ORDER
Rosemary Harlin filed for supplemental security income, claiming disability based on depression. An administrative law judge found that Harlin’s impairments met Social Security’s disability listing 12.09, but determined that she was not disabled because her substance-use disorder materially contributed to her disability. See 42 U.S.C. § 1382c(a)(3)(J); 20 C.F.R. § 416.935. Because the ALJ neither adequately explained why he discounted the opinion of Harlin’s treating psychiatrist nor supported his conclusion that Harlin’s cocaine use materially contributed to her disability, we vacate and remand.
Background
Harlin’s medical record reflects a lengthy battle with depression and substance abuse, and numerous hospitalizations for related symptoms. The first record of her treatment for depression dates back to 1993, when she was hospitalized after a suicide attempt. In the discharge summary, a social worker characterized Harlin as having “psychoactive substance abuse mood disorder” and recounted that Harlin described a previous suicide attempt in 1985.
From 2004 (when the medical record next picks up) until 2007, Harlin was treated by psychiatrist Dr. Bharathi Marri. In a 2005 psychiatric evaluation, Dr. Marri noted that Harlin had a history of depression and cocaine abuse, did not take prescribed medications, and reported using cocaine last in October 2004. In this report Dr. Marri scored Harlin at “60-70” on the Global Assessment of Functioning (“GAF”), a scale used to rate, among other things, the psychological functioning of adults. In April 2007 Dr. Marri completed a form, opining that Harlin’s “depression and relationship problems [led to] substance abuse” and that mental impairment was the “primary cause of disability.” The doctor also thought that Harlin would suffer significant mental impairments even without the use of cocaine because she had attended a partial hospitalization program *566that kept her drug-free yet she continued to experience emotional problems.
Harlin was hospitalized four times between 2005 and 2007 with symptoms of major depression. During each hospitalization, medical records note both Harlin’s depression and her cocaine use. Her symptoms included suicidal thoughts, anxiety, crying, feelings of helplessness and hopelessness.
In 2007 the ALJ held a hearing at which Harlin, a state-agency psychologist, and a vocational expert testified. Harlin described the debilitating effects of her depression, including anxiety attacks, attempts to physically harm herself, and her disinclination to bathe, dress, groom herself, or go outside for days. She explained that she had been off of drugs for three years but still had “trouble focusing on anything.” Dr. Ellen Rozenfeld, the state-agency psychologist, testified that Harlin’s records reflected an affective disorder, but that information was insufficient to determine whether it was bipolar or major depressive. The doctor also questioned Harlin’s truthfulness regarding her drug abuse; she explained that hospital records did not support Harlin’s claim of being sober for three years. Addressing Dr. Marri’s opinion that Harlin’s capacities would be significantly impaired even without substance abuse, Dr. Rozenfeld said she lacked sufficient documentation to give an opinion. Dr. Rozenfeld also observed that Harlin responded well to “treatment,” but did not know whether she could sustain this level of improvement. A vocational expert also testified, concluding that a worker with Harlin’s purported limitations could perform 10-8258 Page 3 the work of office cleaner (5,000 jobs), assembler (8,000 jobs), and packager (5,000-6,000 jobs).
At the hearing’s close, the ALJ asked Harlin to submit additional documents to address Dr. Rozenfeld’s view that the medical record was insufficient to assess whether Harlin’s drug use materially contributed to her disability. After the hearing, Harlin submitted to the ALJ medical documents, including progress notes from Dr. Marri and documentation of an additional hospitalization. The ALJ did not pass along these documents to Dr. Rozenfeld.
The ALJ eventually determined that Harlin was not disabled under the Social Security Act. Applying the five-step process that governs review of disability determinations, 20 C.F.R. § 416.920(a), the ALJ found that Harlin had not engaged in gainful employment since the alleged onset date (step 1); that she had severe impairments (cocaine abuse, depression, and seizure disorder) (step 2); and that her impairments, “including the substance-use disorder,” medically equaled a listed impairment under listing § 12.09. See 20 C.F.R. Pt. 404, Subpart P, App. 1, § 12.09, (step 3). Because of Harlin’s substance addiction, the ALJ next assessed whether Harlin would still be disabled if she stopped using drugs, see 20 C.F.R. § 416.935(b), and the ALJ found that she would not. At step four, the ALJ found that although Harlin had no past relevant work, she had the residual functional capacity to “lifi/carry 20 pounds occasionally and 10 pounds frequently” and “to maintain concentration, persistence and pace 87% of the workday.” At step five, he concluded she had the residual functional capacity — assuming that she stopped the cocaine use — to perform a large range of light work in the Chicago metropolitan area. The Appeals Council declined review and the magistrate judge, proceeding with the parties’ consent, upheld the AL J’s determination; thus the ALJ’s ruling is the final decision of the Commissioner of *567Social Security. See Liskowitz v. Astrue, 559 F.3d 736, 739 (7th Cir.2009).
Discussion
This appeal is complicated by Harlin’s history of cocaine use. Congress eliminated alcoholism or drug addiction as a basis for obtaining social security benefits: “[A]n individual shall not be considered to be disabled for purposes of this subchapter if alcoholism or drug addiction would ... be a contributing factor material to the Commissioner’s determination that the individual is disabled.” See 42 U.S.C. § 1382c(a)(3)(J); see also 20 C.F.R. § 416.935. In other words the inquiry for the ALJ is whether “were the applicant not a substance abuser, she would still be disabled.” Kangail v. Barnhart, 454 F.3d 627, 628-29 (7th Cir.2006)(citing 20 C.F.R. § 416.935); Brueggemann v. Barnhart, 348 F.3d 689, 693 (8th Cir.2003); Bustamante v. Massanari, 262 F.3d 949, 954 (9th Cir.2001); Drapeau v. Massanari, 255 F.3d 1211, 1214 (10th Cir.2001). The claimant bears the 10-3258 Page 4 burden of proving that alcoholism or drug addiction is not a contributing factor. Kluesner v. Astrue, 607 F.3d 533, 537 (8th Cir.2010); Parra v. Astrue, 481 F.3d 742, 748 (9th Cir.2007); Brown v. Apfel, 192 F.3d 492, 498 (5th Cir.1999).
On appeal, Harlin first argues that the ALJ erred by not giving controlling weight to the opinion of her treating physician, Dr. Marri. Had the ALJ properly credited Dr. Marri’s opinion, Harlin says, the ALJ would have concluded that her depression was disabling, even during periods of sobriety. Harlin maintains that, contrary to the ALJ’s opinion, Dr. Marri did not underestimate the frequency of her cocaine use. Harlin points to Dr. Marri’s April 2007 report, in which the doctor diagnosed her with cocaine dependency, noted that she last used drugs three weeks earlier, and specified that her drug use was intermittent.
A treating source’s opinion is entitled to “controlling weight” if it is adequately supported by objective medical evidence and consistent with other substantial evidence in the record. 20 C.F.R. § 416.927(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, he must offer “good reasons” for doing so. Larson v. Astrue, 615 F.3d 744, 749 (7th Cir.2010); Bauer v. Astrue, 532 F.3d 606, 608 (7th Cir.2008).
The ALJ’s reasons for discounting Dr. Marri’s opinion are inconsistent. Apparently believing that Harlin’s drug use was extensive, the ALJ discounted Dr. Marri’s assessment because Dr. Marri was “misinformed by the claimant that the drug use was intermittent.” Elsewhere in the decision, however, the ALJ appears to accept the view that Harlin’s drug use was intermittent; he acknowledged that Harlin enjoyed periods of sobriety. For instance, the ALJ cited Dr. Marri’s progress notes from a “period after the claimant was not using cocaine.... ” Elsewhere the ALJ cited a discharge summary from a hospitalization from June 2007, in which the ALJ noted Harlin’s elevated GAF score and attributed it to her sobriety during her week-long hospital stay. These examples give us pause as to whether the ALJ adequately justified downplaying Dr. Marri’s opinion.
Harlin next argues that the ALJ improperly substituted his opinion for that of Dr. Rozenfeld when he predicted how she would have interpreted the additional records submitted after the hearing. Harlin contends that the ALJ overstepped his role by opining that the doctor would have concluded that these records — in combination with the rest of the longitudal record — “clearly preponderate in favor of the *568proposition that the claimant’s drug abuse is material to her disability.” Harlin says that ALJs should not “play the role of doctor and interpret medical evidence, when he or she is not qualified to do so.”
To the extent that the ALJ projected how Dr. Rozenfeld’s would have testified had she seen the additional documents, the ALJ improperly assumed the role of doctor. Larson, 615 F.3d at 749; Kangail, 454 F.3d at 629. Additionally, it is questionable that the additional records would have altered Dr. Rozenfeld’s opinion because her testimony at the hearing acknowledged all the findings that the ALJ gleaned from the records. These records, the ALJ stated, revealed Harlin’s history of cocaine abuse, her distortion of that history, and her capacity for improvement when medicated as she was during hospitalizations. But Dr. Rozenfeld made these very same points during her testimony.
Harlin also argues the ALJ did not apply the correct legal standard with regard to her drug use. The ALJ applied § 461.935(b) to determine whether her cocaine abuse was a material factor contributing to her depression. Harlin maintains that the regulation was not properly applied because an Emergency Teletype (an internal instruction), issued in 1996 by the Social Security Administration to all adjudication components of the agency, directs a finding of disability unless evidence establishes that the claimant would not be disabled if she stopped using drugs or alcohol. Social Security Teletype, No. EM-96200 at answer 27-29 (Aug. 30, 1996). She also points to an Eighth Circuit decision remarking that “on the issue of materiality of [cocaine abuse], a tie goes to the claimant.” Brueggemann, 348 F.3d at 693 (8th Cir.2003).
Assuming without deciding that the Teletype provisions govern the situation, compare Salazar v. Barnhart, 468 F.3d 615 (10th Cir.2006) (adopting the standard articulated in the Teletype and consequently reversing because the ALJ’s conclusion that the claimant would not be disabled in the absence of substance abuse was not supported by substantial evidence) with Parra, 481 F.3d at 748-50 (treating the Teletype as neither binding nor entitled to deference, and declining to follow it because it “effectively subsidizes substance abuse in contravention of the statute’s purpose”), we believe that the ALJ here has not adequately disentangled the effects of Harlin’s drug abuse from those of her other impairments. The only record evidence reflecting Harlin’s impairments during periods of sobriety is Dr. Marri’s opinion that Harlin’s cocaine use was not a contributing factor material to her depression. As noted earlier, however, the ALJ was inappropriately dismissive of Dr. Marri’s opinion. Further, the evidence that the ALJ chose instead to rely on — discharge summaries showing Harlin’s improved condition at the time of discharge — was hardly remarkable because one would expect a patient with severe mental impairments to improve upon a course of treatment in a structured hospital environment. See e.g., Salazar, 468 F.3d at 624; see also Punzio v. Astrue, 630 F.3d 704, 710 (7th Cir.2011); Bauer v. Astrue, 532 F.3d 606, 609 (7th Cir.2008). The case must be remanded so that the ALJ can reconsider whether Harlin has shown that she would be disabled in the absence of her drug abuse. On remand, the ALJ should also reconsider how much weight to accord Dr. Marri’s opinion, using the factors identified in 20 C.F.R. § 416.927(d)(2).
The judgment is VACATED and the case REMANDED to the Social Security Administration for further proceedings consistent with this order.