ORDER
Frances Czarnecki, a former waitress, applied for Disability Insurance Benefits and Supplemental Security Insurance principally because of back pain and anxiety. An administrative law judge concluded that she was not disabled, reasoning that she was not credible and that her treating physicians were not entitled to deference. But the credibility assessment is flawed, and the ALJ improperly discounted the treating doctors’ opinions. Accordingly, we reverse the district court’s denial of relief and remand with instructions to return this matter to the Acting Commissioner.
I. Background
Czarnecki applied for benefits in September 2009, when she was 48, and alleged an onset date in August 2007. For nearly 20 years previously, she had worked primarily as a waitress, but occasionally she cleaned houses and trained dogs. Most recently Czarnecki had waitressed for a few months in 2008, but she was fired, she says because back pain prevented her from carrying trays weighing 25 to 50 pounds. Czarnecki had first experienced back pain in 1989 after she was pushed down a stairway by her son’s father. Her general practitioner, Dr. John Zielinski, began treating her in 1998. Then in 2000 her son, who was 10, pushed her against a doorknob, hurting her back further.
In May 2007 Czarnecki was hospitalized for three days after drinking heavily and threatening family members. She told doctors that for 20 years she had consumed up to 5 vodka drinks and smoked 2 packs of cigarettes daily. She was diagnosed with alcohol abuse, alcoholic hepatitis (liver inflammation caused by drinking alcohol), alcoholic ketoacidosis (when a body produces high levels of ketones because it cannot produce enough insulin), and thrombocytopenia (abnormally low blood platelet count) caused by alcohol in*638toxication.1 She was prescribed Xanax for anxiety.
The next month Czarnecki saw psychiatrist Brendan Beresford for an initial assessment on the referral of her counselor. He diagnosed her with anxiety and assessed her current Global Assessment of Functioning (“GAF”) at 50 and her highest GAF at 75.2 Dr. Beresford began regular treatment for anxiety, restlessness, and nervousness.
Czarnecki was hospitalized again in January 2008 and treated for alcohol withdrawal and seizures. That March, Dr. Beresford noted that she was depressed. Also in March, Dr. Zielinski noted that Czarnecki was experiencing pain in her lower back and right hand. Then in November, Czarnecki saw Dr. Zielinski after falling from a ladder and hurting her back and neck. A few days later she suffered a seizure and was diagnosed at the emergency room with intermittent aphasia (impaired communication because of a brain injury).3
When Dr. Beresford next treated Czar-necki in December 2008, he noted that she had stopped taking her Xanax and pain medication. In his January 2009 notes, he characterized her mood as anxious yet “ok,” and her sleep as “poor” though helped somewhat by Xanax. In February 2009 Czarnecki had another seizure that led to another ER visit. Dr. Beresford saw her again afterward and observed that she was “doing well.” But in August he noted that she was stressed, sleeping poorly, and not working, though by September her anxiety had diminished.
Czarnecki saw Dr. Zielinski for low back pain almost monthly from February through August 2009. An MRI of her spine disclosed degenerative levorotosco-liosis of the thoracolumbar spine (abnormal leftward curvature of the upper and lower spine), degenerative disc disease, and facet arthropathy (joint disease), with associated central spinal canal stenosis and neural foraminal narrowing (shrinking of nerve-root passageways in the spine).4
In September 2009 Czarnecki went to a hospital complaining of pain in her lower back and leg. She reported having fallen from a ladder twice in the last year. A physician attributed her pain to “very mild” lumbar spinal stenosis and recommended a series of epidural injections. Also that month Dr. Zielinski referred Czarnecki to a pain clinic for her back pain. Meanwhile, Czarnecki continued seeing Dr. Beresford through the middle of 2010 and reported having remained sober for a few months, though she contin*639ued complaining of severe back pain, for which she again was taking medication.
In February 2010 Czarnecki began treatment with Dr. Win Myint and reported chronic back pain, arthritis in her right hand, right club foot, and poor sleep. Dr. Myint diagnosed degenerative disc disease, pain, sleep disturbance, and depression.
Also in February 2010, a state-agency psychologist, Dr. Nathan Wagner, examined Czarnecki. She admitted her history of alcohol abuse but stated that she’d been sober for the “past years,” except for one recent relapse. Dr. Wagner observed Czarnecki’s “dysphoric affect” and gave his diagnostic impression that she had “major depressive disorder, recurrent, moderate” and “panic disorder without agoraphobia.” Dr. Wagner noted that “occupational problems” and “chronic pain” contributed to her disorders, and he assessed her current GAF at 40.5
The same day, a state-agency physician, Dr. Norma Villanueva, examined Czar-necki, who self-assessed her back pain as 10 on a 10-point scale. Dr. Villanueva noted that Czarnecki • had a “slow gait,” could not squat or walk with a tandem gait because of. pain, had “mild difficulty” climbing onto and down from the exam table, and had tenderness in her lower back. Dr. Villanueva concluded that she suffered from arthritis in her lumbar spine, grand mal seizures, panic attacks, and herniated discs.
Two other state-agency doctors, psychologist Donna Hudspeth and physician George Andrews, later reviewed Czar-necki’s medical records — but did not meet with her — and opined that she still had the physical and mental residual functional capacity (“RFC”) to work, albeit with restrictions. Dr. Hudspeth agreed with Dr. Wagner’s diagnoses but, without mentioning Dr. Wagner’s GAF measurement of 40, concluded that Czarnecki could interact and communicate adequately with a supervisor and coworkers and that she demonstrated “sufficient cognition, memory and thought processing skills to perform at least two to three simple repetitive work tasks, within physical limitations.” Dr. Andrews declared Czarnecki able to lift 10 pounds frequently and 20 pounds occasionally; stand, walk, or sit for up to 6 hours per day; and occasionally stoop, crouch, and climb ramps and stairs. Based on these assessments, a disability examiner concluded that Czarnecki at least could perform “a wide range of light work.” Accordingly, in March 2010, her claim for benefits was denied, and she sought reconsideration. Two additional state-agency doctors agreed with Dr. Hudspeth’s and Dr. Andrews’s RFC assessments, and in July 2010 her claim was again denied. Czarnecki then requested a hearing before an ALJ.
Meanwhile, beginning in March 2010, Czarnecki sought treatment from Dr. Na-yeh Mirshed. Czarnecki reported having constant and severe aching pain in her lower back after a fall more than five years earlier. She also reported an anxious mood, decreased appetite, fatigue, and sadness, and she denied alcohol use. Dr. Mirshed concluded that Czarnecki suffered from lumbar spinal stenosis and depressive *640disorder and recommended that she avoid bending, lifting, and stooping.
Czarnecki saw Dr. Myint that same month; he opined that, because of her degenerative joint disease of the lumbosa-cral spine, levoscoliosis of the thoracolum-bar spine, spinal stenosis, anxiety, depression, and chronic obstructive lung disease, she could not sit, stand, or walk more than 10 ten minutes at a time; lift more than 5 pounds; or hold a job. In July 2010 Dr. Myint completed an RFC questionnaire and estimated that Czarnecki could not walk more than lk block without resting, sit for more than 20 or 30 minutes, or stand longer than 15 minutes. He recommended that she use a cane (and a year later he prescribed one). Dr. Myint further estimated that, if employed, Czarnecki would likely be absent from work more than four days per month.
In August 2010 Dr. Beresford completed a Mental Impairment Questionnaire. He listed Czarnecki’s impairments as major depressive disorder, general anxiety disorder, and alcohol dependency (in sustained remission), and he rated both her current GAF and highest GAF in the last year at 50. He opined that her impairments would last at least 12 months and cause her to miss work more than 3 times a month. Dr. Beresford rated as poor Czar-necki’s ability to stay focused for even two hours, to show up for work regularly and on time, to make simple work-related decisions, to coordinate with others, and to perform at a consistent pace without frequent, long breaks. After this assessment, other doctors at the same clinic took over her treatment for depression and anxiety.
In October 2010, at Dr. Myint’s referral, Czarnecki began going to a pain-management center. The center’s initial progress note describes Czarnecki as “disabled” and explains that she had experienced “relatively mild and tolerable” pain in her lower back until July 2010, when it became “unbearable.” Czarnecki continued receiving treatment at the center through May 2011, including an epidural in November 2010 and a medial branch block in January 2011, both of which provided significant but very brief relief. In December 2010, another MRI of Czarnecki’s spine showed mild to minimal bilateral foraminal steno-sis and a subtle disc bulge.
Czarnecki testified before the ALJ in June 2011. She said she had been unable to work since August 2007 due to severe pain in her back, feet, hands, and neck. She had tried working as a waitress and store clerk for a few months in 2008 and 2009, and for several days in 2010, but each time she stopped because of her pain. She testified that her back pain started about 15 years ago but worsened in 2008 while waitressing. Czarnecki described the pain as chronic, radiating up to her neck, and rated it as an 8 on a 10-point scale. She would cope with the pain by taking medication and steroid injections, but the injections caused her to bleed. She also wore braces on her back, hands, and feet. Czarnecki added that her depression has made her want to stay on the couch' for months at a time, though medication helps, and she regularly has had crying spells and panic attacks. Czarnecki admitted abusing alcohol in the past but said she was currently sober and attending Alcoholics Anonymous meetings frequently. She also stated she attends church on Sundays but stands and walks every 15 minutes or so because it’s too painful to remain seated through the service.
A vocational expert (“VE”) was the only other witness. The VE said that none of Czarnecki’s past work would be available to a hypothetical 50-year-old high school graduate who would be limited to routine, repetitive tasks; who could process only simple instructions and make few deci*641sions; who must avoid corfcentrated exposure to hazards or respiratory irritants; and who could crouch and stoop only occasionally and also occasionally climb stairs and ramps but never ladders, scaffolding, or ropes. Yet, the VE continued, employment still would be available as a production assembler, small-parts assembler, or electronics worker. The VE added that these jobs could be performed even by someone further restricted from interacting with the public or meeting fast-paced production requirements. The VE further testified that there were no jobs for someone who would be off-task 30 percent of the time or miss 3 days of work per month.
The ALJ applied the 5-step analysis for assessing disability, see 20 C.F.R. §§ 404.1520(a), 416.920(a), and concluded that Czarnecki was not disabled. The ALJ determined that (1) Czarnecki had not engaged in substantial gainful activity since her alleged onset in August 2007; (2) Czarnecki suffers from degenerative joint disease of the lumbar spine, spinal steno-sis, levorotoscoliosis of the thoracolumbar spine, chronic obstructive pulmonary disease, obesity, depression, anxiety, and past alcohol and drug abuse, all of which are severe; (3) these impairments did not meet the criteria for presumptive disability; (4) Czarnecki had the RFC to perform light work; and (5) jobs were available that Czarnecki could perform. The ALJ discredited Czarnecki’s account of disabling limitations because, the ALJ said, that account contradicted the medical record. That record, the ALJ reasoned, reflected only recent significant limitations and, anyway, Czarnecki’s symptoms were well-controlled with treatment and sobriety.
In assessing Czarnecki’s mental impairments at Step 3, the ALJ concluded that they did not qualify as a listed affective disorder, anxiety-related disorder, or substance-abuse disorder, see 20 C.F.R. Pt. 404, Subpt. P, App. 1, 12.04, 12.06, 12.09. The ALJ reasoned that Czarnecki had experienced only “mild restriction” in her activities of daily living, “mild difficulties” in social functioning, and “moderate difficulties” with concentration, persistence, or pace, noting that Czarnecki could “satisfactorily answer questions relating to memory, calculation, and knowledge” during her February 2010 mental consultative exam. The ALJ added that Czarnecki retained “significant mental abilities when abstaining from drugs and alcohol,” for example, reading books associated with AA. The ALJ found that Czarnecki had experienced one to two episodes of decompensation, each of extended duration. The ALJ did not address Czarnecki’s physical impairments at Step 3, except to mention that she had considered Czarnecki’s obesity “in conjunction with [her] other impairments” and found that none of the listings had been met or equaled.
In August 2012 the Appeals Council denied Czarnecki’s request for review, and in February 2014 a magistrate judge (presiding by consent) rejected Czarnecki’s challenges to the ALJ’s decision.
II. Analysis
Because the Appeals Council denied review, we evaluate the ALJ’s decision as the final word of the Commissioner, see Scrogham v. Colvin, 765 F.3d 685, 695 (7th Cir.2014), and we will uphold that decision if it is supported by substantial evidence, see Pepper v. Colvin, 712 F.3d 351, 361-62 (7th Cir.2013). An ALJ must identify the relevant evidence and build a “logical bridge” between that evidence and her ultimate determination. See Moon v. Colvin, 763 F.3d 718, 721 (7th Cir.2014). Here, the overwhelming evidence shows that the ALJ did not build such a bridge.
*642A. Doctors’ Opinions
On appeal Czarnecki principally argues that the ALJ should have given controlling weight to the opinions of her treating psychiatrist, Dr. Beresford, and treating physicians, Dr. Mirshed and Dr. Myint. “A treating physician’s opinion is entitled to controlling weight if it’s supported by medical findings and consistent with substantial evidence in the record,” Bates v. Colvin, 736 F.3d 1093, 1099 (7th Cir.2013), and even if not controlling, that opinion cannot simply be discarded by án ALJ, Scrogham, 765 F.3d at 697. Here, the ALJ’s refusal to give any weight — let alone controlling weight — to the opinions of Czarnecki’s treating doctors is not supported by substantial evidence.
Starting with Dr. Beresford, we agree with Czarnecki that the ALJ did not provide a sound explanation for rejecting the assessment of her mental impairments. The ALJ criticized the Mental Impairment Questionnaire completed by Dr. Beresford as contradicted by treatment notes, which, according to the ALJ, “document good functioning on medications, and normal memory and concentration.” But the ALJ did not identify which treatment notes purportedly support this characterization, and, in fact, the treatment notes suggest just the opposite. Dr. Beresford began treating Czarnecki in June 2007, when he rated her current GAF at 50 and her highest GAF at 75. Then, before completing the questionnaire, he treated , her twice more in 2007, twice in 2008, five times in 2009, and six times in 2010, and his treatment notes consistently document that Czarnecki was depressed, anxious, restless, and having trouble sleeping. The remaining treatment notes from Dr. Beresford’s colleagues at the clinic in October and December 2010 and in March and April 2011 show that Czarnecki continued to experience depression, anxiety, and difficulty sleeping. And though the notes from Czarnecki’s appointment in May 2011 do indicate that she was then sleeping well, managing her anxiety, and able to “multitask,” the ALJ did not mention these notes, which, despite their few positive comments, still are. generally consistent with Dr. Beresford’s pessimistic views expressed in the August 2010 questionnaire. That document, therefore, should have formed the basis of the ALJ’s determination. See Bates, 736 F.3d at 1099-1100. The ALJ did not explain why she seemingly ignored all but a few lines in Dr. Beres-ford’s treatment records, leaving us unable to engage in meaningful review. See Scrogham, 765 F.3d at 697.
The ALJ also credited, without explanation, the opinions of the two state-agency reviewing psychologists over that of Dr. Wagner, the state-agency psychologist who personally evaluated Czarnecki in 2010, diagnosed her with major depressive disorder and panic disorder, and rated her GAF at 40. Although an ALJ is not required to accept the views of an agency examining physician if there is a contrary opinion from a later reviewer or other compelling evidence, the ALJ still must have a good explanation for rejecting or discounting the examining physician’s opinion. See Beardsley v. Colvin, 758 F.3d 834, 839 (7th Cir.2014). Here, the ALJ gave “great weight” to the state-agency reviewing psychologists simply because she deemed them, without any explanation or record support, “qualified mental health professionals” who are “experts in Social Security disability review.” Thus, the ALJ gave hardly any explanation at all, let alone a valid one, for rejecting Dr. Wagner’s opinion. See Beardsley, 758 F.3d at 839.
The Commissioner is correct that the ALJ did not need to specifically discuss the GAF score from Dr. Wagner. See *643Denton v. Astrue, 596 F.3d 419, 425 (7th Cir.2010) (explaining that GAF score does not necessarily reflect doctor’s opinion of functional capacity because the score measures severity of symptoms and functional level). Indeed, the American Psychiatric Association no longer uses this metric. See Am. Psychiatric Ass’n, Diagnostic & Statistical Manual of Mental Disorders 16 (5th ed.2013). But at the time of Czarnecki’s psychological evaluations, clinicians still used GAF scores to report their judgment of an “individual’s overall level of functioning.” See Am. Psychiatric Ass’n, Diagnostic & Statistical Manual of Mental Disorders 32 (4th ed. text revision 2000). The ALJ did not mention any of the GAF scores from any physicians who examined Czarnecki, even though all of her GAF scores after the alleged onset of disability were between 40 and 50. As in another recent case, the problem here was the ALJ’s “larger general tendency to ignore or discount” favorable evidence, “which included GAF scores from multiple physicians suggesting a far lower level of functioning” than the ALJ assigned. See Yurt v. Colvin, 758 F.3d 850, 860 (7th Cir.2014).
As for Czarnecki’s physical impairments, the ALJ should not have rejected Dr. Myint’s views because he was a treating physician whose opinions about Czarnecki’s physical limitations were consistent with the other medical evidence. See Roddy v. Astrue, 705 F.3d 631, 636 (7th Cir.2013). The ALJ took issue with the fact that Dr. Myint had treated Czarnecki for only a month before providing his first assessment and criticized his examinations as having “minimal findings,” his treatment as “consisting only of prescription medications,” and his assessment as “highly eonclusory.” But Dr. Myint’s views are consistent with his treatment records, and “eonclusory” better describes the ALJ’s analysis. At Czarnecki’s first appointment, Dr. Myint diagnosed her as having degenerative disc disease, sleep disturbance, and depression, and he later opined that these impairments prevented her from sitting, standing, or walking more than ten minutes, lifting more than five pounds, or maintaining meaningful employment. The ALJ doesn’t say why she rejected the conclusion that Czarnecki could not keep a job or explain what difference it would make if. Dr. Myint had treated Czarnecki conservatively with medication, which anyway is inaccurate, since he also referred her to a pain-management center for more aggressive treatment.
The ALJ minimizes this step by declaring that the center’s treatment records indicate that Czarnecki’s back pain was “relatively mild and tolerable until July of 2010.” In fact, this comment is from a single progress note, and the ALJ quoted it selectively; the progress note goes on to say that Czarnecki’s pain had become “unbearable” in July 2010. And other notes made by center staff, such as Czarnecki repeatedly rating her back pain as between 7 and 9 on a 10-point scale, are fully consistent "with • Dr. Myint’s assessment. Moreover, the ALJ does not acknowledge that progress notes from every other physician confirm that Czarnecki had complained about severe back pain — even while taking medication — long before July 2010. See Yurt, 758 F.3d at 860 (explaining that we have “repeatedly forbidden” ALJs from cherry-picking only the medical evidence that supports their conclusion); Moore v. Colvin, 743 F.3d 1118, 1124 (7th Cir.2014) (“The ALJ simply cannot recite only the evidence that is supportive of her ultimate conclusion without acknowledging and addressing the significant contrary evidence in the record.”).
The ALJ similarly gave “little weight” to Dr. Mirshed’s September 2010 *644assessment that Czarnecki should not bend, lift, or stoop. The ALJ characterized this opinion as inconsistent with the medical record, which — the ALJ declared — doesn’t “contain any basis for such extreme restrictions,” and with Dr. Mirshed’s own treatment notes, which— according to the ALJ — document only a “reduced range of motion of the back without other significantly limiting findings.” Again, the ALJ did not identify specific treatment notes contradicting Dr. Mirshed’s assessment or provide any other sound explanation for rejecting it. See Punzio v. Astrue, 630 F.3d 704, 710 (7th Cir.2011). Instead, the ALJ gave “great weight” to the two state-agency reviewing physicians, whom she believed were “qualified specialist physicians” and “experts in Social Security disability review.” But the ALJ did not explain their qualifications or specialties, and in any event, non-examining physicians’ opinions, by themselves, are insufficient to summarily reject the examining physicians’ opinions. See Gudgel v. Barnhart, 345 F.3d 467, 470 (7th Cir.2003).
B. Credibility
Czarnecki also argues that the adverse assessment of her credibility improperly rests on the ALJ’s own spin on the medical record. See Pierce v. Colvin, 739 F.3d 1046, 1049-50 (7th Cir.2014). Again we agree; the ALJ adopted an impermissible “sound-bite” approach in evaluating the record. See Scrogham, 765 F.3d at 698. For example, the ALJ discredited Czarnecki’s allegations of disabling back pain partly because, according to the ALJ, she “was able to help remodel her house in February of 2011.” But there is no information in the record about what exactly Czarnecki was doing to “remodel” her house — she might simply have been holding up paint samples. Moreover, the note that the ALJ relied on from the pain-management center actually says that Czarnecki had experienced significant relief from a medial branch block but then, as a result of trying to help remodel her home, suffered “severe” back pain.
Similarly, the ALJ reasoned that climbing a ladder and waitressing were activities “plainly inconsistent” with Czarnecki’s alleged limited mobility. But the ALJ did not mention that Czarnecki twice fell off a ladder — demonstrating her incapacity to successfully climb ladders — and that she had to stop waitressing because of her pain and inability to meet the physical demands of the job. Beyond that, we have “recognized that even persons who are disabled sometimes cope with their impairments and continue working long after they might have been entitled to benefits.” Shauger v. Astrue, 675 F.3d 690, 697 (7th Cir.2012); see Henderson v. Barnhart, 349 F.3d 434, 435 (7th Cir.2003). Czarnecki should not be penalized for trying to work through her pain. The ALJ at least should have questioned Czarnecki about her reasons for working and climbing ladders after the alleged onset of disability. See Henderson, 349 F.3d at 435-36.
Moreover, the ALJ questioned the timing of Czarnecki filing her application right after her husband lost his job, but the ALJ did not explain why asking for benefits only when they are most needed undermines credibility. To the contrary, we have thought it “fortunate” that persons eligible for benefits sometimes wait until a real need arises before they “feed at the public trough.” See Sarchet v. Chater, 78 F.3d 305, 308 (7th Cir.1996).
Finally, the ALJ declared that “examinations and test results” showed that Czarnecki did not need to use a cane, which she used at the hearing. But one of Czarnecki’s physicians, after more than a year of treating her, had prescribed a cane, *645and the ALJ impermissibly “played doctor” by substituting her own opinion that a cane really wasn’t necessary. See Goins v. Colvin, 764 F.3d 677, 680 (7th Cir.2014).
III. Conclusion
For the foregoing reasons, we conclude that the ALJ’s decision is not supported by substantial evidence. We reverse the district court’s judgment and remand this case to the Commissioner for further proceedings.
. Alcoholic Hepatitis, Mayo Clinic, http://www. mayoclinic.org/diseases-conditions/alcoholic-hepatitis/basics/definition/con-20026160 (last visited Dec. 5, 2014); Diabetic Ketoacidosis, Mayo Clinic, http://www.mayoclinic.org/ diseases-conditions/diabetic-ketoacidosis/ basics/definition/con-20026470 (last visited Dec. 5, 2014); Stedman's Medical Dictionary 1984 (28th ed.2006).
. The GAF Scale is a 100-point metric used to rate overall psychological, social, and occupational functioning on a hypothetical continuum of mental-health illness. See Am. Psychiatric Ass’n, Diagnostic & Statistical Manual of Mental Disorders 32, 34 (4th ed. text revision 2000). A GAF score of 41 to 50 corresponds with "serious symptoms (e.g., suicidal ideation, severe obsessional rituals, frequent shoplifting)” or "any serious impairment in social, occupational, or school functioning (e.g., no friends, unable to keep a job).” Id. at 34. A GAF score of 71 to 80 indicates "transient” symptoms and "expectable reactions to psychosocial stressors,” and "no more than slight impairment in social, occupational, or school functioning.” Id.
. Stedman’s Medical Dictionary 117 (28th ed.2006).
. Stedman’s Medical Dictionary 161, 756, 1121, 1734, 1832, 1982 (28th ed.2006).
. A GAF score of 31 to 40 corresponds with "some impairment in reality testing or communication (e.g., speech is at times illogical, obscure, or irrelevant)” or "major impairment in several areas, such as work or school, family relations, judgment, thinking, or mood (e.g., depressed man avoids friends, neglects family, and is unable to work; child frequently beats up younger children, is defiant at home, and is failing at school)." Am. Psychiatric Ass’n, Diagnostic & Statistical Manual of Mental Disorders 34 (4th ed. text revision 2000).