CARLOS F. LUCERO, Circuit Judge.

Dawn Knight, on behalf of her daughter, P.K., appeals from a district court order affirming the Commissioner's denial of P.K.'s application for Supplemental Security Income benefits ("SSIB"). Exercising jurisdiction under 28 U.S.C. § 1291 and 42 U.S.C. § 405(g),1 we reverse and remand for further proceedings.

I

Knight sought SSIB for P.K. in April 2007, when P.K. was nine years old, claiming that her daughter suffered from a learning disability, hearing loss, attention deficit hyperactivity disorder ("ADHD"), and "temper."

In July 2007, Dr. David LaCourt evaluated P.K. for the New Mexico Disability Determination Services Unit. He noted that P.K. was taking Dextroamphetamine for her ADHD and "had written what was taken as a self-harm note while she was at a doctor visit recently."2 P.K. took a reading test and scored at the second-grade level, despite being in the fourth grade at the time.

P.K.'s third and fourth grade teachers filled out functional-assessment questionnaires indicating that P.K. had obvious or serious problems acquiring and using information. The fourth grade teacher additionally reported that she had "to implement behavior modification strategies" for P.K. because P.K. had been cursing during recess and spreading "malicious gossip."

In September 2009, P.K.'s fifth-grade teacher completed the same questionnaire, finding few obvious or serious problems when it came to P.K.'s abilities to acquire and use information. She did note, however, that P.K. "is very aggressive on the playground and uses bad/inappropriate language outside the classroom."

Dr. E.B. Hall managed P.K.'s medication regimen from 2007 through 2010. His progress notes for that period reveal instances in which P.K. threatened a sibling with a knife, hit another child in the face, was "hearing voices," harmed herself, experienced mood swings, and failed to take her medication. But Dr. Hall also made several notations indicating that P.K. was getting good grades at school. According to Dr. Hall, P.K. was extremely impaired in the areas of attending and completing tasks and interacting and relating with others. She was markedly impaired regarding her overall health and physical well-being, acquiring and using information, and caring for herself.

In March 2010, when P.K. was in the sixth grade, she was examined at Hogares, Inc., a mental-health evaluation and treatment center for children. There, P.K. told the therapist that "she hears people calling her name, thinks that other people are talking about her when they are not, and often talks about killing herself." P.K. also said that "she often feels like nothing is ever going to change or get better and it is not worth getting out of bed." Knight reported to the Hogares therapist that "she [has] found notes stating [that] [P.K.] wanted to kill herself"; that P.K. "scratches out her face in pictures"; that P.K. "does not take her medication without supervision"; and that it is difficult to take P.K. out in public because "she will lay herself down and throw a fit like a two year old."

The therapist diagnosed P.K. with ADHD, oppositional defiant disorder, adjustment disorder with anxiety, and bipolar disorder. The therapist noted that P.K. was currently "doing poorly in school," apparently due to her ADHD.

In June 2010, Knight and P.K. appeared before an administrative law judge ("ALJ"). Knight testified that P.K. fights with the other children in the family; "goes after the [other] kids with a knife"; "likes to try to kill herself"; has broken windows and punched holes in the walls where they've lived; "throw[s] a fit" when asked to help around the house or when they go out in public; and resists taking her medications. Knight also testified that while P.K. was in sixth grade, the school called at least once a week about problems with P.K. fighting, "focusing, staying on task, and not interrupting the classroom." According to Knight, P.K. had no long-term childhood friends and had only recently made a friend.

P.K. testified that she argues with her siblings, has threatened her brother with scissors, has temper tantrums, and has to be reminded to take her medications, brush her teeth, and take a shower. Regarding her medications, P.K. stated that "[s]ometimes [she] get[s] better, but" then "get[s] used to the pills" and they do not help her. She also believed that she did well in sixth grade, but admitted that she had trouble with reading.

The ALJ found that P.K. suffered from hearing loss, a reading disorder, and ADHD, but he concluded that none of her impairments, either singularly or in combination, medically or functionally equaled an impairment listed in the regulations.3 In doing so, the ALJ found that P.K's health problems could "produce the alleged symptoms," but that "statements concerning the intensity, persistence and limiting effects of [P.K.'s] symptoms are credible [only] to the extent they are consistent with finding" her not disabled. He rejected the diagnoses offered by the Hogares therapist, stating that they were not from an acceptable medical source. He also rejected Dr. Hall's opinions, explaining that Dr. Hall "only provides medication management and not therapy"; his opinions "evidently rel[y] heavily on subjective reports concerning areas outside of the scope of his observation"; and his opinions "contrast[ ] sharply with the other evidence of record."

Knight sought review from the Appeals Council, submitting additional therapy notes from Hogares.4 Those notes state that P.K. inconsistently takes her medication and that one of the goals of therapy was to "stop [P.K. from] talking about hurting herself and others." Additionally, Knight had reported to the therapist that the family was "being kicked out of their current living situation because [P.K.] broke a window and punched a hole in the wall." The Appeals Council denied review.

Knight then petitioned the district court for review. A magistrate judge recommended denying the petition. The district court adopted the recommendation and dismissed Knight's petition. This appeal followed.

II

Our review is limited to determining whether the Commissioner applied the correct legal standards and whether the agency's factual findings are supported by substantial evidence. See Barnett v. Apfel, 231 F.3d 687, 689 (10th Cir.2000). Substantial evidence is "more than a scintilla, but less than a preponderance." Lax v. Astrue, 489 F.3d 1080, 1084 (10th Cir. 2007). A decision is not based on substantial evidence "if it is overwhelmed by other evidence in the record." Wall v. Astrue, 561 F.3d 1048, 1052 (10th Cir.2009) (quotation omitted). In conducting our review, "[w]e may neither reweigh the evidence nor substitute our judgment for that of the agency." Barnett, 231 F.3d at 689 (quotation omitted).

Under the Personal Responsibility and Work Opportunity Reconciliation Act ("PRAWOR"), a child is disabled if he or she has a "physical or mental impairment, which results in marked and severe functional limitations, and ... which has lasted or can be expected to last for a continuous period of not less than 12 months." 42 U.S.C. § 1382c(a)(3)(C)(i). If the child is not engaged in substantial gainful activity and has a severe impairment, the child meets the PRAWOR definition if his or her impairment is medically or functionally equivalent in severity to any in the Listings of Impairments contained in 20 CFR pt. 404, subpt. P, App. 1. See 20 C.F.R. § 416.924(a).

The ALJ found that P.K. did not have an impairment that medically equaled a listed impairment, and then assessed functional equivalency. To find an impairment functionally equivalent to one in the list, an ALJ must analyze the impairment's severity in six age-appropriate domains: (1) acquiring and using information; (2) attending and completing tasks; (3) interacting and relating with others; (4) moving about and manipulating objects; (5) caring for oneself; and (6) health and physical well-being. Id. § 416.926a(b)(1). The ALJ must find an extreme limitation in one domain or a marked limitation in two domains to satisfy functional equivalency. Id. § 416.926a(d). Finally, the ALJ must consider "[t]he interactive and cumulative effects of an impairment or multiple impairments." Id. § 416.926a(c).

On appeal, Knight argues that the ALJ improperly discounted her credibility, Dr. Hall's opinions, and the functional equivalency of P.K.'s impairments.

III

Knight asserts that the ALJ's credibility determination was flawed. We agree.

In determining whether a child is disabled, the agency will accept a parent's statement of a child's symptoms if the child is unable to adequately describe them. 20 C.F.R. § 416.928(a). "In such a case, the ALJ must make specific findings concerning the credibility of the parent's testimony, just as he would if the child were testifying." Briggs ex rel. Briggs v. Massanari, 248 F.3d 1235, 1239 (10th Cir. 2001).

The ALJ discussed little of the hearing testimony and mentioned credibility only generically, commenting that "statements concerning the intensity, persistence and limiting effects of [P.K.'s] symptoms are credible to the extent they are consistent with finding" P.K. not disabled. It is unclear from this statement whether the ALJ was referring to the credibility of Knight, P.K., or both. The agency argues that the ALJ was "clearly address[ing]" Knight's credibility because she "completed the paperwork accompanying P.K.'s claim" and because "Knight provided the most substantive testimony." In the absence of the required "specific findings," we fail to see how either Knight's completion of paperwork or the length of her testimony has any bearing on whose testimony the ALJ found not credible.

Moreover, "findings as to credibility should be closely and affirmatively linked to substantial evidence and not just a conclusion in the guise of findings." Raymond v. Astrue, 621 F.3d 1269, 1273 (10th Cir.2009) (quotation omitted). We are left to guess what evidence, if any, belies either Knight's or P.K.'s testimony. The ALJ simply offered a boilerplate credibility assertion without any reference to the evidence. The agency does not contend that the ALJ's error is harmless, and we will not fashion a party's arguments. See Ingram v. Faruque, 728 F.3d 1239, 1251 n. 6 (10th Cir.2013). In the absence of credibility findings, we reverse and remand for a proper credibility determination.

IV

A

Knight challenges the ALJ's determination that P.K. has only a marked limitation in interacting and relating with others. She contends the evidence shows an extreme limitation, which would, by itself, compel a finding of disability. See 20 C.F.R. § 416.926a(d). Knight argues that "[t]he ALJ improperly diminished the importance of Dr. Hall's opinion," which characterized P.K.'s limitation as "extreme."

The difference in degree between a marked limitation and an extreme one depends on whether the impairment "seriously" interferes with the child's ability to independently initiate, sustain, or complete activities (a marked impairment), or whether the impairment "very seriously" interferes with that ability (an extreme impairment). See id. § 416.926a(e)(2) & (3). Our role is limited to determining whether substantial evidence supports the ALJ's decision; we do not reweigh the evidence. Nonetheless, we cannot tell whether the required quantum of evidence is present because the ALJ discounted Dr. Hall's opinion without sufficient explanation.

A treating physician's opinion must be given controlling weight if it "is supported by medically acceptable clinical and laboratory diagnostic techniques and is not inconsistent with other substantial evidence in the record." Hamlin v. Barnhart, 365 F.3d 1208, 1215 (10th Cir.2004). When a treating physician's opinion is not given controlling weight, the ALJ must explain what weight, if any, was assigned to the opinion "using all of the factors provided in 20 C.F.R. §§ 404.1527 and 416.927."5 Watkins, 350 F.3d at 1300 (quotation omitted). Specifically, "the ALJ must give good reasons in the notice of determination or decision for the weight he ultimately assigns the opinion," and if he "rejects the opinion completely, he must then give specific, legitimate reasons for doing so." Id. at 1301 (quotations and alteration omitted).

In this case, the ALJ found that Dr. Hall's opinions were "less persuasive" for three reasons. First, the ALJ observed that Dr. Hall "evidently relie[d] heavily on subjective reports." Although subjective reports do not constitute "medically acceptable clinical and laboratory diagnostic techniques" and can support giving a treating physician's opinion less than controlling weight, see Langley v. Barnhart, 373 F.3d 1116, 1120 (10th Cir.2004), Knight's subjective reports concerning P.K.'s troubling behavior issues are consistent and pervasive throughout the record. P.K. has difficulty making friends, is "very aggressive" and spreads "malicious gossip" on the school playground, and not only fights with her siblings, but apparently has wielded a knife while doing so. The ALJ did not expressly reject any of these reported behaviors or discuss how they failed to support Dr. Hall's finding that P.K. is extremely impaired when it comes to interacting and relating with others.

The ALJ's second reason for discounting Dr. Hall's opinions was that he "only provide[d] medication management and not therapy." But the ALJ failed to indicate how Dr. Hall's opinion was compromised by not providing therapy. Dr. Hall, a psychiatrist, treated P.K. for three years, managing P.K.'s intake of multiple potent psychostimulant, antipsychotic, and antidepressant medications. This constitutes a valid treatment relationship with a specialist, and the ALJ should have provided further explanation as to its effect on the weight of Dr. Hall's opinion.

Finally, the ALJ discounted Dr. Hall's opinions because they "contrast[ed] sharply with the other evidence." We cannot discern from the ALJ's decision what that evidence is in the context of P.K.'s ability to interact and relate with others. The ALJ takes issue with Dr. Hall's opinion because it "does not reconcile with the consistent reports of [P.K.] making good grades in school" and "complet[ing] her homework." Those reports, however, have little to no bearing on a child's ability to interact and relate with others. And other evidence the ALJ cited actually supports Dr. Hall's opinion for this domain: The ALJ noted that P.K.'s "behavior is her most significant problem," as demonstrated by her aggressive play, fighting with siblings, and "anger outburst[s]."

Because the ALJ's decision is deficient in its treatment of Dr. Hall's opinion about the severity of P.K.'s limitations in interacting with and relating to others, we reverse and remand for a proper analysis.

B

Knight challenges the ALJ's determination that P.K. has less than a marked limitation in the domain of caring for herself. This domain focuses on the child's "personal needs, health and safety." 20 C.F.R. § 416.926a(k)(1)(i). It requires that the child use age-appropriate strategies "to identify and regulate [her] feelings, thoughts, urges, and intentions," id. § 416.926a(k)(1)(iii), and to respond to "circumstances in safe and appropriate ways, making decisions that do not endanger [herself]," id. § 416.926a(k)(1)(iv). The regulations specifically provide that "suicidal thoughts or actions ... or refusal to take ... medication" exemplify a child's limited functioning in caring for herself. Id. § 416.926a(k)(3)(iv).

There is substantial evidence in the record that P.K. has suicidal thoughts and inconsistently takes her medication. Nevertheless, in the caring-for-herself domain, the ALJ failed to mention P.K.'s medication compliance and stated only that her suicidal ideation was "being addressed with therapy." The fact that her suicidal ideation is "being addressed" says nothing about the severity of P.K.'s impairment in this domain. It seems obvious that the therapy's progress and success are key to determining whether P.K. is extremely impaired, markedly impaired, or less than markedly impaired. But the ALJ apparently assumed that the mere fact of therapy resolved the issue of suicidal ideation.

Moreover, to support his statement that P.K.'s suicidal ideation is "being addressed in therapy," the ALJ cited teacher questionnaires and the Comprehensive Assessment completed by the Hogares therapist in March 2010. But the teachers' questionnaires are silent about therapy for suicidal ideation, and the ALJ indicated elsewhere in his decision that the Hogares assessment was not a record of therapy.

In short, the ALJ's conclusion that P.K. is less than markedly impaired in caring for herself is devoid of support. Remand is necessary to determine the severity of P.K's impairment in this domain.

C

Knight also argues that the record does not contain substantial evidence for the ALJ's findings that P.K. has less than a marked impairment in three of the remaining domains: acquiring and using information; attending and completing tasks; and health and well-being. We do not need to determine whether the ALJ's findings in these domains are supported by substantial evidence or "overwhelmed by other evidence in the record." Wall, 561 F.3d at 1052 (quotation omitted). On remand, the ALJ will be required to perform a proper credibility determination of the hearing testimony; revisit the evidence regarding P.K.'s abilities to interact with others and to care for herself; and reassess and identify the weight assigned to Dr. Hall's opinions. And under the agency's "whole child approach," the ALJ must consider "the interactive and cumulative effects of the child's impairment(s), including any impairments that are not `severe.'" SSR 09-1p, Title XVI: Determining Childhood Disability Under the Functional Equivalence Rule-The "Whole Child" Approach, 2009 WL 396031, at *3 (S.S.A. Feb. 17, 2009); see also 20 C.F.R. § 416.926a(c). Thus, on remand, the ALJ should reevaluate P.K.'s functioning throughout all domains.

V

The judgment of the district court is REVERSED. We REMAND the case to the district court with directions to remand the matter, in turn, to the agency for further proceedings consistent with this decision.