ORDER
Marla Phillips applied for disability insurance benefits and supplemental security income, claiming that she was disabled by major depression, borderline personality disorder, and other physical impairments. The Social Security Administration denied her claim, and a magistrate judge, presiding by consent, upheld the agency’s decision. Phillips appeals, arguing primarily that the Administrative Law Judge erred by discounting the opinions of her treating doctors, improperly weighing the evidence of her mental illnesses, and discrediting her testimony. Because we conclude that the ALJ’s decision is not supported by *879substantial evidence, we reverse and remand for further proceedings.
Background
Phillips is 30 years old. At the age of four, her father allegedly molested her and tried to drown her in the bathtub, and since then she has been treated for depression and other mental impairments. She spent five years in counseling as a child, and at the age of ten, she made her first of at least six attempts to commit suicide with drugs or by cutting her wrists. At one point she was hospitalized for six weeks for psychiatric treatment.
Phillips worked beginning at age 15 as a housekeeper, clerk, and certified nurse’s assistant. But by December 2004 her physical and psychiatric impairments had left her unable to work, and she applied for disability benefits. Three times in the 25 months immediately preceding that application, Phillips had been hospitalized for suicidal ideation or attempted suicide.
The first of these incidents was in November 2002 when, at age 22, Phillips was admitted to Harrisburg Medical Center in southern Illinois, complaining of severe depression and suicidal ideation. At the time her Global Assessment of Functioning (“GAF”) score was 20 to 25, indicating serious impairments in her ability to function and a danger that she might hurt herself or others. The severity of her symptoms led the examining psychiatrist to conclude that Phillips, who until this episode had been enrolled in a college nursing program and working as a nursing assistant, would “require an extensive treatment plan” and was “a good candidate for a partial hospitalization program.” The psychiatrist noted that Phillips had been dropped from the nursing program because her depression had led to poor attendance. She also had been fired from her job after a misdemeanor arrest for receiving $1900 in unauthorized food stamps, which she maintained were for her son, who was then two years old and living with her mother. Phillips remained at the hospital for twelve days until doctors found a workable combination and dosage of antipsychotic drugs.
Just ten days after her discharge, however, Phillips was readmitted to Harrisburg following two successive suicide attempts — first overdosing on her antidepressants and then, when that did not work, slashing her wrists. Phillips cried throughout the interview as she told the same psychiatrist who had treated her in November that she had “no desire to do anything,” that she didn’t have any will to live, and that she “just can’t get rid of it.” The psychiatrist observed that Phillips had embraced the view that her situation was hopeless and that eventually she would succeed in killing herself. He recommended that planning be done “to facilitate the patient’s entering a group home or similar kind of arrangement” because her “high degree of suicidality and lethal potential makes her a very difficult case to manage.” Her depression gradually subsided after more adjustments to her medications, and by the time she was discharged after eight days, her GAF score had increased from 20 to 50, still showing that she suffered from serious symptoms.
Then in July 2004, after relocating to Kansas, Phillips was admitted to Mount Carmel Regional Medical Center for severe depression and suicidal ideation. Doctors there noted that Phillips presented with hypersomnolence, refusing to get out of bed for her intake evaluation and declining to participate in any psychological testing or therapy sessions. This led one psychologist to conclude that Phillips “appear[ed] to be treating her hospital stay as more of a vacation than a treatment program.” Her medications were *880adjusted, though, and her condition improved by her discharge eight days later.
In January 2005, following her return to southern Illinois and about a month after she applied for disability benefits, Phillips again was hospitalized at Harrisburg when she attempted to take her life by cutting her wrist. Dr. Aline Gilbert-Johnson, a psychiatrist, noted that Phillips experiences bouts of serious depression a few times a year, each time lasting three weeks to two months. He adjusted her medications, adding Wellbutrin (because “that’s the only antidepressant she hasn’t tried yet”), Ability (used to treat schizophrenia, bipolar disorder, and depression) and Tegretol (for bipolar disorder). Her GAF score on admission had been 30, indicating serious impairments in her ability to function, but this had increased to 70, reflecting mild symptoms, by the time she left.
After her discharge, Phillips continued outpatient treatment with Mohamed Elsamahi, a physician’s assistant at Harrisburg with a Ph.D. in Philosophy, and Dr. Rakesh Chandru, a psychiatrist at the Egyptian Health Department. On February 14, 2005, Phillips reported to Elsamahi that she had insomnia, had been crying when depressed, and was hearing voices. Elsamahi noted in his file that Phillips had “improved 50%” but her affect was still restricted. On that same day Phillips also participated in a “comprehensive mental health assessment” at the Egyptian Health Department. The assessment was completed by Amy Bates, a social worker on staff, but reviewed and signed by Dr. Chandru. Bates concluded that, although Phillips’s ability to care for herself was not impaired by her mental-health issues, she had moderate impairments in social and educational/vocational functioning and a serious impairment related to her “symptom distress.”
On March 5, 2005, Phillips was admitted to Harrisburg because of “homicidal ideation and auditory hallucinations telling her to kill people.” Phillips reported feeling angry and agitated and terrified of the voices she was hearing. She was still taking Wellbutrin, Ability, Trazodone, and Tegretol for her depression, as well as Klonopin for panic attacks, and Atarax for anxiety. Dr. Gilbert-Johnson increased her dose of Ability, and after the adjustment he noted that she “made good improvement.” Her GAF on admission had been 40, indicating a major impairment in several areas of functioning, and before her discharge three days later it had increased to 75, reflecting only transient symptoms or normal reactions to stress.
On March 18, 2005, Dr. David Warshauer, a state-agency clinical psychologist, examined Phillips. Her affect, he reported, was one of “agitated depression,” and she cried while explaining how she “feels that she has failed as a mother” and “has nothing to live for.” He commented that her GAF was 45, indicating a serious impairment in functioning, and opined that her “borderline personality disorder is quite severe.”
In May 2005, Dr. Lionel Hudspeth, a state-agency psychologist, assessed Phillips’s mental residual functional capacity based upon the medical record. Dr. Hudspeth acknowledged her diagnoses of a depressive disorder and a borderline personality disorder characterized by persistent disturbances in mood or affect, intense and unstable interpersonal relationships, and impulsive and damaging behavior. On the agency’s form assessment, however, he checked boxes indicating that these impairments were causing only mild restrictions in Phillips’s daily activities and her ability to maintain concentration, persistence and pace, and a moderate limitation in her social func*881tioning. Dr. Hudspeth also checked a box corresponding to his count of “one or two” extended episodes of decompensation caused by her mental impairments. At the end of the form he offered his two-sentence assessment: “This claimant has cognition, memory and thought processes are intact. She does have difficulty in interpersonal relationships, and would be best suited in job tasks requiring no contact with the public, and minimal contact with coworkers, and supervisors.” Dr. Hudspeth did not explain any of his findings, or discuss the extensive medical record, or even identify the portions of the medical record he deemed significant. Indeed, he left blank the page for “Consultant’s Notes” except to remark, “25yo female who has DX: of Depressive Disorder and Borderline Personality.”
In mid-June, Phillips went to the emergency room at Harrisburg, complaining of severe depression, suicidal ideation, and auditory hallucinations. She was not admitted, and there are only two pages of records from her visit, but the examining doctor noted that she was depressed, angry, and agitated. Progress notes from Elsamahi and Dr. Chandru show that Phillips’s condition continued to fluctuate throughout 2005.
On March 20, 2006, Phillips was again admitted to Harrisburg for severe depression and another attempted suicide; she had overdosed on 20 tablets of Ativan, an antianxiety medication. Dr. Naeem Qureshi, a psychiatrist, noted that Phillips had experienced episodes of major depression since she was a child and that “in between the depressive episodes [she] sometimes experiences also mild to moderate depression of between 3-4 days weekly with excessive mood swings.” Phillips told Qureshi that she had been hallucinating and hearing voices which told her to kill herself and others. When she was admitted Phillips was taking Geodon (which is used to treat schizophrenia and mania), Zoloft, Effexor XR (for depression), Ativan, and Trazodone. Qureshi adjusted her medications and observed that Phillips “improved rather fast and the depression disappeared quickly” after Ability was added to the drug regimen.
On April 4, 2006, Elsamahi, the physician’s assistant, authored a letter describing his treatment of Phillips and summarizing his findings. He noted that Phillips had been seeing him regularly since January 2005. According to Elsamahi, Phillips’s mental illnesses caused severe symptoms, which in turn meant that she could not “function normally in public” or “tolerate normal occupational stressors, or have the energy to attend work regularly.” He corroborated her diagnoses of depressive disorder, generalized anxiety disorder, panic disorder with agoraphobia, and borderline personality disorder, and opined that Phillips “cannot hold an average job for a reasonable period of time.” The name of the psychiatrist who treated Phillips at Harrisburg in 2002 also appears on the letterhead, but it is unclear whether that physician had reviewed and approved Elsamahi’s correspondence.
Around the same time, Linda Redick, a physician’s assistant for the primary-care physician Phillips had been seeing regularly for the past three years, completed a “Multiple Impairment Questionnaire.” Redick noted under “prognosis” that “mentally [Phillips] is very unstable,” and she opined that Phillips was likely to be absent more than three times a month and was incapable of tolerating even low-stress work because she “has very low selfesteem” and “gets very emotional during stressful times.” According to Redick, Phillips was not a malingerer.
Phillips continued to attend monthly sessions with Elsamahi, who in April 2007 *882completed a mental-health assessment. He concluded that her prognosis was “guarded” because she responded only partly to medications, her anxiety was severe, and her panic attacks and depressive episodes were frequent. According to Elsamahi, Phillips experienced marked limitations in five functional areas, including in her ability to maintain regular attendance, to complete a normal workweek without interruptions from psychologically based symptoms, to perform at a consistent pace without an unreasonable number of breaks, and to interact appropriately with others. Like Redick, Elsamahi concluded that Phillips was not a malingerer, that she would likely miss work more than three times a month, and that she was incapable of tolerating even low-stress work because even mild stress had triggered serious deteriorations in her mental stability.
At the hearing before the ALJ in April 2008, Phillips testified about the debilitating effect of her mental impairments. Phillips explained that her medications would help for a while, but doctors had to change her prescriptions frequently in order to manage her symptoms. When asked how her conditions affected her ability to work, Phillips responded that “being criticized or being told to do something or told that I’ve done something wrong makes me very anxious and nervous and depressed even.” Phillips explained that everything caused her anxiety, that she regularly experiences nightmares and flashbacks, and that she still thinks about committing suicide every day. She informed the ALJ that, about a month before the hearing, her son, then seven years old, and a daughter born in 2005 had been removed from her home by social workers after she had been “declared unstable and unfit in a court of law.” She testified that she had done her best to care for her children but struggled with daily activities. She spent a lot of time sleeping or watching television, she said, but even these activities were affected because she experiences “a lot of racing thoughts” that hinder her concentration. When asked about her use of illegal drugs, Phillips denied current use but acknowledged past use of marijuana. Her medical records document disclosures of periodic marijuana use to her physicians, and Phillips offered that she sometimes resorted to marijuana when her combination of prescription drugs no longer controlled her symptoms.
The ALJ and Phillips’s counsel posed hypothetical questions to a vocational expert, who testified that an employee with the same age and work history as Phillips, and with the limitations described by the ALJ, could perform the sedentary, unskilled work of a packer, an inspector, or an assembler. The vocational expert conceded, however, that none of these jobs— or any others — would be available unless the employee could maintain her concentration, persistence, or pace for at least seven hours a day. The vocational expert also conceded that no work would be available for someone who could not be punctual and follow a schedule, would miss work more than three times a month, and could not interact with the general public and get along with coworkers.
After the hearing Phillips submitted an additional mental-health assessment from Dr. Julie Handwerk. Handwerk is a psychiatrist who treated Phillips at Harrisburg from August through December 2007. She diagnosed Phillips with bipolar disorder, post-traumatic stress disorder, panic disorder, generalized anxiety disorder, and borderline personality disorder. Dr. Handwerk opined that Phillips exhibited marked limitations in nine work-related functions, including her abilities to maintain attention and concentration, to accept instructions and respond appropriately to criticism, and to maintain regular attendance. According to Dr. Handwerk, Phil*883lips’s GAF score was 45, and her prognosis was “poor.” Like Redick and Elsamahi, Dr. Handwerk also opined that Phillips was not a malingerer, that she would likely be absent from work more than three times a month, and that she was incapable of tolerating even low-stress work because of her “frequent decompensation requiring hospitalization.”
Even after considering the additional medical evidence, the ALJ denied Phillips’s claim. The ALJ performed the requisite five-step analysis, see 20 C.F.R. § 404.1520, concluding that Phillips had not engaged in gainful work since filing her application in December 2004, and that her depression with anxiety and her borderline personality disorder constituted severe impairments. At Step 3, relying principally on the checkbox form submitted by Dr. Hudspeth in May 2005, the ALJ concluded that these impairments did not meet a listing requirement and that Phillips had only a mild limitation in her activities of daily functioning and moderate limitations in social functioning and concentration. Despite Phillips’s numerous hospitalizations and suicide attempts, the ALJ also concluded that she “experienced no episodes of decompensation.” At Step 4, the ALJ explained that he incorporated these findings into Phillips’s RFC and determined that she could perform a reduced range of sedentary jobs at which she would be “limited to simple repetitive tasks, occasional contact with coworkers and supervisors, and no contact with the public.” Finally, the ALJ concluded at Step 5 that Phillips was not disabled because she could perform the work of a packer, inspector, or assembler.
In reaching these conclusions, the ALJ discredited Phillips’s statements regarding the severity of her mental impairments. The only specific ground he offered was that she had given inconsistent answers about her use of marijuana. The ALJ gave little or no weight to the opinions of Dr. Handwerk and Dr. Warshauer, commenting that Dr. Handwerk’s findings were “not consistent with the weight of the evidence presented and are only based upon a 4 month period of treatment” and that Dr. Warshauer, the state-agency psychologist, examined Phillips only once. The ALJ similarly discounted the conclusions of Elsamahi, the physician’s assistant at Harrisburg, because, according to the ALJ, they were “not consistent with the evidence contained within the file” and lacked any documentation, treatment notes, or “explanatory remarks.” The ALJ failed to mention any records from Dr. Chandra, social worker Bates, Phillips’s two hospitalizations in 2002, her emergency room visit in June 2005, her history of sexual abuse, or the fact that she has been experiencing severe depression since she was a child. The Appeals Council denied review.
Discussion
We review the ALJ’s decision deferentially and will uphold the denial of benefits unless it is not supported by substantial evidence or is based on an error of law. 42 U.S.C. § 405(g); Terry v. Astrue, 580 F.3d 471, 475 (7th Cir.2009); Rice v. Barnhart, 384 F.3d 363, 368-69 (7th Cir.2004). An ALJ need not address every piece of evidence but must build a “logical bridge” between the evidence and his findings and adequately discuss the issues so that we can evaluate the validity of the agency’s findings. Jones v. Astrue, 623 F.3d 1155, 1160 (7th Cir.2010); Villano v. Astrue, 556 F.3d 558, 562 (7th Cir.2009). We confine our review to the reasons offered by the ALJ and will not consider post-hoc rationalizations that the Commissioner provides to supplement the ALJ’s assessment of the evidence. See SEC v. Chenery Corp., 318 U.S. 80, 93-95, 63 S.Ct. 454, 87 L.Ed. 626 (1943); Spiva v. Astrue, 628 F.3d 346, 353 *884(7th Cir.2010); Campbell v. Astrue, 627 F.3d 299, 307 (7th Cir.2010).
Phillips principally contends that, under the treating source rule, the ALJ should have given controlling weight to Elsamahi’s opinion that she is unable to work because of her mental impairments. According to that rule, the opinion of a “treating source” — typically a physician— is entitled to “controlling weight” if it is adequately supported by objective medical evidence and is consistent with other substantial evidence in the record. 20 C.F.R. § 404.1527(d)(2); Schaaf v. Astrue, 602 F.3d 869, 875 (7th Cir.2010); Skarbek v. Barnhart, 390 F.3d 500, 503 (7th Cir.2004). If the ALJ discounts the opinion of a claimant’s treating physician, the ALJ must offer “good reasons” for doing so. Larson v. Astrue, 615 F.3d 744, 751 (7th Cir.2010); Bauer v. Astrue, 532 F.3d 606, 608 (7th Cir.2008). The ALJ must then go on to determine what weight to give to the opinion using the factors listed in § 404.1527(d)(2), including the length of treatment, the physician’s specialty, and the types of diagnostic tests performed. Moss v. Astrue, 555 F.3d 556, 561 (7th Cir.2009); Elder v. Astrue, 529 F.3d 408, 415 (7th Cir.2008).
But, as the Commissioner points out, not every provider will qualify as a treating source. See SSR 06-03p; Bowman v. Astrue, 511 F.3d 1270, 1274-75 (10th Cir.2008); Frantz v. Astrue, 509 F.3d 1299, 1301-02 (10th Cir.2007). Only “acceptable medical sources,” such as licensed physicians, can be characterized as treating sources whose opinions are generally entitled to controlling weight. 20 C.F.R. § 404.1513(a); SSR 06-03p. Physician’s assistants like Elsamahi are considered “other medical sources.” See 20 C.F.R. § 404.1513(d)(1); SSR 06-03p. Although the opinions of other medical sources are important and should be considered when evaluating “key issues such as impairment severity and functional effects,” their findings cannot “establish the existence of a medically determinable impairment.” SSR 06-03p. In deciding how much weight to give to opinions from these “other medical sources,” an ALJ should apply the same criteria listed in § 404.1527(d)(2). Id.; Cruse v. Comm’r of Soc. Sec., 502 F.3d 532, 541 (6th Cir.2007). Thus, Elsamahi’s assessment that Phillips was unable to work and his findings that she had marked limitations in numerous work-related abilities do not support an automatic finding of disability. Moreover, even measured against the criteria listed in § 404.1527(d)(2), we can understand how an ALJ might be skeptical of Elsamahi’s qualifications to render a psychiatric opinion. Elsamahi was involved to a significant degree in Phillips’s psychiatric care, but he is not a licensed psychiatrist or psychologist. He is a certified physician’s assistant. He has a Ph.D., but that degree is in philosophy. Thus, even though Elsamahi saw Phillips regularly, we cannot tell from this record whether the ALJ might have had good reason to give diminished weight to his views.
What is certain, however, is that the ALJ did not cite Elsamahi’s credentials or any other legitimate reason for rejecting his conclusions. Instead, the ALJ offered two cursory reasons — both flawed— for giving “little weight” to this evidence: Elsamahi’s findings, said the ALJ, are inconsistent with the rest of the medical evidence and are unexplained and unsupported by documentation or treatment notes. But the ALJ never identified what evidence contradicts Elsamahi’s findings, and, as we explain below, his conclusions — at least enough of them to support a determination of disability at Step 3 under listings 12.04 and 12.08 — are corroborated by a wealth of medical evidence accumulated over six years by a range of professionals at different facili*885ties. And, in fact, the record does include some treatment notes from Elsamahi, whose written assessments about Phillips are much more extensive than the few sentences that Dr. Hudspeth offered to explain his conclusions to the effect that Phillips remained capable of some types of work.
And that is the crux of the problem with the ALJ’s decision. The ALJ did not simply discard the conclusions of Elsamahi; rather, the ALJ’s decision belittled the views of every medical professional who treated or examined Phillips and adopted the terse conclusions of the one doctor who had never met her. Even if we put Elsamahi’s opinions aside entirely, which we should not, that would still leave many others — both treating physicians and other treating medical sources — who uniformly concluded that Phillips suffered from severe mental impairments that effectively rendered her disabled, even under the stringent standards of the Social Security Act. Most important, Dr. Handwerk, who does qualify as a treating physician, opined that Phillips was markedly limited in nine functional areas related to her work abilities. According to Dr. Handwerk, Phillips’s prognosis was poor and she was incapable of even low-stress work. The ALJ gave Dr. Handwerk’s opinion “little weight” because she treated Phillips for only four months, yet he did not explain the logic of then crediting only the opinion of Dr. Hudspeth, who never even examined Phillips. The ALJ also found that, as with Elsamahi’s opinions, the “weight of the evidence” contradicted Dr. Handwerk’s conclusions, but with the exception of Dr. Hudspeth’s evaluation offered without examination, we are unable to identify any other inconsistent findings in the record.
In fact, Dr. Handwerk’s conclusions are supported by the opinions and observations of other doctors, including the psychiatrists who treated Phillips during her hospitalizations, Dr. Warshauer, Dr. Chandru, and physician’s assistant Re-dick. Dr. Gilbert-Johnson and Dr. Qureshi, who treated Phillips during her psychiatric hospitalizations at Harrisburg, both commented on the episodic nature of her depression, which they described as consisting of several extended periods of severe depression each year and mild to moderate bouts of depression every week. Other hospital doctors relayed concerns that the volatility and severity of her impairments made it extremely difficult to manage Phillips’s case and opined that intensive forms of treatment would be necessary to help her. Even Dr. Warshauer, the state-agency clinical psychologist who examined Phillips just seven weeks before Dr. Hudspeth gave his assessment, concluded that Phillips’s borderline personality disorder was “quite severe.” Dr. Chandru concurred with the assessment of the social worker Bates that Phillips lacked impulse control and was seriously impaired by her symptom distress. And Redick echoed Dr. Handwerk’s conclusions that Phillips was very unstable, likely to miss work more than three times a month, and could not tolerate any work-related stress. Treatment notes from these medical professionals show that all of them made observations about Phillips’s condition that track Dr. Handwerk’s opinions about her limitations.
In addition to these notes from her doctors, the record is replete with objective medical evidence that Phillips suffered from a debilitating mental illness. By the time of her hearing, Phillips had attempted suicide at least six times and had been repeatedly hospitalized for severe bouts of depression and suicidal or homicidal ideation. To manage her anxiety, irritability, and depression, Phillips had tried every kind of antidepressant, according to Dr. Qureshi, and she was often prescribed a combination of four or five antipsychotic *886medications. And even though there are no details about the state’s decision to remove her children from the home, we do know that they were taken away from her at least once because she was not stable enough to care for them.
The ALJ’s assessment of the medical record also demonstrates a misunderstanding about the nature of mental illness. Each time the ALJ described Phillips’s hospitalizations in his decision, he stressed only the few hopeful observations, increased GAF scores, and any comments about her speedy improvement. What the ALJ failed to recognize is that, in every instance, Phillips improved only after the doctors had adjusted her medications and discovered a new combination of prescriptions that would adequately manage her symptoms. Many mental illnesses are characterized by “good days and bad days,” rapid fluctuations in mood, or recurrent cycles of waxing and waning symptoms. See Spiva, 628 F.3d at 351-52; Larson, 615 F.3d at 751; Bauer, 532 F.3d at 609; Kangail v. Barnhart, 454 F.3d 627, 629 (7th Cir.2006). We are also troubled by the ALJ inexplicable conclusion that Phillips had never experienced an episode of decompensation, a conclusion suggesting that the ALJ does not understand the meaning or significance of that term. See Larson, 615 F.3d at 750. Under listings 12.04 and 12.08, a claimant is deemed to have a disabling mental impairment if she has marked limitations in either two categories of functional limitation or one category coupled with evidence of three or more episodes of decompensation. 20 C.F.R. Pt. 404. Subpart P., App. 1, § 12.00. Such episodes may include incidents like hospitalizations, which signal the need for a different treatment, or they “may be inferred from medical records showing a significant alteration in medication.” Id.; Larson, 615 F.3d at 750. Because the ALJ did not explain why none of Phillips’s suicide attempts, hospitalizations, or frequent medication adjustments evidenced episodes of decompensation, he failed to build a logical bridge between the evidence and his finding that she could perform full-time work.
Most puzzling to us is that, despite this overwhelming medical evidence, the ALJ seized on the one outlier opinion in the record and this from a nonexamining source. Dr. Hudspeth’s evaluation suggests that he was unfamiliar with a significant portion of the medical record. By May 2005, when he reviewed Phillips’s file, she had been hospitalized five times for suicide attempts or suicidal thoughts, not including her attempt to take her life when she was still a child, yet he concluded she experienced only “one or two” extended episodes of decompensation. And Dr. Hudspeth’s opinion was rendered before Phillips’s two subsequent hospitalizations in June 2005 and March 2006, which is significant because Phillips’s insured status did not end until September 2009, and thus she is entitled to benefits if her impairments were disabling before the ALJ made his decision. Given these deficiencies, we cannot conclude that Dr. Hudspeth’s evaluation provides substantial evidence to support the ALJ’s RFC determination and the ultimate decision to deny benefits.
Phillips also argues that the ALJ improperly discredited her testimony regarding the severity of her symptoms. Although we generally defer to an ALJ’s credibility determination, we will overturn that finding if we conclude that it is “patently wrong.” See Prochaska v. Barnhart, 454 F.3d 731, 738 (7th Cir.2006); Sims v. Barnhart, 442 F.3d 536, 538 (7th Cir.2006). Here, in addressing Phillips’s credibility, the ALJ employed the same boilerplate language that we have often criticized, stating that although “the claimant’s medically determinable impairments could rea*887sonably be expected to produce the alleged symptoms ... [her] statements concerning the intensity, persistence and limiting effects of these symptoms are not credible to the extent they are inconsistent with” the RFC. See Spiva, 628 F.3d at 348; McClesky v. Astrue, 606 F.3d 351, 352 (7th Cir.2010); Parker v. Astrue, 597 F.3d 920, 921-22 (7th Cir.2010). The only specific reason the ALJ gave for doubting her credibility was his perception that Phillips had not always been candid about her use of marijuana. On this record it is far from certain that Phillips had tried to conceal her marijuana use — on one occasion after testing negative she still confessed to a physician that she had used the drug — but the important point is that Phillips may have been afraid to fully disclose her marijuana use to medical providers. See McClesky, 606 F.3d at 353. And what is more unsettling is the ALJ’s failure to mention record evidence that was plainly relevant to the question of Phillips’s credibility, i.e., that three of her healthcare providers— Dr. Handwerk, Elsamahi, and Redick— all unequivocally stated that she was not a malingerer, and there is no suggestion of symptom exaggeration in the record.
Finally, we note that much of the Commissioner’s appellate brief is devoted to trying to rewrite the ALJ’s decision by providing additional rationalizations not articulated by the ALJ and by attempting to bolster the significance of Dr. Hudspeth’s opinion. These post-hoc rationalizations not only undermine our confidence in the accuracy of the Commissioner’s representations of the record, but we have repeatedly warned that attempts to supplement the ALJ’s decision are inappropriate. Twice recently, we again emphasized that the government may not provide the missing justification for an ALJ’s decision. See Spiva, 628 F.3d at 353; Campbell, 627 F.3d at 307.
Accordingly, the judgment is REVERSED, and the case is REMANDED to the agency for further proceedings.