagnostic procedures. Even a promising detection signal would need medical investigation. Dogs may contribute to research and the development of new tools, but identifying a suspicious sample does not by itself confirm a cancer diagnosis.
References
Source | Full URL |
Cancer Research — Abstract P93: Multi-Cancer Screening Using AI-Enhanced Canine Olfactory Detection: Interim Results from a Multi-Center Breath Analysis Study in India. Conference abstract, 2026. | |
BMJ — Olfactory Detection of Human Bladder Cancer by Dogs: Proof of Principle Study. Original research, 2004. | |
Integrative Cancer Therapies — Diagnostic Accuracy of Canine Scent Detection in Early- and Late-Stage Lung and Breast Cancers. Original research, 2006; indexed in PubMed. | |
European Respiratory Journal — Canine Scent Detection in the Diagnosis of Lung Cancer: Revisiting a Puzzling Phenomenon. Original research, 2011; indexed in PubMed. | |
PLOS ONE — Feasibility of Integrating Canine Olfaction with Chemical and Microbial Profiling of Urine to Detect Lethal Prostate Cancer. Original research, 2021. | |
National Cancer Institute — Cancer Screening Overview. | |
National Cancer Institute — Symptoms of Cancer. | |
Centers for Disease Control and Prevention — Symptoms of Skin Cancer. | |
Cancer Council Australia — Can Animals Sniff Out Cancer? |
Research note: The 2026 study discussed in this article was reported as an interim conference abstract. Its findings should be interpreted within that reporting format and should not be treated as proof that the method is ready for routine cancer screening.
Comments