OPINION

Bernard Laborin appeals the district court’s judgment affirming an administrative law judge’s (ALJ). denial of his applications for disability benefits and supplemental security income under Title II and Title XVI of the Social Security Act, 42 U.S.C. §§ 301-1397mm. The ÁLJ did not credit Laborin’s testimony regarding the intensity, persistence, and limiting effects of his symptoms to the extent that testimony was “inconsistent with the above residual functional capacity assessment [ (RFC) j,” ^[1] This boilerplate statement encourages an inaccurate assessment of a[*1153] claimant’s credibility and also permits determination of RFCs that are inconsistent with truly credible testimony. The approach taken by the ALJ .was inconsistent with the Social Security Act and should not be used in disability decisions. Because the ALJ also did not give clear and convincing reasons for rejecting Laborin’s symptom testimony, and for the reasons, provided in the concurrently filed memorandum, disposition, we reverse and remand. ^[2]

The “RFC is an administrative assessment of the extent to which an individual’s medically determinable impairment(s), -including any related symptoms, such as pain, may cause physical or mental limitations or restrictions that may affect his or. her capacity to do work-related physical and mental activities.” SSR 96-8p, 61 Fed. Reg. 34474, 34475 (July 2, 1996). ^[3] It “is the most [a" claimant] can still do despite [his or her] limitations.” 20 C.F.R. § 416.945(a)(1).

When an individual does not have a severe medical impairment that meets or equals one of the listed impairments, the scope of the RFC plays a crucial role in the ALJ’s determination of whether an individual is disabled and entitled to benefits under the Social Security Act. See generally id. § 416.920(a)(4) (describing the five-step sequential evaluation process); SSR 96-8p, 61 Fed. Reg. 34474 (explaining the RFC), The ALJ assesses a claimant’s RFC between steps three and four of the five-step sequential evaluation process used for disability determinations. See 20 C.F.R. § 416.920(a)(4). The RFC is first used at step four to determine whether an individual can do relevant past work. Id. § 416.920(a)(4)(iv). If the claimant cannot do relevant past work, the RFC is then used again at step five—along with the claimant’s age, education, and work experience—to determine whether the claimant “can make an adjustment to other work.” Id. § 416.92Q(a)(4)(iv)-(v). If the claimant can do other work, the claimant is found to be not disabled. See id. § 416.920(a)(v).

The ALJ assesses, a claimant’s RFC “based on all the relevant evidence in [the] case record.” Id. § 416,945(a)(l)..The ALJ must consider both the medical evidence and “descriptions and observations of [the claimant’s] limitations from [the claimant’s] impairment(s), including limitations that result from [the claimant’s] symptoms, such as pain, provided by” the claimant, family, friends, and .other people. Id. § 416.945(a)(3). The RFC assessment must “[e]ontain a thorough discussion and analysis of the objective medical and other evidence, including the. individual’s complaints of pain and other symptoms and the adjudicator’s personal observations, .if appropriate.” SSR 96-8p, 61 Fed. Reg. at 34478 (emphasis added). In other words, the ALJ must take “the claimant’s subjective experiences of pain” into account when determining the RFC. Garrison v, Colvin, 759 F.3d 995, 1011 (9th Cir. 2014); see also Lingenfelter v. Astrue, 504 F.3d 1028, 1035 (9th Cir. 2007) (“[T]he ALJ failed to provide clear and convincing reasons for finding [the claimant’s] alleged pain and symptoms not credible, and therefore was required to include these[*1154] limitations in his assessment of [the claimant’s] RFC.”). ^[4]

Despite this clear directive from the regulations, Social Security Rulings, and our case law, which all require ALJs to take the claimant’s symptom testimony into account in determining the RFC, ALJs with frequency include the boilerplate language discrediting the claimant’s symptom testimony because it is “inconsistent with” the RFC in their disability determinations. See, e.g., Treichler v. Comm’r of Soc. Sec. Admin., 775 F.3d 1090, 1103 (9th Cir. 2014). By doing so, the ALJ indicates that he or she did not properly “incorporate a claimant’s testimony regarding subjective symptoms and pain into the RFC finding, as [he or she] is required to do.” Trevizo v. Berryhill, 862 F.3d 987, 1000 n.6 (9th Cir. 2017); see also Mascio v. Colvin, 780 F.3d 632, 639 (4th Cir. 2015) (holding that this boilerplate language conflicts with the regulations and rulings). This practice “inverts the responsibility of an ALJ, which is first to determine the medical impairments of a claimant based on the record and the claimant’s credible symptom testimony and only then to determine the claimant’s RFC.” Trevizo, 862 F.3d at 1000 n.6.

Not only does the ALJ err by discrediting symptom testimony to the extent it is inconsistent with an RFC, but the ALJ’s analysis is also illogical. Because the claimant’s symptom testimony must be taken into account when the ALJ assesses the claimant’s RFC, it cannot be discredited because it is inconsistent with that RFC. By discrediting a claimant’s “statements concerning the intensity, persistence and limiting effects of [the claimant’s] symptoms ... to the extent they are inconsistent with the” RFC, the ALJ puts the cart before the horse.

We are not the first circuit to recognize that this boilerplate language is problematic. Both the Seventh and the Fourth Circuits have concluded that it is incorrect as a matter of law. See Filus v. Astrue, 694 F.3d 863, 868 (7th Cir. 2012) (noting that finding a claimant’s symptom testimony is not credible to the extent it is inconsistent with an RFC “puts the cart before the horse, in the sense that the determination of capacity must be based on the evidence, including the claimant’s testimony, rather than forcing the testimony into a foregone conclusion”); Bjornson v. Astrue, 671 F.3d 640, 645 (7th Cir. 2012) (noting that this boilerplate language “gets things backwards”); Mascio, 780 F.3d at 639 (concluding the boilerplate language “ ‘gets things backwards’ by implying ‘that ability to work is determined first and is then used to determine the claimant’s credibility’” (quoting Bjornson, 671 F.3d at 645)). This improper procedure both inverts and subverts the way an RFC must be determined relying on credible evidence, including credible testimony.

Like our sister circuits, we have stated that inclusion of this flawed boilerplate language is not, by itself, reversible error and can be harmless. See Trevizo, 862 F.3d at 1000 n.6; Mascio, 780 F.3d at 639; Filus, 694 F.3d at 868. It does not, however, add anything to the ALJ’s deter[*1155] mination of either the RFC or the claimant’s credibility. We cannot infer from this language “that the ALJ rejected [the claimant’s] testimony to the extent it conflicted with [the] medical evidence.” Treichler, 775 F.3d at 1103. To discredit a claimant’s symptom testimony when the claimant has provided objective medical evidence of the impairments which might reasonably produce the symptoms or pain alleged and there is no evidence of malingering, the ALJ must give “specific, clear, and convincing reasons, for rejecting” the testimony by identifying “which testimony [the ALJ] found not credible” and explaining “which evidence contradicted that testimony.” Brow n-Hunter v. Colvin, 806 F.3d 487, 489, 494 (9th Cir. 2015). This boilerplate language does neither.

As we elaborate in the concurrently filed memorandum disposition, here the ALJ did not give clear and convincing reasons for rejecting Laborin’s testimony regarding the severity of his pain. The ALJ’s inclusion of the boilerplate statement dobs not save her decision.

REVERSED and REMANDED.