ORDER AND JUDGMENT*
After examining the briefs and appellate record, this panel has determined unanimously to grant the parties’ request for a decision on the briefs without oral argument. See Fed. R.App. P. 34(f); 10th Cir. R. 34.1(G). The case is therefore ordered submitted without oral argument.
Plaintiff-appellant Robert Oslin appeals from the district court’s order affirming the Commissioner’s denial of his applications for disability benefits and supplemental security income benefits. Oslin argues the Commissioner should have given greater weight (1) to his treating physician’s opinions concerning his physical and mental limitations and (2) to his combination of impairments. “We review the Commissioner’s decision to determine whether the factual findings are supported by substantial evidence and whether the correct legal standards were applied.” Angel v. Barnhart, 329 F.3d 1208, 1209 (10th Cir.2003). Because the correct legal standards were not applied, we reverse and remand for further proceedings.
Oslin asserts disability beginning on August 15, 1996 due to depression, bipolar disorder, and hepatitis B with accompanying liver disease. The Administrative Law Judge (ALJ) denied benefits at step 5 of the five-step sequential evaluation *944process, see 20 C.F.R. §§ 404.1520, 416.920,1 finding that although Oslin suffered from severe impairments and could not return to his past work, he retained the residual functional capacity (RFC) to perform sedentary work, limited by his inability to concentrate. The ALJ decided Oslin could perform sedentary work requiring only simple one- and two-step instructions and occasionally detailed instructions. Relying on the vocational expert’s testimony, the ALJ determined Oslin could perform work as an assembler, miscellaneous laborer, and newspaper clipper. The ALJ’s decision became the final decision of the Commissioner when the Appeals Council denied Oslin’s request for review. The district court affirmed the Commissioner’s decision.
I
Before addressing Oslin’s appellate arguments, we first consider the Commissioner’s contentions that Oslin improperly raises arguments that he did not present to the district court and improperly cites authority he did not bring to the attention of the ALJ.2 With respect to the contention that Oslin improperly raises new arguments, the Commissioner specifically submits Oslin did not argue before the district court (1) that the ALJ failed to give proper weight to Dr. Yarborough’s report; (2) that the ALJ improperly discounted Oslin’s credibility; (3) that Oslin’s medications caused problematic side effects; (4) that the ALJ should have ordered a consultative examination and therefore failed in his investigative duty; and (5) that the ALJ should have considered Oslin’s impairments in combination. Extremely liberal reading of Oslin’s district court brief allows us to conclude Oslin barely sufficiently raised the first, third, and fifth arguments by implication.3 Oslin argued without specificity that the ALJ disregarded his treating physician’s opinions. App. vol. 1, at 30. Although he listed his medications in his district court brief, he did not expressly assert that they caused side effects. Id. at 28-30. While not raised with precise clarity, Oslin did assert that the ALJ found his allegations not fully credible. Id. at 27, 28. Generally, as the district court recognized, Oslin did argue the ALJ’s decision was not supported by substantial evidence. Prior to this appeal, however, Oslin, who at all times has been represented by counsel, has never asserted the ALJ should have ordered a consultative examination and the ALJ failed to fulfill his investigative duty. We will consider this newly raised argument only if there are compelling reasons to do so. See Crow v. Shalala, 40 F.3d 323, 324 (10th Cir.1994). As we discuss below, there are compelling reasons to deviate from our general rule. See id. Thus, we consider all issues Oslin raises on appeal.
The Commissioner also contends this court should not consider new exhibits included in Oslin’s appendix that were not *945presented to the ALJ. See App. vol. 1, at 62A7-62A26. Our review is limited to the administrative record. See 42 U.S.C. § 405(g); see also O’Dell v. Shalala, 44 F.3d 855, 858-59 (10th Cir.1994) (reviewing evidence presented to Appeals Council and to ALJ). Even if we were to consider the exhibits, they are merely on-line medical publications describing some of Oslin’s medications and their possible side effects. Thus, the exhibits alone cannot establish that Oslin actually suffers side effects from his medications. Nor does Oslin even cite to them in his appellate briefs.
II
Having addressed these preliminary matters, we turn now to Oslin’s first argument that the ALJ should have given greater weight to the opinions of his treating physicians. “An ALJ is required to give controlling weight to a treating physician’s well-supported opinion, so long as it is not inconsistent with other substantial evidence in the record.” Drapeau v. Massanari, 255 F.3d 1211, 1213 (10th Cir. 2001); see also 20 C.F.R. §§ 404.1527(d)(2); 416.927(d)(2). “When an ALJ decides to disregard a medical report by a claimant’s physician, he must set forth specific, legitimate reasons for his decision.” Drapeau, 255 F.3d at 1213 (quotation omitted). “In choosing to reject the treating physician’s assessment, an ALJ may not make speculative inferences from medical reports and may reject a treating physician’s opinion outright only on the basis of contradictory medical evidence and not due to his ... own credibility judgments, speculation or lay opinion.” McGoffin v. Barnhart, 288 F.3d 1248, 1252 (10th Cir.2002) (quotations and italics omitted).
A
Oslin first contends that although the ALJ agreed with the exertional limitations set forth by Dr. Yarborough, that Oslin can lift and carry up to ten pounds and can sit for six hours, stand for two hours, and walk for one hour during an eight-hour day, the ALJ improperly rejected the opinions of Dr. Yarborough and Dr. Briggs that he could not work. Dr. Yarborough stated in an undated letter that Oslin was currently unable to work. Neither this brief, conclusory letter, nor other record evidence specifically indicated why Dr. Yarborough believed Oslin could not work. His opinion therefore may be rejected. See Castellano v. Sec’y of HHS, 26 F.3d 1027, 1029 (10th Cir.1994) (“A treating physician’s opinion may be rejected if his conclusions are not supported by specific findings.”). Further, the exertional limits Dr. Yarborough set forth do not preclude sedentary work. Cf. 20 C.F.R. §§ 404.1567(a), 416.967(a) (defining sedentary work as lifting up to ten pounds, mostly sitting, and walking and standing occasionally). Dr. Yarborough referred Oslin to Dr. Briggs for treatment of hepatitis B and liver disease. That treatment included a short, but discontinued, Interferon treatment. In a letter dated October 23, 1997, Dr. Briggs stated that Oslin could work at a desk job, because he had not experienced any of the end-stage complications of liver disease.
Other medical evidence noted Oslin had complained of joint pain of increasing severity, primarily in his knees, elbows and back, for two years, but he did have a normal range of motion in his joints. Dr. Yarborough prescribed medication for the joint pain. At the ALJ’s hearing, Oslin himself testified that his joint pain is only a 2 or 3 on a scale of 1 to 10 when he takes pain medication. At the time of the ALJ proceedings, the medical evidence did not establish that Oslin’s physical impairments limited his ability to perform sedentary *946work. Accordingly, we conclude the ALJ correctly assessed the opinions of Dr. Yarborough and Dr. Briggs.
B
Oslin next challenges the ALJ’s failure to specifically address the mental RFC form completed by Dr. Roman and Ms. Snook, his psychiatrist and case manager at Parkside Clinic. The form indicated that Oslin would have marked to moderate limitations in performing basic work functions, his depression would interfere with the activities of daily living, and he cannot maintain a stable living for more than a few months at a time.4 Although the ALJ recognized that Dr. Roman and Ms. Snook had completed this mental assessment, which indicated Oslin is unable to function in a work setting,5 the ALJ rejected that assessment as inconsistent with the objective medical evidence, pointing to evidence that Oslin was doing well and had no problems with sleep or hallucinations. ALJ’s decision at 5. Also, the ALJ found that Dr. Roman’s diagnosis of a moderate bipolar disorder was inconsistent with a complete inability to do work-related mental activities. Instead, the ALJ found on his Psychiatric Review Technique form and recited in his decision that Oslin has a slight limitation in performing daily activities and in social functioning, he often has problems concentrating, and he never has episodes of deterioration or decompensation in work settings.
Contrary to the ALJ’s findings, the totality of the Parkside Progress Notes, completed by Dr. Roman, Ms. Snook, and others, do not conclusively establish that Oslin always does well and has no problems. Rather, the progress notes suggest he experienced changing symptoms and no sustained periods of improved functioning. The ALJ, however, did not address or explain his rejection of this evidence. Nor does he assess whether Oslin’s changing symptoms permit him to perform work on a regular, consistent, or sustained basis.
Likewise, the ALJ disregarded Oslin’s Global Assessment of Functioning (GAF)6 *947scores, ranging between 45 and 55. A GAF rating between 41 and 50 indicates “[sjerious symptoms (e.g., suicidal ideation ...) OR any serious impairment in social, occupational, or school functioning (e.g., ... unable to keep a job);” whereas a GAF rating between 51 and 60 indicates moderate symptoms or impairments in these categories. DSM-IV-TR at 34 (bold print omitted). Although the GAF rating may indicate problems that do not necessarily relate to the ability to hold a job, Dr. Roman and Ms. Snook indicated Oslin could not work or his impairments interfered with his ability to keep a job. GAF scores of 50 or less do suggest an inability to keep a job. Id. The ALJ did not discuss the GAF scores and did not provide a reason to disregard the opinion of Oslin’s treating source in this respect. See Frey v. Bowen, 816 F.2d 508, 513 (10th Cir. 1987).
Although the ALJ did not specifically discuss the complete set of mental progress notes and the GAF scores in his decision, he did indicate that he had considered all of the evidence.7 See Clifton v. Chater, 79 F.3d 1007, 1009-10 (10th Cir. 1996) (“The record must demonstrate that the ALJ considered all of the evidence, but an ALJ is not required to discuss every piece of evidence.”). Nonetheless, the ALJ “may not ignore evidence that does not support his decision, especially when that evidence is significantly probative.” Briggs ex rel. Briggs v. Massanari, 248 F.3d 1235, 1239 (10th Cir.2001) (quotation omitted); see also Clifton, 79 F.3d at 1010 (“[I]n addition to discussing the evidence supporting his decision, the ALJ also must discuss the uncontroverted evidence he chooses not to rely upon, as well as significantly probative evidence he rejects.”). Because the ALJ’s discussion of the mental impairment evidence does not mention the GAF scores or fully discuss the mental progress notes, it is unclear whether the ALJ considered or rejected this evidence, favorable to Oslin.
We agree with Oslin’s assertion that the ALJ failed to properly consider Oslin’s subjective complaints, which are not inconsistent with the conclusions of his mental health treating sources. Oslin testified at the hearing that three to four times a month, and lasting for two to four days, his mental problems preclude him from functioning and his bipolar disorder is not completely controlled with medication. He also complained of drowsiness caused by his medication. Oslin’s complaints of drowsiness, depression, and moodiness were affirmed by the testimony of Bruce Beggley, who is the Senior Executive Director of the HIV ER Center where Oslin volunteers three days a week for two to three hours, and who had allowed Oslin to stay in his home. Lumping all of Oslin’s subjective complaints together, and using boiler-plate language, see, e.g., Angel, 329 F.3d at 1213, the ALJ rejected Oslin’s allegations of limitations as not credible “because, but not limited to, the objective findings, or the lack thereof, by treating and examining physicians, the lack of med*948ication for severe pain, the frequency of treatments by physicians and the lack of discomfort shown by the claimant at the hearing.”8 ALJ’s decision at 3-4.
“Although this court ordinarily defers to the ALJ as trier of fact on credibility, ... deference is not an absolute rule.” Thompson v. Sullivan, 987 F.2d 1482, 1490 (10th Cir.1993); see also Briggs, 248 F.3d at 1239 (“Credibility determinations are peculiarly the province of the finder of fact, and we will not upset such determinations when supported by substantial evidence.”) (quotation omitted). The ALJ’s analysis was legally insufficient. See White v. Barnhart, 287 F.3d 903, 909 (10th Cir.2001) (recognizing ALJ must give specific reasons when rejecting claimant’s subjective complaints); Kepler v. Chater, 68 F.3d 387, 390-91 (10th Cir.1995) (rejecting ALJ’s conclusory decision dismissing claimant’s subjective complaints without explaining reasoning and specific evidence underlying that determination). Because “[a] significant portion of the record supports [Oslin’s] testimony,” “the ALJ must explain why he has determined that the testimony is not credible.” Briggs, 248 F.3d at 1239.
The ALJ also failed to explain adequately why he would discredit Oslin’s testimony concerning his daily activities. Oslin testified that he sometimes cooks, does dishes, helps with sweeping and vacuuming, does cross-stitch as long as his concentration continues, collects Barbie dolls, goes to the flea market every three or four weeks, and volunteers two to three hours three times per week at the HIV ER Center. Oslin’s minimum daily activities do not establish he can work on a consistent basis. See Thompson, 987 F.2d at 1490 (“sporadic performance of [household tasks or work] does not establish that a person is capable of engaging in substantial gainful activity”) (quotation omitted).
Oslin argues the ALJ should have obtained a consultative examination and fulfilled his investigative duty. “It is the ALJ’s duty to investigate the facts and develop the arguments both for and against granting benefits.” Sims v. Apfel, 530 U.S. 103, 111, 120 S.Ct. 2080, 147 L.Ed.2d 80 (2000). “This duty to develop the record pertains even if the claimant is represented by counsel.” Thompson, 987 F.2d at 1492; see also Hawkins v. Chater, 113 F.3d 1162, 1167 (10th Cir.1997) (“the ALJ may ordinarily require counsel to identify the issue or issues requiring further development”). “In the absence of such a request by counsel, we will not impose a duty on the ALJ to order a consultative examination unless the need for one is clearly established in the record.” Hawkins, 113 F.3d at 1168. If evidence from the claimant’s treating doctor is inadequate to determine if the claimant is disabled, the Commissioner must first recontact the treating doctor to determine if additional needed information is available. See 20 C.F.R. § 404.1512(e)(1) (“We will seek additional evidence or clarification from your medical source when the report from your medical source contains a conflict or ambiguity that must be resolved, the report does not contain all the necessary information, or does not appear *949to be based on medically acceptable clinical and laboratory diagnostic techniques.”); id. § 416.912(e)(1); see also White, 287 F.3d at 905 (“obligation [to recontact the treating physician] arises whenever the information received from the treating doctor is inadequate”). Where the medical evidence in the record is inconclusive, “a consultative examination is often required for proper resolution of a disability claim.” Hawkins, 113 F.3d at 1166; see also 20 C.F.R. §§ 404.1512(f) (“If the information we need is not readily available from the records of your medical treatment source, or we are unable to seek clarification from your medical source, we will ask you to attend one or more consultative examinations at our expense.”); id. § 416.912(f).
Here, the Parkside Clinic progress notes are insufficient to draw reliable conclusions about Oslin’s alleged disability, inasmuch as they indicate at times that Oslin is improved and doing well, but at other times indicate that he is having problems with depression, mania, medication side effects, hallucinations, suicidal thoughts, anxiety, memory and concentration. The progress notes do, however, establish that Oslin presented sufficient mental health medical evidence to warrant further investigation. See Hawkins, 113 F.3d at 1169. We remand to the ALJ with directions to obtain a detailed evaluation from Dr. Roman with a clear assessment of Oslin’s condition and/or a detailed evaluation from a consulting doctor who personally examines Oslin. The further information should clarify how Oslin’s mental problems impact his ability to work. Also, in remanding, we direct the ALJ to make specific findings fully discussing the medical evidence.
C
Oslin argues the ALJ should have considered his physical and mental impairments in combination. In light of our decision to remand for further proceedings, we also direct the ALJ to again consider the totality of the evidence after more fully developing the record with respect to Oslin’s mental impairments. The judgment of the district court is REVERSED, and this case is REMANDED to the district court with directions to REMAND to the Commissioner for further proceedings assessing Oslin’s credibility, the effect of his mental impairments on his ability to work, and the combined effect of his physical and mental impairments.