GAD-7 accuracy: what an anxiety screening score can and can’t tell you

Getting 11 on an anxiety check-in, with the result “Moderate anxiety symptoms” and a suggestion to talk to a professional, can make you wonder first, “Does this mean I have an anxiety disorder?” It is also common to take it again two weeks later and get 8, “Mild anxiety symptoms”, or to get a different number on another anxiety test. Then you start to wonder how far to trust the score.

The short answer: the GAD-7 is a well-studied screening tool. The Cochrane researchers who pooled the studies describe it as one of the most frequently used anxiety screening tools, and across the 35 studies they pooled on generalized anxiety disorder, a cut-off of 10 picked up 64% of people with generalized anxiety disorder confirmed by a diagnostic interview and ruled out 91% of people without it (Aktürk et al., 2025). That is a lower detection rate than the 89% of the original study: one study’s figure and the average of many studies can differ (Spitzer et al., 2006). And screening is not diagnosis. In an illustrative calculation by the researchers for 1,000 people, 12% of whom have generalized anxiety disorder, only about half of those scoring 10 or more would have it (Aktürk et al., 2025), and because the questions ask about the past two weeks, it is natural for the score to move from one check-in to the next.

This guide covers how the accuracy of a screening test is judged, how a cut-off of 10 performed when the studies were pooled, why a score of 10 or more may not mean an anxiety disorder and what a low score can miss, why scores move when you check in again, which kinds of anxiety the GAD-7 tends to miss, and how to read and use your result. Where a score on the mind.io anxiety check-in (GAD-7) sits compared with other people is covered in average GAD-7 score, and how anxiety actually shows up in anxiety symptoms.

A screening test such as the GAD-7 does not confirm an illness; it is a first step that picks out people who may need a closer look. So its accuracy is judged against a diagnostic interview. Sensitivity is the share of people found to have generalized anxiety disorder at interview who score at or above the cut-off; specificity is the share of people without it who score below. The yardstick is a structured or semi-structured diagnostic interview, such as the MINI or the SCID. A diagnosis of generalized anxiety disorder needs excessive anxiety and worry on most days for more than six months, difficulty controlling the worry, and at least three of six symptoms: restlessness, fatigue, trouble concentrating, irritability, muscle tension and sleep problems (Aktürk et al., 2025).

The GAD-7 was first validated at 15 primary care clinics in the United States. Of 2,740 patients who completed the questionnaire, 965 chosen at random from those who agreed were interviewed by telephone by a mental health professional within a week, and 73 of them had generalized anxiety disorder. Against that interview, a score of 10 or more had a sensitivity of 89% and a specificity of 82%. The seven items were the ones, out of 13 candidates, most closely related to the total score, and internal consistency, how consistently the items move together, was .92 (Spitzer et al., 2006).

The developers said scores of 5, 10 and 15 could be read as mild, moderate and severe anxiety, and put forward 10 as a reasonable cut-off for picking out generalized anxiety disorder. Raising the cut-off to 15 lifted specificity to 95% but lowered sensitivity to 48%. People found to have generalized anxiety disorder at interview averaged 14.4, and people without it 4.9 (Spitzer et al., 2006). So where a score sits tells you more than whether it crossed a line.

The mind.io anxiety check-in (GAD-7) uses the seven GAD-7 items as they are. The Korean items follow Pfizer’s official Korean version, the French ones the official French version, and the English ones the original wording. The Japanese items are a translation that follows the original meaning. The site that distributes the official forms has no Japanese version, so mind.io does not present them as an official translation. Each item is answered from 0 to 3, the 0 to 21 total is read in the same four ranges, and from 10 mind.io suggests talking with a professional. But mind.io has not studied how accurate it is as an online check-in. Every figure in this guide comes from GAD-7 research in clinics and communities.

The original study confirmed generalized anxiety disorder in 73 people. Since then, a Cochrane review pooled 48 studies with 19,228 people, found in searches up to January 2024. They covered the GAD-7 in 24 languages from 27 countries, and left out studies of people seeking mental health care or recruited because of their symptoms. In the 35 studies (15,274 people) that looked at generalized anxiety disorder, a cut-off of 10 (or 8 to 9 where 10 was not reported) had a sensitivity of 64% and a specificity of 91%. For any anxiety disorder, the 22 studies (10,245 people) gave a sensitivity of 48% and a specificity of 91% (Aktürk et al., 2025).

Individual studies, however, ranged widely, with sensitivities from 9% to 91%, and the researchers stressed that these figures should be read as rough averages that may differ substantially in specific situations. Sensitivity was higher in samples where generalized anxiety disorder was more common, while differences by type of interview or study quality were not statistically significant. In the 5 studies in non-clinical settings (3,458 people: the general population, migrant workers, students, older people and others), sensitivity was 50% and specificity 93%, but differences between settings were not statistically significant either (Aktürk et al., 2025). In an earlier meta-analysis of 12 samples (5,223 people), the 11 samples on generalized anxiety disorder gave a sensitivity of 83% and a specificity of 84% at a cut-off of 8, with similar results from 7 to 10 (Plummer et al., 2016, from the abstract). The Cochrane review included four times as many studies (Aktürk et al., 2025).

Where to set the cut-off is a trade-off. In the same review’s separate analyses by cut-off, 7 gave a sensitivity of 84% and a specificity of 78%, 8 gave 75% and 85%, 10 gave 62% and 90%, and 15 gave 25% and 97% (being analyses by cut-off, they differ slightly from the 64% and 91% above). The combined accuracy was highest at 7, but in the researchers’ calculation, which used another model’s 80% and 78% for a cut-off of 7 and assumed that 12% of people have generalized anxiety disorder, a cut-off of 7 would flag 290 of 1,000 people, only 33% of whom would have it. The researchers judged that switching to a lower “optimal” cut-off is unlikely to solve the problem of many people being misclassified (Aktürk et al., 2025).

Studies have also been done within single countries. In Korea, 1,157 adults recruited through online advertisements and hospital introductions answered the GAD-7 and had a face-to-face MINI diagnostic interview; 90 (7.7%) had generalized anxiety disorder. This study used the same Pfizer Korean version as mind.io; a cut-off of 10 had a sensitivity of 72.2% and a specificity of 89.3%, and the best balance came at 8 (sensitivity 81.1%, specificity 84.5%) (Ahn et al., 2019). A Korean study of 112 psychiatric outpatients found the scores reliable, but the GAD-7 poor at picking out anxiety disorders in that setting (Lee et al., 2022, from the abstract). A study of the Japanese version (J-GAD-7) in primary care and other settings also exists, but its sample had an unusually high share of people with generalized anxiety disorder, and the number with the disorder differs between its abstract and its tables, so it needs to be read with care (Muramatsu et al., 2025). Whether mind.io’s Japanese items use the same wording as that version has not been checked. In France, among 145 patients of a specialist epilepsy clinic in Marseille (49 with generalized anxiety disorder on the MINI), the best-balanced cut-off was 7 (Micoulaud-Franchi et al., 2016, from the abstract).

So the best cut-off shifts with who is tested and against which interview. mind.io uses 10 in all four languages, the cut-off set out by the original study and used for the Cochrane review’s main analysis. Lowering a cut-off finds more people with generalized anxiety disorder but also flags more people without it; raising it does the opposite. A score just below the line, such as 8 or 9, is worth taking seriously too. A short version using only the first two items (the GAD-2, with a cut-off of 3) was not significantly different in accuracy from the GAD-7 (Aktürk et al., 2025), but mind.io shows only the seven-item total.

A sensitivity of 64% and a specificity of 91% do not mean that 91% of people scoring 10 or more have generalized anxiety disorder. The share of people scoring 10 or more who actually have it (the positive predictive value) depends heavily on how common generalized anxiety disorder is among the people being tested. The rarer it is, the more people without it will cross 10 by comparison.

Here are some real figures. In the original study, 23% of primary care patients scored 10 or more, and among the 965 interviewed (73 with generalized anxiety disorder), 29% of those scoring 10 or more had it (Spitzer et al., 2006). In an illustrative calculation by the Cochrane researchers, if 1,000 people, 12% of whom have generalized anxiety disorder, took the test, 156 would score 10 or more, of whom 77 (49%) would have it and 79 would not. In a calculation for primary care patients, with a rate of 8%, the share was 42% (Aktürk et al., 2025). In the Korean study (7.7% with generalized anxiety disorder), it was 36.3% (Ahn et al., 2019). How Korean and German adults score on the GAD-7 is covered in average GAD-7 score.

That does not mean a score of 10 or more without generalized anxiety disorder is nothing. In the Korean study, the 56 people with major depression but without generalized anxiety disorder averaged 10.13 on the GAD-7, people with another anxiety disorder 6.66, and people with no mental disorder 2.34 (Ahn et al., 2019). An analysis of the same 965 patients as the original study reported that the GAD-7 also worked for picking out panic disorder, social anxiety disorder and post-traumatic stress disorder (Kroenke et al., 2007, from the abstract), and in people seen for specific medical conditions, sensitivity tended to be higher and specificity lower (Aktürk et al., 2025). Depression, other kinds of anxiety and physical illness can all raise the score. So a score of 10 or more is a signal worth talking over with a professional, whether or not it turns out to be an anxiety disorder.

Errors go the other way too. Taking the sensitivity of 64% at face value, about 36 in 100 people with generalized anxiety disorder would score below 10. For any anxiety disorder, sensitivity was 48%, so more than half could score below 10 (Aktürk et al., 2025). Even with a low score, if worry and tension have lasted more than two weeks or daily life feels hard, trust that feeling more than the number and ask for help.

Because of these errors, views on screening differ. According to the 2025 Cochrane review, the US Preventive Services Task Force in 2023 recommended anxiety screening for adults aged 64 or younger, but to the researchers’ knowledge no other country had a similar recommendation, and evidence that screening improves outcomes was lacking. The researchers questioned whether screening with fixed cut-offs outside healthcare, or among all patients seeking care, is an efficient approach (Aktürk et al., 2025). A self-check done on your own is best used not as a verdict but as a way to start a conversation with someone.

Every GAD-7 item asks how often, over the last two weeks, you have been bothered by a problem. A diagnosis of generalized anxiety disorder, by contrast, rests on symptoms lasting more than six months, and the developers pointed out this difference (Spitzer et al., 2006). A fortnight packed with stressful events, such as exams or a job change, can push the score up for a while, and long-standing worry can score lower if the last two weeks were a little easier. Scores changing when circumstances change is normal.

Measured again a short time later, scores are quite close. In the original study, 236 people who answered the same questionnaire again within a week gave two scores that agreed well (intraclass correlation 0.83), and the scores people gave themselves matched scores from an interviewer asking the same items just as well (0.83) (Spitzer et al., 2006). These were measured separately: one compares the same questionnaire taken twice, the other two different ways of answering. Even so, a difference of 1 or 2 points can come from how you feel that day.

So how big a change is meaningful? A study using data from 261 people in a treatment trial for chronic depression estimated 4 points on the GAD-7 as the minimal clinically important difference; the researchers noted that it needs confirming in people treated for anxiety (Toussaint et al., 2020, from the abstract). A single difference of 1 to 3 points therefore tells you less than the direction over several check-ins.

On mind.io, the band edges can also make a score look as if it changed a lot. If one answer out of seven moves by one step, the total changes by 1 point, yet the bands split between 4 and 5, 9 and 10, and 14 and 15. So 9, “Mild anxiety symptoms”, and 10, “Moderate anxiety symptoms”, are one answer apart, and from 10 the result page also suggests talking with a professional. The result page shows no separate note when a score sits near a band edge, so look at the score itself rather than the band name. If an earlier result is saved in the same browser, the result page shows how many points it changed and notes that scores reflect the past two weeks and naturally move up and down, and you can add a calendar reminder to check in again in two weeks.

The GAD-7 was built to find generalized anxiety disorder, and its developers listed its focus on a single anxiety disorder as a limitation (Spitzer et al., 2006). When studies were pooled, its sensitivity for any anxiety disorder was 48%, lower than for generalized anxiety disorder (Aktürk et al., 2025), and the Korean researchers concluded that using the GAD-7 to screen for anxiety disorders in general (including generalized anxiety disorder) calls for caution (Ahn et al., 2019). Fears that arise only in particular situations, or panic that comes on suddenly, may show less in the score, so see the guides on panic attacks, social anxiety, phobias, OCD and health anxiety as well. mind.io has no separate test for these kinds of anxiety.

It also overlaps a good deal with depression. In the Korean study, GAD-7 scores were more strongly related to two depression scales (.80 and .83) than to a worry scale (.72) (Ahn et al., 2019). The developers found that anxiety and depression symptoms often occur together but form separate dimensions (Spitzer et al., 2006). If your mood has been low or you have lost interest in things, take the depression check-in (PHQ-9) as well. It is not a diagnosis either. How to read that score is covered in PHQ-9 accuracy.

Look at your body and your age too. Anxiety symptoms overlap with the symptoms of physical illnesses (Aktürk et al., 2025). New chest pain, breathlessness or fainting need medical assessment first; if they come on suddenly or severely, call emergency services. Only 4 of the studies in the Cochrane review had a mean participant age of 65 or over (Aktürk et al., 2025), and a German study of 438 people aged 58 to 82 living at home concluded that a lower cut-off, such as 5, may suit older people (Wild et al., 2014, from the abstract). The Cochrane review included only people aged 18 or over, so a teenager is better off reading the result with a parent, a school counsellor or a professional.

The GAD-7 has no question about self-harm: whatever your score, if you have thoughts of harming yourself, do not carry them alone. Call or text 988 now in the US or Canada, call Samaritans on 116 123 in the UK or Ireland, or find a line at findahelpline.com elsewhere.

First, look at the score and how far it is from the line, not just the band name. A 9 and a 10 have different names and messages, but they are one answer apart. In the Korean study the best balance came at 8 (Ahn et al., 2019), so at 8 or 9, rather than deciding you are fine, look at which items were high; at 11, read it not as a diagnosis but as a signal to talk to someone.

Second, check in again about two weeks apart, in similar circumstances. The GAD-7 asks about the past two weeks, so checking more often measures overlapping periods. Look at the direction over several check-ins, and at whether the score moved by 4 points or more, rather than at a single difference of 1 to 3 points (Toussaint et al., 2020, from the abstract).

Third, note what the score leaves out. The GAD-7 asks only how often symptoms came up over the past two weeks, so how many months the worry has lasted, what you have started to avoid, and how much it gets in the way of work and relationships are not part of the score. A diagnostic interview looks at these as well.

Fourth, if your total is 10 or more, or whatever your score, if worry and tension have lasted more than two weeks, or anxiety has led you to avoid more things or got in the way of daily life, talk to a professional. Outside Korea, your family doctor or a local mental health service is a good place to start; in Korea, a community mental health welfare centre or a psychiatry clinic. Saying “I took this check-in and this is my score” is an easy way to begin. What happens at a first visit is in seeing a psychiatrist for the first time.

What not to use it for: do not use a GAD-7 score alone to decide that you or someone else has an anxiety disorder, or to dismiss a hard time because the score was low. Being a worrier is not the same as having an anxiety disorder, and a high score may come from depression or a physical problem rather than anxiety. A diagnosis is made by a professional who looks at the whole picture in an interview.

Q. Is the GAD-7 a reliable test? A. As a tool for screening anxiety symptoms, it has been studied a great deal. In a Cochrane review of 48 studies with 19,228 people, the 35 studies on generalized anxiety disorder found that a cut-off of 10 picked up 64% of people with generalized anxiety disorder and ruled out 91% of people without it (Aktürk et al., 2025). But it is a screening tool, so a score alone is not a diagnosis. The mind.io check-in uses the same seven items and scoring.

Q. Does a score of 10 or more mean I have an anxiety disorder? A. Not on its own. In the researchers’ illustrative calculation assuming that 12% of people have generalized anxiety disorder, only about half (49%) of those scoring 10 or more would have it (Aktürk et al., 2025), and in the Korean study the share was 36.3% (Ahn et al., 2019). Still, 10 or more is a signal worth talking over with a professional, because even without generalized anxiety disorder there may be other difficulties, such as depression or another kind of anxiety.

Q. Does a score of 9 or less mean I am fine? A. Not necessarily. Taking the sensitivity of 64% at face value, about 36 in 100 people with generalized anxiety disorder score below 10 (Aktürk et al., 2025). If worry and tension have lasted more than two weeks or daily life feels hard, you can ask for help whatever your score.

Q. The original study said 89%. Why are the figures different? A. The 89% and 82% come from one study of 965 US primary care patients, 73 of them with generalized anxiety disorder (Spitzer et al., 2006), while the 64% and 91% are an average over 35 studies from many countries (Aktürk et al., 2025). The figures depend on who is tested and against which interview, and they varied widely between studies.

Q. I have heard you should start paying attention at 8. Is that right? A. In the Korean study, 8 gave the best balance (Ahn et al., 2019), and across the pooled studies, 7 did. But lowering the cut-off lowers the share of positive results that are true cases (Aktürk et al., 2025). mind.io uses 10 as its line, but a score of 8 or 9 is still worth a closer look.

Q. Can the GAD-7 tell me about panic or social anxiety? A. It depends on the study. A US primary care study reported that it worked for these too (Kroenke et al., 2007, from the abstract), but across the pooled studies its sensitivity for any anxiety disorder was 48% (Aktürk et al., 2025), and the Korean researchers also advised caution in using it to screen for anxiety disorders in general (including generalized anxiety disorder) (Ahn et al., 2019). The guides on panic attacks and social anxiety may help more.

Q. My score is different from last time. Which one is right? A. Both reflect the two weeks before each check-in. Scores from people who answered again within a week agreed well (Spitzer et al., 2006), but a difference of 1 to 3 points can come from how you feel that day, and a meaningful change has been estimated at about 4 points (Toussaint et al., 2020, from the abstract). Look at the direction over several check-ins rather than at one difference.

Q. How often can I take it again? A. The GAD-7 asks about the past two weeks, so about two weeks apart works well. You can add a calendar reminder for two weeks later from the mind.io result page. At 15 or more, the result page suggests checking again alongside talking to a professional rather than keeping an eye on things alone.

Q. It’s so hard I’m having thoughts of harming myself. A. Please do not carry them alone. Reach out to a crisis line right away: call or text 988 in the US and Canada, call Samaritans on 116 123 in the UK and Ireland, or find a line at findahelpline.com elsewhere. If you are in immediate danger, contact your local emergency services.

References: Ahn, J.-K., Kim, Y., & Choi, K.-H. (2019), The psychometric properties and clinical utility of the Korean version of GAD-7 and GAD-2, Frontiers in Psychiatry 10, 127, https://doi.org/10.3389/fpsyt.2019.00127; Aktürk, Z., Hapfelmeier, A., Fomenko, A., Dümmler, D., Eck, S., Olm, M., Gehrmann, J., von Schrottenberg, V., Rehder, R., Dawson, S., et al. (2025), Generalized Anxiety Disorder 7-item (GAD-7) and 2-item (GAD-2) scales for detecting anxiety disorders in adults, Cochrane Database of Systematic Reviews 2025(3), CD015455, https://doi.org/10.1002/14651858.CD015455; Kroenke, K., Spitzer, R. L., Williams, J. B. W., Monahan, P. O., & Löwe, B. (2007), Anxiety disorders in primary care: Prevalence, impairment, comorbidity, and detection, Annals of Internal Medicine 146(5), 317-325, https://doi.org/10.7326/0003-4819-146-5-200703060-00004; Lee, S.-H., Shin, C., Kim, H., Jeon, S. W., Yoon, H.-K., Ko, Y.-H., Pae, C.-U., & Han, C. (2022), Validation of the Korean version of the Generalized Anxiety Disorder 7 self-rating scale, Asia-Pacific Psychiatry 14(1), e12421, https://doi.org/10.1111/appy.12421; Micoulaud-Franchi, J.-A., Lagarde, S., Barkate, G., Dufournet, B., Besancon, C., Trébuchon-Da Fonseca, A., Gavaret, M., Bartolomei, F., Bonini, F., & McGonigal, A. (2016), Rapid detection of generalized anxiety disorder and major depression in epilepsy: Validation of the GAD-7 as a complementary tool to the NDDI-E in a French sample, Epilepsy & Behavior 57, 211-216, https://doi.org/10.1016/j.yebeh.2016.02.015; Muramatsu, K., Maeda, S., Miyaoka, H., Kamijima, K., Muramatsu, Y., Tanaka, Y., Hosaka, M., Miwa, Y., Fuse, K., Yoshimine, F., et al. (2025), Diagnostic performance of the Japanese version of the GAD-7 in primary care and comparing its accuracy with other language versions, Medical Research Archives 13(1), https://doi.org/10.18103/mra.v13i1.6247; Plummer, F., Manea, L., Trepel, D., & McMillan, D. (2016), Screening for anxiety disorders with the GAD-7 and GAD-2: A systematic review and diagnostic metaanalysis, General Hospital Psychiatry 39, 24-31, https://doi.org/10.1016/j.genhosppsych.2015.11.005; Spitzer, R. L., Kroenke, K., Williams, J. B. W., & Löwe, B. (2006), A brief measure for assessing generalized anxiety disorder: The GAD-7, Archives of Internal Medicine 166(10), 1092-1097, https://doi.org/10.1001/archinte.166.10.1092; Toussaint, A., Hüsing, P., Gumz, A., Wingenfeld, K., Härter, M., Schramm, E., & Löwe, B. (2020), Sensitivity to change and minimal clinically important difference of the 7-item Generalized Anxiety Disorder Questionnaire (GAD-7), Journal of Affective Disorders 265, 395-401, https://doi.org/10.1016/j.jad.2020.01.032; Wild, B., Eckl, A., Herzog, W., Niehoff, D., Lechner, S., Maatouk, I., Schellberg, D., Brenner, H., Müller, H., & Löwe, B. (2014), Assessing generalized anxiety disorder in elderly people using the GAD-7 and GAD-2 scales: Results of a validation study, The American Journal of Geriatric Psychiatry 22(10), 1029-1038, https://doi.org/10.1016/j.jagp.2013.01.076