*614ORDER
Debbie Eskew seeks review of the denial of her application for disability insurance benefits. She argues that the administrative law judge committed a series of analytical and factual errors while applying the five-step test used to determine eligibility for benefits. Because the ALJ’s assessment of Eskew’s credibility was patently wrong, we reverse and remand for further proceedings.
Eskew maintains that she is disabled by a combination of several impairments, including asthma; abnormal heart rhythms; bronchitis; early cataracts; chronic obstructive pulmonary disease (COPD); depression; diabetes; emphysema; inflammation of the esophagus; pneumonia; and lymph nodes in her lungs. She claims that she became disabled in March 2005, when she was hospitalized after complaining of chest pain, a persistent cough, and difficulty breathing. She was diagnosed with severe COPD, pneumonia, and anxiety. Shortly after this hospitalization, Eskew applied for unemployment benefits, for which her treating physician, Dr. Martin Brauweiler, prepared a note stating that she had been able to work since early March, adding that she had been unable to do work requiring her to lift more than 20 pounds or be around smoke.
In early 2006 Eskew reported to an emergency room complaining of chest pain and told the attending physician that she had not been taking albuterol because her COPD had been “okay.” Her condition improved after she received treatment. A CT scan taken at the hospital revealed a mass on her lung, and she was strongly urged to return for a follow-up in two weeks, but the record shows no such visit.
During an emergency room visit in March 2006, she reported being unable to afford her medications and was directed by her doctor to contact a local free clinic. She apparently did not do so. Three months later she reported to an emergency room doctor that the free clinic would not help her because she was still on Medicaid, and she could not find a local provider who would accept Medicaid. Her Medicaid coverage expired one week later, and since then she has received treatment from the free clinic.
In August 2006 Dr. Virgilio Pilapil, a state-agency medical consultant, reviewed Eskew’s medical records. He concluded that she could lift 20 pounds occasionally and 10 pounds frequently, and could stand for about 6 hours in an 8-hour workday or sit for the same amount of time.
At a hearing before an ALJ to determine her eligibility for benefits, Eskew was asked about her daily activities and the reasons for several hospital visits. She testified that, even with her medication, she still became winded walking short distances, had difficulty breathing on humid days, and was unable to do laundry. She reported being unable to stand in place for longer than ten minutes. A vocational expert testified that Eskew could no longer perform her prior relevant work and, that if a person with her impairments were limited to light work that did not involve exposure to dust and smoke or more than occasional climbing of ramps and stairs, such a person could work as a cashier, receptionist, courier, or general office clerk.
The ALJ concluded that Eskew was not disabled under the Social Security Act. Applying the five-step process that governs review of disability determination, 20 C.F.R. § 404.1520(a), the ALJ found that Eskew had not engaged in gainful employment since the alleged onset date (step 1); that she had a severe impairment (COPD) (step 2); that she did not meet a listed impairment under Listing 3.02A (step 8); that she could not perform her prior relevant work but that she had the residual *615functional capacity (RFC) to perform light duty work (step 4); and that her RFC would allow her to perform other jobs existing in significant numbers (step 5). In concluding that she could perform light duty work, the ALJ stated that he found her testimony “not entirely credible,” noting gaps in her pursuit of treatment and inconsistencies in her testimony regarding her ability to afford treatment. He noted that the medical evidence in the record belied the degree of seriousness with which she described her level of impairment. The Appeals Council denied her request for review. The magistrate judge, proceeding with the parties’ consent, later granted summary judgment for the agency.
On appeal, Eskew faults the ALJ for (1) failing to consider all of her impairments at step 2; (2) ignoring medical records which showed that her condition met the requirements of Listing 3.02A — thereby qualifying her automatically for disability benefits; (3) overlooking whether her impairments, acting in combination, constituted the functional equivalent of Listing 3.02A; (4) improperly discrediting her testimony; (5) finding her capable of light work at step 4 without addressing countervailing evidence; and (6) posing a hypothetical to the vocational expert at step 5 that failed to accurately describe her impairments.
We review the ALJ’s decision directly. Simila v. Astrue, 573 F.3d 503, 513 (7th Cir.2009). We do not displace the ALJ’s judgment by reconsidering facts or evidence, or by making independent credibility determinations. Id. Instead, we review the ALJ’s decision to determine whether it is supported by substantial evidence. Myles v. Astrue, 582 F.3d 672, 676 (7th Cir.2009).
Eskew first argues that the ALJ erred by failing to assess the combined severity of her impairments at step 2, and she insists that this oversight tainted his analysis at later steps. Step 2 requires the ALJ to determine whether the claimant has a “severe medically determinable physical or mental impairment ... or a combination of impairments that is severe....” 20 C.F.R. § 404.1520(a)(4)(ii). But step 2 is also a threshold step and “[a]s long as the ALJ determines that the claimant has one severe impairment, the ALJ will proceed to the remaining steps of the evaluative process.” Castile v. Astrue, 617 F.3d 923, 926-27 (7th Cir.2010). The ALJ here determined that Eskew’s COPD was a severe impairment, and proceeded accordingly to the remaining steps of the analysis.
Eskew next argues that the ALJ erred at step 3 by finding that her condition did not meet Listing 3.02A. Under that listing, if a person suffering from COPD has a forced expiratory volume in one second (FEV1) at a particular value, she automatically qualifies for benefits. 20 C.F.R. Pt. 404, Subpt. P, App. 1. § 3.02(A). FEV1 is measured through use of an instrument called a “spirograph” both before and after the patient inhales a “bronchodilator” medication. See § 3.00(E). Both measurements produce FEV1 values, but only the highest post-bronchodilator result is used to assess the severity of the respiratory impairment. See id. For a person of Eskew’s height, an FEV1 value at or below 1.25 qualifies for benefits. See § 3.02(A) tbl.l. According to Eskew, an August 2006 spirograph in which she produced an FEV1 of 1.15 should have satisfied the Listing, but was ignored by the ALJ. But she mischaracterizes this spiro-graph result. The 1.15 reading was an inapplicable prebronchodilator result; her highest post-bronchodialator result from the August 2006 spirograph was 1.85. The ALJ correctly ignored the pre-bronchodi-lator result and properly concluded that *616the 1.85 FEV1 value was too high to meet Listing 3.02A.
Eskew insists, in the alternative, that even if her impairments did not satisfy Listing 3.02A, the ALJ failed to explain why they did not functionally equal its requirements. 20 C.F.R. § 414.1520(a)(4)(iii). But the burden of proof at step 3 rests with the claimant, and the ALJ has no duty to analyze equivalence when the claimant, like Eskew, presents no substantial evidence of it. Scheck v. Barnhart, 357 F.3d 697, 700-01 (7th Cir.2004); Clifford v. Apfel, 227 F.3d 863, 868 (7th Cir.2000).
Eskew next argues that the ALJ improperly discounted her credibility by drawing inappropriate inferences from her gaps in treatment, disregarding her explanation that she could not afford prescribed medication, and speculating that the absence of work restrictions in her medical record undermined her claims of disabling impairments.
We are troubled by the ALJ’s failure to build a “logical bridge” between the record and his credibility determination. Castile, 617 F.3d at 929; Simila, 573 F.3d at 517. The absence of major work restrictions in Eskew’s medical records does not illuminate the question of her credibility — she was after all unemployed throughout the time in question. Further, the ALJ did not identify the “gaps” in Eskew’s treatment history he alludes to beyond one instance in which she apparently ignored a doctor’s recommendation that she make a follow-up visit. He also failed to address Eskew’s explanation that the gaps in treatment stemmed from her difficulty finding medical providers who accepted Medicare. See SSR 96-07p,1996 WL 374186, at *7 (July 2,1996); see also Myles, 582 F.3d at 677; Craft v. Astrue, 539 F.3d 668, 679 (7th Cir.2008). And he summarily dismissed Eskew’s explanation for not taking prescribed medication simply by noting her ability to buy cigarettes during that time -even though the record contains no information about either the price of her medication or the cost of her cigarette habit.
Finally, the ALJ assesses Eskew as “not entirely credible” — a phrase we have derided as “meaningless boilerplate.” Parker v. Astrue, 597 F.3d 920, 922 (7th Cir.2010). As we have explained, the phrase fails to specify which statements are or are not credible. Martinez v. Astrue, 630 F.3d 693, 695 (7th Cir.2011). The ALJ’s wording leaves us with no basis to review whether his conclusion is supported by substantial evidence or to understand how his determination influenced his RFC assessment.
Although the question is close, the ALJ’s flawed credibility assessment cannot be brushed aside merely as harmless. An erroneous credibility finding requires remand unless the claimant’s testimony is incredible on its face or the ALJ explains in his decision how he would have reached the same conclusion even if he had found the claimant credible. See Allord v. Barnhart, 455 F.3d 818, 821 (7th Cir.2006). Here, the ALJ did not provide an alternate justification for his decision, and Eskew’s account of her impairments is not so contradicted by the medical record that it is wholly unbelievable; Dr. Brauweiler’s assessment covers only the first months after the onset date and Dr. Pilapil’s “check-box form” is the sort of medical evaluation this court has discounted in the past. See Larson v. Astrue, 615 F.3d 744, 751 (7th Cir.2010).
We need not address Eskew’s remaining arguments; on remand the RFC determination and questions to the vocational expert may be altered after the ALJ develops a properly supported credibility detennination. The judgment is therefore VACATED and the case is REMANDED *617to the Social Security Administration for further proceedings.