Kalulu v. Bondi

FOR PUBLICATION UNITED STATES COURT OF APPEALS FOR THE NINTH CIRCUIT MILLY KALULU, No. 21-895 Petitioner, Agency No. A213-592-589 v. ORDER AND PAMELA BONDI*, Attorney General, AMENDED OPINION Respondent. On Petition for Review of an Order of the Board of Immigration Appeals Argued and Submitted April 20, 2023 San Francisco, California Filed March 11, 2024 Amended February 13, 2025 Before: Lawrence VanDyke and Gabriel Sanchez, Circuit Judges, and Stephen J. Murphy, III, District Judge. ** * We have substituted Attorney General Pamela Bondi as defendant- appellee pursuant to Federal Rule of Appellate Procedure 43(c). ** The Honorable Stephen J. Murphy, III, United States District Judge for the U.S. District Court for the Eastern District of Michigan, sitting by designation. 2 KALULU V. BONDI Order; Opinion by Judge VanDyke; Partial Concurrence and Partial Dissent by Judge Sanchez SUMMARY ** Immigration The panel filed an order (1) amending the majority opinion and dissent filed on March 11, 2024; (2) denying the petition for rehearing en banc; and (3) indicating that no further petitions for rehearing shall be filed. In the amended opinion, the panel granted Milly Kalulu’s petition for review of the Board of Immigration Appeals’ decision upholding the denial of asylum and related relief, and remanded, holding that although substantial evidence supported the agency’s adverse credibility determination, the agency did not properly evaluate whether Kalulu’s supporting evidence independently supported her claims of past persecution in her native Zambia on account of her sexual orientation. The panel concluded that while some of the reasons the agency relied upon did not support its credibility finding, the administrative record as a whole did not compel a conclusion different than the agency’s, even after the record was stripped of any of the agency’s erroneous findings. Specifically, at least five of the factual bases underlying the agency’s adverse credibility determination ** This summary constitutes no part of the opinion of the court. It has been prepared by court staff for the convenience of the reader.  KALULU V. BONDI 3 were supported by the record, including four identified inconsistencies, as well as the IJ’s demeanor finding. The panel concluded that the IJ failed to properly consider and evaluate the evidentiary weight of multiple documents Kalulu offered into the record independent of her noncredible testimony, and the BIA made clear factual errors when it reviewed those documents. The panel therefore remanded for the IJ or BIA to consider whether those documents, when properly read, independently proved Kalulu’s past persecution claim. Concurring in part and dissenting in part, Judge Sanchez agreed with the majority that remand was required because the agency failed to consider whether Kalulu’s supporting evidence independently proved her claims. However, Judge Sanchez wrote that because the bulk of the agency’s credibility findings were based on significant errors, the REAL ID Act, principles of administrative law, and precedent require remand to the BIA to determine whether the few remaining factors are sufficient in light of the totality of the circumstances. COUNSEL Claire Weintraub (argued) and Natalie Kaliss (argued), Certified Law Students; Amalia Wille (argued) and Judah Lakin, Supervising Attorneys; Erwin Chemerinsky, University of California, Berkeley School of Law, Berkeley, California; Gia L. Cincone, Kilpatrick Townsend & Stockton LLP, San Francisco, California; for Petitioner. Robert D. Tennyson Jr. (argued), Trial Attorney, and Paul Fiorino, Senior Litigation Counsel, Office of Immigration 4 KALULU V. BONDI Litigation; Brian Boynton, Principal Deputy Assistant Attorney General, Civil Division; United States Department of Justice, Washington, D.C.; for Respondent. Kristin Macleod-Ball, National Immigration Litigation Alliance, Brookline, Massachusetts; Glenda M. Aldana Madrid, Matt Adams, Leila Kang, and Aaron Korthuis, Northwest Immigrant Rights Project, Seattle, Washington; for Amici Curiae Northwest Immigrant Rights Project and National Immigration Litigation Alliance. Andrew R. Davies and Mark L. Hanin, Wilmer Cutler Pickering Hale and Dorr LLP, New York, New York; Sonia Sujanani, Wilmer Cutler Pickering Hale and Dorr LLP, Boston, Massachusetts; Mary Georgevich, National Immigrant Justice Center, Chicago, Illinois; Bridget Crawford, Immigration Equality, Brooklyn, New York; Liz Bradley and Kari Hong, Florence Immigrant & Refugee Rights Project, Tucson, Arizona; for Amici Curiae National Immigrant Justice Center, Immigration Equality, and The Florence Immigrant and Refugee Rights Project. ORDER The majority opinion and dissent filed on March 11, 2024, are hereby amended. The amended opinion and amended dissent will be filed concurrently with this order. The full court has been advised of the petition for rehearing en banc and no judge has requested a vote on whether to rehear the matter en banc. Fed. R. App. P. 40. Petitioner’s petition for rehearing en banc (Dkt. No. 48) is thus DENIED. No further petitions for rehearing shall be filed.  KALULU V. BONDI 5 OPINION VANDYKE, Circuit Judge: Milly Kalulu petitions for review of a Board of Immigration Appeals (BIA) decision dismissing her appeal of a removal order. We have jurisdiction under 8 U.S.C. § 1252 and grant her petition. Even though substantial evidence supports the agency’s adverse credibility determination, the agency did not properly evaluate documents Kalulu introduced into the record to support her claims of past persecution in her native Zambia on account of her sexual orientation. We grant the petition and remand this case with instructions for the agency to consider whether those documents, when properly read, independently prove those claims. I. BACKGROUND 1 Kalulu is a native of Zambia who identifies as a lesbian. Homosexual activity is illegal in Zambia, and Kalulu says she began to experience persecution on account of her sexual orientation after brothers of her girlfriend discovered their relationship in 2019. She recounts two episodes when those brothers attacked her in her hometown, and then another episode when they harried her from a restaurant in Zambia’s capital city after she fled there. Shortly after the last alleged attack, Kalulu entered the United States on a tourist visa to attend a world scouting jamboree in West Virginia with her Zambian girl scout 1 Some facts provided in this section are based on parts of Kalulu’s testimony the IJ found not to be credible. The court presents them here only as background and does not rely on those facts in its analysis. 6 KALULU V. BONDI troop. Her visa permitted her to remain in the United States for up to six months at a time, renewable for up to three years, and Kalulu chose to reside in California with her naturalized aunt for approximately five months after the jamboree ended. She then took an extended weekend trip to Mexico so that on reentry she could reset the six-month clock on her stay in the United States. Kalulu legally reentered without difficulty at a California port of entry. But when she then tried to extend her visa in person at the port, she was directed to a building where her wallet and phone were inspected. That inspection uncovered a California public health benefits card and WhatsApp messages describing her paid babysitting work in California. Her tourist visa did not permit receipt of those benefits or earned income, so border officials cancelled her visa and placed her in removal proceedings. Kalulu did not mention any past persecution or fear of future persecution to border officials during her interactions with them. By the time Kalulu applied for asylum, withholding of removal, and CAT relief about three months later, however, she claimed to have suffered past persecution (and to fear future persecution) on the basis of her sexual orientation. She offered various documents into the record before her hearing to support her claims of past persecution, including purported declarations from eyewitnesses to the three alleged attacks described above and a purported medical report describing injuries from the second attack. To support her claims of a well-founded fear of future persecution, Kalulu offered news articles and a State Department report on Zambia that described ongoing persecution of homosexuals there.  KALULU V. BONDI 7 At her removal hearing, Kalulu initially offered her own testimony in support of her asylum application. But when the IJ began to question her about her story, the IJ found some of her testimony evasive and contradictory. After detailed examination of the perceived inconsistencies and evasiveness, the IJ determined that they rendered her testimony not credible. And because the IJ did not believe that the testimony from Kalulu’s other witnesses or supporting documents rehabilitated her credibility, independently established her claims of past persecution, or demonstrated that she was more likely than not to be tortured if she returned to Zambia, the IJ entered an order denying asylum, withholding of removal, and CAT relief. On appeal, the BIA found no clear error in the IJ’s adverse credibility determination or in the IJ’s determination that Kalulu’s supporting documents did not rehabilitate her credibility or independently establish her claims of past persecution. The BIA explained: The affidavits from the respondent’s cousin, neighbor, and friend are not signed or sworn …, which undermines their evidentiary value. The respondent’s cousin’s letter does not mention any reason for the June 2019 attack on the respondent. The neighbor’s statement also does not mention any underlying reason for the attack. The friend’s statement indicates that the respondent would be in danger if she returned 8 KALULU V. BONDI to Zambia but does not mention whether she was aware of the respondent’s sexuality. The BIA likewise found no clear error in the IJ’s determination that Kalulu’s medical report failed to rehabilitate her testimony or support even the second purported attack because the report (1) had been signed by a police officer even though Kalulu testified that she did not report that attack to police and (2) omitted injuries Kalulu had testified to during her hearing. The BIA thus concluded that Kalulu could not meet her burden to establish eligibility for asylum. A fortiori, it reasoned, she could not meet the higher burden of proof of persecution to establish eligibility for withholding of removal. The BIA similarly upheld the IJ’s denial of CAT relief because Kalulu had offered only her noncredible testimony as to the likelihood that the Zambian government would persecute her or acquiesce to her persecution by a private party. II. STANDARD OF REVIEW This court’s review of an agency order denying asylum, withholding of removal, and CAT relief “is limited to the BIA’s decision, except to the extent that the IJ’s opinion is expressly adopted.” Singh v. Garland, 57 F.4th 643, 651 (9th Cir. 2022) (internal quotation marks omitted). We look to the IJ’s decision “as a guide to what lay behind the BIA’s conclusion” when “the BIA reviewed the IJ’s credibility- based decision for clear error and relied upon the IJ’s opinion as a statement of reasons but did not merely provide a boilerplate opinion.” Dong v. Garland, 50 F.4th 1291, 1296 (9th Cir. 2022) (cleaned up).  KALULU V. BONDI 9 The agency’s factual findings are reviewed for substantial evidence, which exists when the agency’s conclusions “are supported by reasonable, substantial, and probative evidence in the record.” Gutierrez-Alm v. Garland, 62 F.4th 1186, 1194 (9th Cir. 2023) (cleaned up). Under that extremely deferential standard of review, this court may not independently weigh the evidence and reverse the agency unless “the evidence compels a conclusion contrary to the BIA’s.” Umana-Escobar v. Garland, 62 F.4th 1223, 1228 (9th Cir. 2023). As this court oftentimes phrases the same point in the inverse, we must accept agency factual findings “as conclusive unless any reasonable adjudicator would be compelled to conclude to the contrary.” Antonio v. Garland, 58 F.4th 1067, 1073 (9th Cir. 2023) (cleaned up); 8 U.S.C. § 1252(b)(4)(B). The agency may deny asylum, withholding of removal, and CAT relief because it deems an applicant’s testimony, under the totality of the circumstances, not to be credible, and the lack of credible testimony renders the applicant unable to meet her burden. Iman v. Barr, 972 F.3d 1058, 1064–65 (9th Cir. 2020). Such an adverse credibility determination may be based on inconsistencies in an applicant’s testimony even if no uncovered inconsistency goes to the heart of the applicant’s claim. 8 U.S.C. § 1158(b)(1)(B)(iii); Lalayan v. Garland, 4 F.4th 822, 835 (9th Cir. 2021). It also may be based on a finding that a petitioner’s demeanor undermined her credibility. 8 U.S.C. § 1158(b)(1)(B)(iii); Munyuh v. Garland, 11 F.4th 750, 758 (9th Cir. 2021). III. DISCUSSION Even though substantial evidence supports the agency’s adverse credibility determination in this case, the agency 10 KALULU V. BONDI misread some of Kalulu’s supporting documents. Because that mistake prevented the agency from properly evaluating whether the documents independently prove Kalulu’s claim of past persecution, this court remands with instructions for the agency to reexamine their evidentiary value when properly characterized. a. Substantial Evidence Supports the Agency’s Adverse Credibility Determination. The agency based its adverse credibility determination on (1) twelve inconsistency and implausibility findings and (2) Kalulu’s demeanor during her hearing. The adverse credibility determination is supported by substantial evidence because at least four of those inconsistencies, as well as the IJ’s demeanor finding, are supported by the record. At the threshold, the parties disagree about which of the IJ’s findings related to the adverse credibility determination the BIA adopted. Kalulu argues the BIA adopted only some of the IJ’s findings because it did not affirmatively and expressly adopt each one or invoke Matter of Burbano, 20 I. & N. Dec. 872 (BIA 1994), and that those findings not expressly adopted by the BIA are not properly before this court. The government, on the other hand, argues that all the IJ’s findings are properly before the court. Kalulu’s argument fails to appreciate that this court, in agreement with every other circuit, considers an adverse credibility determination to be a factual finding informed by other factual findings made by the IJ (such as inconsistencies and demeanor). Mukulumbutu v. Barr, 977 F.3d 924, 925 (9th Cir. 2020). As such, the IJ’s adverse credibility determination, together with all the findings that fed into that determination, are subject to clear error review by the BIA.  KALULU V. BONDI 11 See 8 C.F.R. § 1003.1(d)(3)(i). Reviewing for clear error, the BIA may not overturn an IJ’s factual finding sub silentio. 2 This court must review the agency’s conclusions in this case for substantial evidence, which means that the court must review the agency’s decision against the whole record and “must accept ‘administrative findings,’” including the adverse credibility determination, “as ‘conclusive unless any reasonable adjudicator would be compelled to conclude to the contrary.’” Garland v. Ming Dai, 141 S. Ct. 1669, 1677 (2021) (quoting 8 U.S.C. § 1252(b)(4)(B)). The two pre- REAL ID Act cases Kalulu relies on to argue for less deference, believing the IJ questioned her too intensively, do not suggest otherwise, not least of all because this court in both cases ultimately employed the same standard of review we apply here: substantial evidence. See Arulampalam v. Ashcroft, 353 F.3d 679, 688 (9th Cir. 2003); Garrovillas v. I.N.S., 156 F.3d 1010, 1015–16 (9th Cir. 1998). In any event, both those cases are dissimilar to this one. There, IJs bullied petitioners during their removal hearings with pervasive “haranguing,” “derisive innuendos,” and “inexplicable outbursts.” Garrovillas, 156 F.3d at 1014–15; see Arulampalam, 353 F.3d at 682, 687. Here, Kalulu asserts only that the IJ (1) asked lots of clarifying questions and (2) grew frustrated when she concluded Kalulu was trying to evade those questions. Neither warrants or even permits departure from the substantial evidence standard of 2 Kalulu relatedly argues that some of the agency’s findings that inform the IJ’s adverse credibility determination should not be reviewed deferentially by our court. Again, the law unambiguously requires us to review the agency’s factual findings deferentially. 8 U.S.C. § 1252(b)(4)(B); see Mukulumbutu, 977 F.3d at 926–27. 12 KALULU V. BONDI review. An IJ who perceives inconsistencies in a petitioner’s testimony may certainly ask clarifying questions so she has an opportunity to reconcile the inconsistencies. And this court has made clear that while an IJ’s expression of frustration during a hearing might not be optimal, it does not taint her findings such that substantial evidence review no longer holds. Gonzalez-Veliz v. Garland, 996 F.3d 942, 950 (9th Cir. 2021). Moving onto the findings themselves, the agency offered thirteen factual reasons for its adverse credibility determination. This court does not draw a bright line as to what number or percentage of the underlying factual reasons identified by the agency must be supported by the record for the agency’s ultimate determination to survive substantial evidence review. See Barseghyan v. Garland, 39 F.4th 1138, 1142 (9th Cir. 2022). Instead, this court examines the totality of the record, Alam v. Garland, 11 F.4th 1133, 1137 (9th Cir. 2021) (en banc), considers all the factually supported reasons for the agency’s determination, and defers to that conclusion unless no reasonable adjudicator considering the factual record as a whole could come to the same conclusion that the agency did, Antonio, 58 F.4th at 1072–73. Here, while some of the reasons relied on by the agency do not support its ultimate credibility finding, once those unsupported reasons are disregarded, substantial evidence remains to support the agency’s determination that Kalulu was not credible. Put inversely, the administrative record as a whole does not compel a conclusion different than the agency’s, even after the record is stripped of any of the agency’s erroneous findings. Here, at least five of the factual bases underlying the agency’s adverse credibility determination are supported by the record.  KALULU V. BONDI 13 First, the record supports the agency’s finding that Kalulu was inconsistent about when she made plans to remain in the United States after the jamboree rather than return to Zambia. She initially testified that she did not make the plan while still in Zambia. But her testimony shortly thereafter contradicted that timeline, indicating that she had made the plan “[w]hen [she] was in Zambia.” Kalulu did not take advantage of opportunities the IJ afforded her to clarify the apparent inconsistency, and it was not unreasonable for the IJ to find that she could not have both made and not made the plans while in Zambia. 3 Second, the record supports the agency’s finding that Kalulu was inconsistent about her alleged fear of future persecution if she returned to Zambia. She did not disclose such a fear during her first border interview. Yet she then claimed in her asylum application and hearing testimony only that she was afraid because of her sexual orientation. Kalulu does not endeavor to argue that these changing claims were in any way consistent. And there is nothing in the record that compels a contrary conclusion. 4 3 Kalulu argues that “the IJ never notified [her] or her counsel that she perceived as inconsistent [her] testimony about her plans to stay in the United States.” While the IJ never uttered the word “inconsistent” during the lengthy colloquy with Kalulu on this point, the IJ repeatedly advised her that the IJ was trying to reconcile two apparently irreconcilable statements. That sufficed to meet the IJ’s obligation to give Kalulu notice of the substance of the perceived inconsistency and an opportunity to explain it. See Lizhi Qiu v. Barr, 944 F3d 837, 846 (9th Cir. 2019); Rizk v. Holder, 629 F.3d 1083, 1088 (9th Cir. 2011). 4 Although Kalulu does not argue that her varying descriptions of whether she was afraid and why she was afraid are consistent, she does claim she failed to disclose her fear of repatriation at the border because 14 KALULU V. BONDI Third, the record supports the agency’s finding that Kalulu was inconsistent about whether she planned to attend college in California. Her border incident report indicates she told a border officer that she intended to enroll at the California State University campus in San Marcos and that her California-based aunt would pay her tuition there. she was scared. But the agency was not required to credit Kalulu’s proffered explanation and disregard or discount the inconsistency. Li v. Garland, 13 F.4th 954, 960–61 (9th Cir. 2021). Kalulu relies on two cases to argue to the contrary, but neither supports her position. In Mousa v. Mukasey, a female petitioner said she did not declare her fear of rape upon repatriation to a male border official because she was scared and ashamed to disclose it. 530 F.3d 1025, 1027–29 & n.3 (9th Cir. 2008). The agency relied on this lack of disclosure as the “primary reason” for its adverse credibility determination. This court reversed, holding that nondisclosure of rape is sui generis, not “a bellwether of truth,” and unreliable as a primary reason for an adverse credibility determination. Id. at 1027. Here, Kalulu has not articulated a fear of rape or other sexual assault, and her failure to disclose fear at the border is just one of multiple independent reasons the IJ offered for the adverse credibility determination. And in any event the Mousa decision did not require the agency to entirely disregard the inconsistency arising from the petitioner’s failure to disclose. Kalulu’s second case, Singh v. I.N.S., did not even involve a petitioner’s claimed fear of border officials. 292 F.3d 1017 (9th Cir. 2002). Instead, a petitioner who spoke Punjabi and only “a little” Hindi needed to communicate with an American border interviewer who only spoke English through a Hindi-English translator who could not speak Punjabi. Id. at 1022–23. His interview report indicated he did not disclose fear of persecution upon repatriation, but also strongly suggested he had not understood the border official’s question about fear as it went through two layers of translation. Id. at 1022. No similar linguistic barrier was present here. And like the Mousa decision above, the Singh decision did not ultimately require the agency to entirely disregard the failure to disclose during its credibility analysis. Id.  KALULU V. BONDI 15 During her hearing, though, Kalulu insisted she had never actually told the border official she planned to attend college. When asked to explain why the border report indicated she had, she said the border official must have misunderstood her. But the IJ was not required to accept that proffered explanation, see Rivera v. Mukasey, 508 F.3d 1271, 1275 (9th Cir. 2007), and the record does not compel a conclusion that Kalulu’s testimony about her college plans was consistent with the border incident report. Fourth, the record supports the agency’s finding that Kalulu was inconsistent about her health condition before she was diagnosed with HIV in 2020. During her hearing, Kalulu testified that she had not felt or been sick in any way prior to that diagnosis. But when the IJ later asked Kalulu about why she had obtained a state public health benefits card in 2019, Kalulu testified that it was because “I wasn’t feeling well with my health…. I had health problems, so that’s why [medical clinic staff] advised me to have the, the insurance.” 5 Kalulu did not take advantage of an opportunity the IJ gave her to explain this inconsistency, and the record does not compel a conclusion that she was consistent when she said, on the one hand, that she was not sick before 2020 and, on the other hand, that she was sick in 2019. Fifth, the record supports the agency’s finding that Kalulu’s demeanor during her removal hearing further undermined the credibility of her testimony because of her evasiveness and non-responsiveness while answering 5 The IJ misquotes Kalulu’s testimony here as “I had heart problems.” But the inconsistency remains because the record, when properly read, shows that Kalulu testified both that she had and did not have health problems before her HIV diagnosis. 16 KALULU V. BONDI several of the IJ’s questions. A short excerpt from the lengthy verbal exchange between the IJ and Kalulu about when Kalulu decided and planned to stay in the United States rather than return to Zambia is illustrative: JUDGE TO MS. KALULU So when did you make the arrangements with your aunt to leave West Virginia and go to California? MS. KALULU TO JUDGE I was talking to her a while on the phone. JUDGE TO MS. KALULU Okay. But that still doesn’t answer my question. When did the two of you decide that you would not go back to Zambia and that the plan was for you [to] come to California to be with her? MS. KALULU TO JUDGE When they just give [sic] me my visa, I knew I wasn’t going to come back to, to go back to Zambia? JUDGE TO MS. KALULU Okay. But did your aunt know that as well? MS. KALULU TO JUDGE No, that’s right. This back and forth continues for another three pages of the hearing transcript without a clear answer to the IJ’s question even though Kalulu twice confirmed that she understood what the IJ was asking her. It only ends when Kalulu, after taking “a long pause,” finally answers the question. The  KALULU V. BONDI 17 record does not compel the conclusion that Kalulu was not evasive, particularly given that this court accords demeanor findings special deference because the IJ uniquely observes demeanor first-hand while this court must try to extract it from “cold records.” Dong, 50 F.4th at 1298; accord Manes v. Sessions, 875 F.3d 1261, 1263–64 (9th Cir. 2017) (per curiam). Kalulu argues that even if the agency was right about those inconsistencies and her demeanor, those facts cannot justify an adverse credibility determination because they are counterbalanced by other consistencies in her testimony and statements. But this court repeatedly has emphasized that an adverse credibility determination is not a balancing exercise, where inconsistent or untrue testimony is weighted against the other testimony in the record that has not been shown to be inconsistent or untrue. A petitioner might (at least appear to) be consistent in much of her testimony, including all aspects of that testimony that go to the heart of her claim. But if the record demonstrates that some of her testimony is inconsistent, even if that testimony is peripheral to the heart of her claim, an IJ is permitted to make an adverse credibility determination. See Lalayan, 4 F.4th at 835. 6 The whole 6 Kalulu correctly notes in her opening brief that this court has held this enormous discretion is not entirely unbounded: “‘An utterly trivial inconsistency … will not by itself form a sufficient basis for an adverse credibility determination.’” And she characterizes many inconsistencies the agency found as being utterly trivial, apparently interpreting this court’s use of the term to be roughly synonymous with “insignificant in comparison to counterevidence.” But Kalulu in her brief misuses ellipses to hide key words from the quote in Shrestha: “[A]n utterly trivial inconsistency, such as a typographical error, will not by itself form a sufficient basis for an adverse credibility 18 KALULU V. BONDI point of an adverse credibility determination is to conclude that a person is not generally credible because of some observed indicia of lack of truthfulness, which then permits the factfinder to disregard the person’s other testimony about which there is no evidence of untruthfulness. Because of that, once a person has sufficiently demonstrated she is not credible in part of her testimony, no amount of ostensibly credible testimony elsewhere can rehabilitate her. Thus, the INA does not require an IJ to gather all credible and noncredible assertions in the record and then somehow expressly weigh them against each other to determine whether a petitioner is more credible than not. Instead, in making an adverse credibility determination, the IJ need only discuss inconsistencies in a petitioner’s testimony (or between that testimony and other evidence in the record). 7 determination.” Shrestha v. Holder, 590 F.3d 1034, 1043 (9th Cir. 2010) (emphasis added). In the fourteen years since this court decided Shrestha, it has resisted characterizing as “utterly trivial” an inconsistency arising from anything other than a typographical error or a minor inconsistency about a date lacking any nexus to the petitioner’s claim. Rizk, 629 F.3d at 1088. None of the inconsistencies supported by the record that the agency relied on here can naturally be read to be akin to a third-party typographical error or a minor difference in an immaterial date. They are not “utterly trivial.” 7 Kalulu is correct that the agency is required to consider all evidence in the record as a whole. So the agency may not rely on a supposed inconsistency when other evidence in the record shows there is actually no inconsistency. But beyond that, our court has made clear that the obligation to consider the entire record is typically satisfied when, as here, the IJ makes a “general statement that the agency considered all the evidence before it.” Cole v. Holder, 659 F.3d 762, 771–72 (9th Cir. 2011) (cleaned up). Absent a reason to think otherwise, we do not assume the agency failed to comply with its legal obligations. B.R. v. Garland, 26 F.4th 827, 836 (9th Cir. 2022). And as explained, here the  KALULU V. BONDI 19 Lastly, Kalulu argues that even if the inconsistencies are supported by the record and cannot be counterbalanced by consistencies in the record, she still could have explained them if the IJ had only given her a meaningful opportunity to do so. But the IJ repeatedly gave her such opportunities, so much so that Kalulu elsewhere in her briefing characterizes the IJ as badgering her by asking too many questions to clarify perceived inconsistencies in her testimony. As discussed above, the IJ engaged in frequent— and often quite lengthy—colloquies with Kalulu to give her multiple chances to explain an inconsistency the IJ expressly identified for her. Indeed, Kalulu’s own attorney, also recognizing the inconsistencies, at one point attempted to interject herself into at least one of those colloquies to help Kalulu try to resolve whether she decided with her aunt to stay in the United States while she still was in Zambia. At that point, the IJ, while not allowing Kalulu’s attorney to testify on behalf of Kalulu, did expressly allow the attorney “to ask [Kalulu] questions … to clarify” the inconsistency. And the IJ again at the end of the hearing gave Kalulu’s attorney another opportunity to ask any questions she wanted on redirect to try to resolve any inconsistent testimony—an opportunity the attorney declined to take. b. The Dissent’s Reasoning Largely Tracks the Unsuccessful Arguments Made by Kalulu and is Therefore Unconvincing for Many of the Same Reasons. agency’s lack of discussion of other “consistent” testimony that is irrelevant to its adverse credibility determination provides no reason to think that the agency failed to consider that testimony or otherwise failed to review the entire record. 20 KALULU V. BONDI Looking to the inconsistencies relied on by the agency, and focusing its attention on other perceived consistencies in Petitioner’s unrelated testimony, the dissent suggests that the appropriate course of action is to remand the agency’s adverse credibility determination for reconsideration in light of only those factors that are supported by substantial evidence. But it is important to recognize at the outset that, with one minor exception, the dissent does not disagree with the majority on the decisive issue that many of the agency’s findings regarding credibility are supported by the record. 8 Nor does it contend that the agency would be unjustified in relying on those findings to make an adverse credibility determination. Indeed, there appears to be complete agreement that these facts could constitute substantial evidence. Yet despite this critical—and in our view dispositive—agreement, the dissent nevertheless advocates for remand because it repeats many of the same errors made by the Petitioner and rejected above by this court. First, the dissent asserts that in evaluating credibility, the agency must “consider the totality of the circumstances” and all “relevant factors.” That much is undoubtedly true, which is indeed what the agency here did. But the dissent missteps, 8 The dissent disagrees as to whether there was substantial evidence to support the finding of an inconsistency in Petitioner’s testimony relating to her health issues. While the dissent implies that her testimony on this one issue could be read as consistent, the agency as the fact-finder was not required to read it as such. And ultimately, our narrow disagreement on just this one basis for the agency’s adverse credibility determination is inconsequential. The dissent agrees that there remain at least four independent bases, which together constitute substantial evidence, supporting the agency’s credibility determination. We don’t read anything in the dissent as concluding that our minor disagreement about the agency’s conclusion on the health issues alone controls anything in this case.  KALULU V. BONDI 21 contending that because other parts of the record contain unrelated testimony that has not been shown to be inconsistent, those parts should be re-weighed against Petitioner’s inconsistencies to reevaluate her credibility. Again, that is not how an adverse credibility determination works. And that is certainly not how we deferentially review such a determination. The agency’s obligation to examine the record as a whole in making its credibility determination does not carry with it an obligation to weigh or counterbalance all of a petitioner’s consistent statements against all of her inconsistent statements. Instead, if there is sufficient inconsistent testimony in the record to constitute “substantial evidence” supporting the conclusion that a petitioner is not credible, those statements may be used by the agency fact- finder to make an adverse credibility determination. Such a factual determination that the petitioner is not credible then properly supports the agency’s discounting of all the testimony of a petitioner—regardless of whether the petitioner’s other testimony may appear credible or not. Put differently, once the agency catches a petitioner in some lies, the agency does not have to weigh those lies against all the other testimony where the agency couldn’t tell if the petitioner was lying. Of course, in making an adverse credibility determination the agency must consider the “record as a whole” and the “totality of the evidence” in the sense that the agency cannot disregard other evidence in the record directly showing that the petitioner was not, in fact, inconsistent or evasive about the issue or issues that formed the basis for the agency’s credibility finding. But the idea that a petitioner’s other, unrelated testimony can somehow outweigh or overcome the fact that she elsewhere lied to the agency has no basis in either logic or our precedent. Nor is 22 KALULU V. BONDI there any support for the similar idea that corroboration of a petitioner’s unrelated testimony can overcome a showing that she elsewhere lied. In looking at the record and circumstances as a whole, if the agency finds that there are sufficient testimonial inconsistencies in part of that evidentiary record, then the agency can properly rely on those findings to render all the petitioner’s other testimony non-credible. As long as the agency has some evidence showing non-trivial inconsistencies in the petitioner’s testimony, that is all that is needed to make an adverse credibility determination. It is irrelevant how much ostensibly consistent testimony was presented about other issues by the petitioner. 9 Similarly, the dissent contends that independent, corroborating documents somehow undercut the adverse credibility determination and potentially revive Petitioner’s testimony. As explained, however, a petitioner’s lack of credibility supported by substantial evidence in the record cannot be rehabilitated just because some of her testimony about issues unrelated to the agency’s adverse credibility determination is corroborated by other evidence. Once the agency has properly concluded that the petitioner lied about 9 Perhaps part of the dissent’s erroneous understanding of our standard of review stems from its fundamental misunderstanding of the nature of substantial evidence review—particularly the misconception that when engaging in this review this court can only rely on the facts the BIA expressly identified. That is wrong. See I.N.S. v. Elias-Zacarias, 502 U.S. 478, 481 (1992) (instructing that in conducting our substantial evidence review the record must be considered “as a whole”); Ramirez- Villalpando v. Holder, 645 F.3d 1035, 1039 (9th Cir. 2011) (stating that in this consideration, the court is not limited to the “evidence expressly identified in the BIA’s decision”); 8 U.S.C. § 1252(b)(4)(B) (“administrative findings of fact are conclusive unless any reasonable adjudicator would be compelled to conclude to the contrary”).  KALULU V. BONDI 23 some things, the fact that corroborating evidence indicates that she may not have lied about everything does not somehow repair her credibility. A petitioner may be able to sufficiently support her claims by relying on evidence other than her own testimony—indeed, that possibility is precisely why the majority is remanding to the agency in this case— but that is irrelevant to her damaged credibility. 10 Of course, there are certainly circumstances in which independent documentary evidence could be directly relevant to an adverse credibility determination. This could occur in situations where the independent evidence directly 10 In arguing for a remand on the adverse credibility issue, the dissent inexplicably cites to cases stating that remand is appropriate where an agency has not had the first chance to answer a question delegated to it. Of course. But those cases are inapplicable here, because nobody can dispute that the agency has already addressed Kalulu’s credibility. Instead, the appropriate standard of our review is the well-established substantial evidence standard. And applying that extremely deferential standard, we do not remand just because the agency’s decision is not supported by all of the evidence it purported to rely on. We instead ask only whether, with any factual errors corrected, substantial evidence remains that nonetheless supports the agency’s decision. So long as there is enough evidence supporting the agency’s decision that the record doesn’t compel a remand, we will defer to the agency, even where the agency’s decision is supported by less evidence than the agency may have thought. The dissent’s WWiJD- (“What Would an Immigration Judge Do”) the-second-time-around-with-a-corrected-record approach is something other than the substantial evidence standard. And notwithstanding the dissent’s attempt to portray its revisionist approach as somehow deferential to the agency, it is clearly much less so than a proper substantial evidence review. The dissent’s standard would result in much more frequent remands to the agency. Cases would be remanded whenever a court was unsure how the agency might redecide a case once any evidentiary mistakes were corrected, even in those cases (like here) where, once those evidentiary mistakes are accounted for, substantial evidence still supports the agency’s original decision. 24 KALULU V. BONDI relates to the basis for the adverse credibility determination. For example, if documents prove that those statements the agency found to be inconsistent are in fact consistent, then that would render the adverse credibility determination unsupported by substantial evidence. But that is not the case here. The so-called “corroborating evidence” that the Petitioner and the dissent reference in this case concerns wholly unrelated aspects of Kalulu’s testimony; it has nothing to do with the inconsistencies that formed the basis of the agency’s adverse credibility determination. As such, it is irrelevant to that determination. Finally, the dissent elsewhere seems to be inventing and applying a novel ratio test applicable to the agency’s factual bases for its credibility determination: compare the total number of facts the agency relied upon for its determination to the subset of just those facts the court finds reliable. If that ratio is too low (although the dissent nowhere tells us precisely where the cutoff is), then apparently remand of the credibility issue is appropriate, even if the agency’s determination remains amply factually supported by the record. Here, the dissent emphasizes that “about two-thirds of the findings identified by the Agency are not supported by the record.” Because that ratio is too low, the dissent asserts, we must send the credibility issue back. 11 11 The dissent claims the majority does not “dispute that the vast majority of the agency’s credibility findings” are not supported by sufficient evidence. But we do. By the majority’s count, at least five of the agency’s factual bases for its adverse credibility determination are supported. Even assuming all the other findings are unsupported, barely more than one-half of the agency’s findings related to credibility are unsupported, hardly a “vast majority.” Even by the dissent’s count, fully  KALULU V. BONDI 25 The dissent cites no authority whatsoever for this faux quantitative approach, and for good reason. As described above, it runs contrary to both the way in which a credibility determination works and the concept of substantial evidence, which defers to an agency’s decision even if the agency made some factual errors so long as sufficient evidence nonetheless supports the agency’s conclusions. The dissent’s error is also highlighted by the odd results it would lead to. As the dissent admits, there are at least four adequate bases for the agency’s adverse credibility determination in this case. It would be strange to conclude that had the agency mentioned and relied only upon these four correct facts (without mentioning any others), then its decision would stand; but because it relied on additional facts (right or wrong), remand is appropriate. And while the dissent acknowledges our precedent stating there is no bright line rule for how many factual bases must properly support the agency’s credibility finding, that seems to be precisely what the dissent’s ratio test would eventually lead to. Ultimately, the dissent’s new ratio test is not compatible with proper substantial evidence review, which both the Supreme Court one-third of the agency’s findings remain supported. Apparently “vast majority” too is being affected by inflation. We don’t point this out to be pedantic or to manufacture disagreement where our dissenting colleague is admirably trying to find common ground. Rather, this ungenerous characterization of the majority’s position relates back to the fundamental problems with the dissent’s novel ratio test. Our job is not to effectively redo the agency’s decision- making process based on some comparative ratio of its supported and unsupported findings. Nor is it our job to ask what decision we think the agency might make if any of its evidentiary mistakes are fixed. Instead, our job is simply to determine whether there is substantial evidence in the record to support the finding that the agency did make—here, that the Petitioner is not credible. 26 KALULU V. BONDI and this court have made clear requires that we must deny a petition unless, reviewing the record as a whole, “any reasonable adjudicator would be compelled to conclude to the contrary.” Ming Dai, 141 S. Ct. at 1677 (quoting 8 U.S.C. § 1252(b)(4)(B) (emphasis added)). 12 Once this court finds substantial evidence in the entire record to support the agency’s conclusion, our review is over, and we do not apply novel ratio tests or speculate about what the agency might have done had it read the evidentiary record precisely as we do. Contrary to the dissent’s inexplicable characterization, this is clearly not a case in which the rejected findings “all but gut” the basis for the agency’s determination. By the dissent’s own admission, there remain at least four independent and supported factual findings that together constitute more than substantial evidence supporting the agency’s credibility determination. This is all that is needed for us to deny the petition on this point. What this court cannot do, and what the dissent is effectively asking us to do, is reweigh evidence and make 12 It is worth adding that the dissent’s ratio approach is additionally problematic because it fails to take into account potential differences in the strength of the various facts supporting an adverse credibility determination. Imagine two different situations in which the agency relied on twelve factual findings to make its credibility determination. In the first, the court concludes that only four of the factual bases are supported by the record, but these four facts are non-trivial and strongly support the conclusion that the petitioner is not credible. In the second, the court concludes that seven of the factual bases are supported by the record, but are all very trivial and only questionably support a non- credibility finding. Under the dissent’s ratio approach, the former would be remanded while the latter would not, even though the agency’s credibility determination would be much more supported by “substantial evidence” in the former case. 93.2% of statistics give the illusion of quantitative certainty while providing very little in the way of substance.  KALULU V. BONDI 27 our own credibility determination. See Don v. Gonzalez, 476 F.3d 738, 743 (9th Cir. 2007). * * * Oddly, the dissent accuses the majority of “usurp[ing] the agency’s role by concluding that the agency would make the same adverse credibility determination on remand.” Not true. The majority takes no position about what the agency might do on remand because that isn’t the appropriate question to ask. We are simply taking the agency’s decision as a given, and asking if what the agency already did is supported by substantial evidence. It is the dissent’s novel approach that would require the court to partially assume the role of the agency decisionmaker, asking if the court is confident the agency would necessarily reach the same conclusion if the case was remanded on a corrected record. Whatever else that might be, it is not the substantial evidence review standard. c. The Case Nevertheless Is Remanded for the Agency to Properly Consider Whether Supporting Documents Independently Prove Kalulu’s Past Persecution Claim. If a petitioner who has been found noncredible provides independent evidence to support her claims, the agency must evaluate whether that evidence independently proves her claims. Here, Kalulu offered such documentary evidence in the form of purported eyewitness declarations to three attacks on her in Zambia and a purported medical record detailing injuries of the second attack. Because the agency discounted the documents’ evidentiary value based on a clear misreading of them, this case is remanded with instructions for the agency to reconsider whether, when 28 KALULU V. BONDI properly read, they independently prove Kalulu’s past persecution claim. The first document the agency must review properly is the declaration Kalulu offered from her second cousin, who Kalulu testified housed her after the first alleged attack and witnessed the second alleged attack. Kalulu’s cousin recounts in that declaration that: In the middle of the night Milly came into the house. I was asleep. She woke me up and told me that she was attacked. But she did not tell me who attacked her. She was very scared and lived in fear. Then I asked her why she doesn’t even want to go out anymore. That’s when she explained the story that made me sick to my stomach [that Kalulu was lesbian]. She first asked me if I can keep a secret? I told her that you are my favorite cousin and yes I will. Then she told me that Amina [Kalulu’s alleged girlfriend in Zambia] and her were a couple they are in a relationship with. She is into girls and the day she got attacked it was Amina’s brothers that beat her up. The IJ misread the declaration to expressly state that all these events occurred on the same night, which would be inconsistent with Kalulu’s testimony that weeks passed between the first attack and when she disclosed her sexual orientation to her cousin. But the declaration does not in fact state any timeline, neither its syntax nor grammar implies one, and common sense might even suggest that Kalulu’s  KALULU V. BONDI 29 cousin would not ask her “why she doesn’t even want to go out anymore” right after Kalulu allegedly arrived at her house bloodied and beaten in the middle of the night. In any event, the IJ’s misreading of the declaration led her to impermissibly discount its evidentiary value based on a clearly nonexistent inconsistency. The second document the agency must review anew is the declaration Kalulu offered from her second cousin’s neighbor, who she testified intervened to protect Kalulu during the second alleged attack and then drove her to the hospital. On its face, it purports to contain a stamp from a Zambian commissioner for oaths certifying it as a true copy of the original document. Yet the IJ characterized the declaration as lacking information to establish its authenticity, and the BIA characterized it as unsworn. When the IJ and BIA decisions are read together, the agency’s characterization makes sense only if it deemed the stamp inauthentic. It may or may not be. But it would be “illogical,” and thus impermissible, for the agency to discount the declaration on that basis, De Leon, 51 F.4th at 1000, given that it elsewhere discounted a different declaration as possibly inauthentic because it lacked the same stamp. The third eyewitness document the agency must reconsider is the declaration Kalulu offered from a friend in Zambia who Kalulu testified witnessed the third attack at the restaurant in the capital city of Lusaka. That declaration on its face purports to contain the friend’s handwritten signature. But the BIA incorrectly stated that it is “not signed or sworn … which undermines [its] evidentiary value.” The agency then improperly discounted the evidentiary weight of the declaration on that inaccurate basis. 30 KALULU V. BONDI Last, the agency must reconsider the medical record Kalulu offered as evidence, which on its face purports to be from the hospital where Kalulu allegedly received emergency care to stitch a stab wound she received during the second attack. The report mentions a deep cut on Kalulu’s chest. The IJ observed that this statement “is consistent with Respondent’s claim that she was stabbed in the chest during the [second] attack.” The BIA disagreed, finding without analysis or explanation that the cut to Kalulu’s chest does not “fully comport with” Kalulu’s testimony about being stabbed in the chest. IV. CONCLUSION This court grants a petition for review of an agency denial of asylum, withholding of removal, and CAT relief only under the most extraordinary circumstances. See Gutierrez-Alm, 62 F.4th at 1194; Sharma v. Garland, 9 F.4th 1052, 1060 (9th Cir. 2021). This is one of those rare instances. For the reasons discussed above, the agency’s adverse credibility determination is amply supported by substantial evidence. But the IJ failed to properly consider and evaluate the evidentiary weight of multiple documents Kalulu offered into the record independent of her testimony, and the BIA made clear factual errors when it reviewed those documents. Because the agency’s decision therefore “cannot be sustained upon its reasoning,” this case must be remanded for the IJ or BIA to reconsider its decision. De Leon, 51 F.4th at 1008 (internal quotation marks omitted). On remand, the agency must reexamine the three declarations and medical document discussed in section III(b) to consider whether they, when properly read alongside other nontestimonial evidence in the record, independently prove Kalulu’s claims for asylum or  KALULU V. BONDI 31 withholding of relief. This court takes no position on whether those documents provide such proof or whether Kalulu merits any of the relief for which she applied. PETITION GRANTED. SANCHEZ, Circuit Judge, concurring in part and dissenting in part: Petitioner Milly Kalulu, a native of Zambia, alleges she was persecuted because she is a lesbian in a country that criminalizes same-sex relationships. When her relationship with a woman was discovered by her girlfriend’s brothers, she was beaten, whipped, injected with an unknown substance, stabbed in the chest, doused with gasoline, and threatened with death over several violent encounters. Kalulu submitted documentary evidence corroborating her claims, including a copy of her medical report, a declaration from her aunt in California, and declarations from several Zambians who witnessed the attacks on her. The agency, however, dismissed this evidence based on unsupportable or trivial grounds. I agree with the majority that the agency failed to consider whether Kalulu’s supporting evidence independently proves her claims for asylum, withholding of removal, and relief under the Convention Against Torture (CAT). “Where potentially dispositive testimony and documentary evidence is submitted, the BIA must give reasoned consideration to that evidence.” Cole v. Holder, 659 F.3d 762, 772 (9th Cir. 2011); see also Antonio v. Garland, 58 F.4th 1067, 1077 (9th Cir. 2023) (“[W]here there is any indication that the agency did not consider all of 32 KALULU V. BONDI the evidence before it the decision cannot stand.” (cleaned up)). Remand is required where, as here, the agency did not give reasoned consideration to highly probative evidence that may independently support Kalulu’s claims of past persecution. But the agency’s failure to consider the documentary evidence was emblematic of other significant errors underlying its adverse credibility determination. The most egregious example? Disbelieving Kalulu’s claim that she is a lesbian because she had not visited gay clubs or participated openly in “LGBT activities” during her first five months in the United States. As the majority recognizes, two-thirds of the factors cited by the agency for its adverse credibility determination were based on dubious stereotyping, mischaracterizations of the testimony, or purported inconsistencies not found in the record. Where “the bulk of the [agency’s] credibility findings . . . are infirm,” our precedent requires that we “remand to the BIA to determine whether the few remaining factors are sufficient—in light of the totality of the circumstances—to support such a finding.” Kumar v. Garland, 18 F.4th 1148, 1151 (9th Cir. 2021). The majority ignores our precedent and instead concludes that the agency would have reached the same adverse credibility determination in the absence of these unsupported findings. That approach contravenes the REAL ID Act, binding circuit precedent, and fundamental principles of administrative law. I respectfully dissent. I. The agency made two findings regarding Kalulu’s eyewitness declarations and medical documentation: (1) the documents did not rehabilitate her testimony, and (2) they  KALULU V. BONDI 33 did not independently establish her claims of past persecution. As the majority correctly observes, those findings were “based on a clear misreading” of the documents. The agency manufactured inconsistencies, applied arbitrary authentication requirements, and overlooked or mischaracterized key portions of the documents to discredit them. The majority correctly concludes that the agency “failed to properly consider and evaluate” the documents and therefore “the agency’s decision . . . cannot be sustained upon its reasoning.” Based on the majority’s analysis, one might expect for the court to set aside both of the agency’s document-related findings. After all, the eyewitness declarations and hospital documents corroborated Kalulu’s testimony that she was attacked on multiple occasions, driven to the hospital, and received emergency treatment for a stab wound to her chest. The agency’s failure to consider whether this evidence independently established her claims of past persecution necessarily requires reexamining whether the documents also rehabilitate her testimony—that is, whether she testified truthfully about these violent incidents and about her sexual orientation as the underlying basis for these attacks. Inexplicably, however, the majority sets aside the second finding while upholding the first, concluding that “the agency’s adverse credibility determination is amply supported by substantial evidence.” The majority’s inconsistent reasoning contravenes the REAL ID Act, which requires the agency to consider “the totality of the circumstances” and “all relevant factors” when making a credibility determination. See 8 U.S.C. § 1158(b)(1)(B)(iii). “To ignore . . . relevant record evidence [is] to make an adverse credibility determination on less than the total circumstances in contravention of the 34 KALULU V. BONDI REAL ID Act’s text.” Shrestha v. Holder, 590 F.3d 1034, 1044 (9th Cir. 2010). The majority’s conclusion that the agency “failed to properly consider and evaluate” Kalulu’s corroborating evidence means the agency also ignored relevant evidence when making its adverse credibility determination. See Yan Rong Zhao v. Holder, 728 F.3d 1144, 1149 (9th Cir. 2013) (The BIA “is required to consider the evidence in its entirety . . . and where its failure to do so could have affected its decision, remand is appropriate.” (cleaned up)); see also Garcia v. Wilkinson, 988 F.3d 1136, 1142 (9th Cir. 2021). Because the agency failed to consider the evidence in its totality, I would remand for reconsideration of its adverse credibility determination. II. The agency’s adverse credibility determination must be set aside for a second reason: about two-thirds of the findings identified by the agency are not supported by the record. The IJ relied on thirteen factual reasons for its adverse credibility finding. The BIA explicitly addressed twelve of these factors, which we review for substantial evidence. See Shrestha, 590 F.3d at 1039 (“When the BIA conducts its own review of the evidence and law rather than adopting the IJ’s decision,” we review only the BIA’s decision “except to the extent that the IJ’s opinion is expressly adopted.” (cleaned up)). Of the dozen adverse credibility factors expressly adopted by the BIA here, two-thirds of those findings are not supported by substantial evidence in the record. My colleagues do not seriously dispute that the vast majority of the agency’s credibility findings are based on dubious stereotyping, mischaracterizations of the evidence, or purported inconsistencies that are not inconsistencies at all  KALULU V. BONDI 35 or that the agency did not allow Kalulu to address. As discussed below, only four of the agency’s dozen credibility findings survive our deferential standard of review. A. Implausibility Findings The agency found that Kalulu testified implausibly regarding her (1) status as a lesbian, and (2) delay in applying for asylum. Both of those findings relied on speculation and conjecture, which “cannot form the basis of an adverse credibility finding, which must instead be based on substantial evidence.” See Shah v. INS, 220 F.3d 1062, 1071 (9th Cir. 2000). The most egregious example was the BIA’s conclusion that the IJ “permissibly determined that the record evidence was insufficient to support the respondent’s assertion that she was a lesbian.” The IJ deemed Kalulu’s lesbian identity “implausible” because she did not visit “locations or clubs geared towards the LGBT community” after arriving in the United States. 1 The agency may not rely on “dubious” stereotypes to determine how the petitioner “ought to act.” See Munyuh v. Garland, 11 F.4th 750, 764 (9th Cir. 2021); Tan Xia Zhu v. Mukasey, 537 F.3d 1034, 1039 (9th Cir. 2008). The agency also found that Kalulu’s delay in applying for asylum undermined her claim that she was afraid to return to Zambia because of her sexual orientation. The agency reasoned that Kalulu’s “lack of effort to investigate her eligibility for fear-based relief undermines her claim that she was so afraid to return to Zambia that she decided to 1 The government, to its credit, acknowledges that the agency’s reasoning “reflects rank speculation and stereotyping” and agrees this factor is not supported by substantial evidence. 36 KALULU V. BONDI remain in the United States as soon as she received her tourist visa[.]” This finding lacks support in the record. In Guo v. Ashcroft, 361 F.3d 1194, 1201–02 (9th Cir. 2004), we rejected an adverse credibility finding based on the petitioner’s failure to apply for asylum immediately upon entering the United States. We concluded that the agency’s reasoning—that waiting to apply for asylum meant the petitioner was doing so only “as an afterthought”—was “conjecture and speculation [that] cannot substitute for substantial evidence.” Id. at 1202 (internal quotation marks omitted). Here, when asked why she did not apply for asylum during the five months she lived with her aunt, Kalulu testified that she “wasn’t aware” that asylum existed. Kalulu explained she had been “looking for ways to live in America” and traveled to the U.S.-Mexico border to speak with an immigration officer because she did not want to have any problems with an expired visa. Kalulu’s testimony establishes that she made “efforts” to remain in the United States, contrary to the agency’s unsupported conjecture. B. Inconsistencies in Kalulu’s Testimony The BIA relied on six purported inconsistencies in Kalulu’s testimony in determining that she lacked credibility. Inconsistencies that “form the basis of the IJ’s adverse credibility determination” must be “actually inconsistent in light of all ‘relevant record evidence.’” Bhattari v. Lynch, 835 F.3d 1037, 1043 (9th Cir. 2016) (quoting Shrestha, 590 F.3d at 1043–44). The agency may not base an adverse credibility determination on inconsistencies “when [it] did not ask [petitioner] about these discrepancies or give [petitioner] an opportunity to reconcile them.” Soto-Olarte v. Holder, 555 F.3d 1089,  KALULU V. BONDI 37 1092 (9th Cir. 2009). Finally, “[i]f the applicant gives a reasonable and plausible explanation” for the inconsistency, the agency “must state a specific and cogent reason for rejecting it.” Barseghyan v. Garland, 39 F.4th 1138, 1145 (9th Cir. 2022) (internal quotation marks omitted). The agency found that Kalulu testified inconsistently regarding six topics: (1) whether she feared returning to Zambia; (2) whether she planned to attend university in California; (3) the date she made arrangements with her aunt to remain in the United States; (4) whether she attended the Girl Scout Jamboree as a “regular scout” or a “leader”; (5) the state of her health before she was detained; and (6) whether anyone in her family knew she was a lesbian. The agency “considered these inconsistencies in the aggregate . . . to conclude that the respondent had not set forth a credible claim.” I agree that the first three inconsistency findings are supported by substantial evidence, but the remaining three inconsistency findings lack support in the record. 2 The agency found that Kalulu “initially testified that she attended the jamboree as a regular scout, but when questioned about the age limitations for the event, she changed her testimony to state that she attended as a leader.” Kalulu’s exchange with the IJ regarding her role in the jamboree follows: IJ: In what role were you attending this Jamboree? 2 I also agree that substantial evidence supports the agency’s adverse demeanor finding. The remaining three adverse findings arose from the agency’s misreading of the documentary evidence as discussed above. 38 KALULU V. BONDI Kalulu: Well, I was a participant, Your Honor. IJ: I’m sorry. You want to repeat that again? Kalulu: Participant. Participant. IJ: As a—as a participant? Kalulu: Yes, Your Honor. IJ: Just like a regular scout? Kalulu: Yes, Your Honor. [. . .] IJ: And isn’t, generally, the requirements to attend as a participant that you have to be between the ages of 14 and 17? Kalulu: Yes, Your Honor. IJ: Okay. But you were 22 at the time that you attended. So how . . . could you attend a youth jamboree of scouts at age 22? Kalulu: Because I was one of the leaders, so we’re all participating. IJ: Well, that’s interesting ma’am. Because that’s what I just asked you. How were you participating? And you said as a participant. And I just said as a regular scout participant, and you said yes. So why are you now changing to say that you were a leader?  KALULU V. BONDI 39 Kalulu: No, Your Honor. Because everybody, even the other people, we’re all participants. Immediately after this exchange, the government informed the IJ that Kalulu’s visa corroborated her testimony about participating in the jamboree. There was no actual inconsistency between Kalulu’s statements. One can, of course, be a “participant” and a “leader” at the same event, and as Kalulu testified, “everybody. . . we’re all participants.” 3 The agency “manufacture[d] a discrepancy by characterizing the factual situation as an ‘either/or’ situation.” Barseghyan, 39 F.4th at 1145. It is clear from the context of Kalulu’s testimony that she did not testify inconsistently, and even if she had, the IJ failed to address Kalulu’s reasonable explanation for her testimony. The agency found that Kalulu testified inconsistently regarding her health because she “initially testified that she did not have any health issues prior to being diagnosed with HIV in June 2020 while she was in the detention facility, but subsequently admitted that she had previously not felt well and went to a clinic for heart problems and obtained a pap smear.” 4 This finding is not supported by substantial 3 See Participant, OXFORD ENGLISH DICTIONARY (3d ed. 2005) (“A person who participates in something; a person who experiences something in common with others”). 4 While both the IJ and BIA found that Kalulu testified she had received treatment for “heart problems,” the transcript contains no such testimony. Instead, it appears that both the IJ and BIA misread the transcript in which Milly testified that she “had health problems.” The majority inappropriately corrects the record, replacing “heart problems” 40 KALULU V. BONDI evidence. When the IJ asked Kalulu if she had received a medical exam since she had been taken into custody, Kalulu responded, “[y]es . . . I was diagnosed [] with HIV.” The IJ then asked the compound question, “[b]efore that, did you ever have any indication that you had been HIV positive or that you were sick in any way?” Kalulu replied, “No, Your Honor.” Later, when asked why she had a Medi-Cal Benefits Identification Card in her possession when she was detained at the U.S.-Mexico border, Kalulu informed the IJ that she had been visiting a medical clinic in the United States for “health problems” and had obtained a pap smear. Even if the IJ perceived an inconsistency in Kalulu’s testimony as to whether receiving a “pap smear” constituted a preexisting “health issue,” the IJ deprived Kalulu of the opportunity to address that inconsistency. See Soto-Olarte, 555 F.3d at 1092. After questioning Kalulu about treatment she received at the clinic, the IJ changed the topic and began questioning her about how she obtained a California ID card. The IJ never mentioned that she perceived Kalulu’s statements as inconsistent, which the government tacitly acknowledges. Although the majority makes the threadbare assertion that the IJ gave Kalulu an opportunity to explain the purported inconsistencies in her statements regarding her health, the record does not support it. Kalulu’s supposedly inconsistent statements are in response to different lines of questioning at different parts of the hearing, separated by nearly fifty pages in the record transcript. The agency did not afford Kalulu an opportunity to explain any supposed inconsistency in her health testimony, and therefore cannot with “health problems”—substantively altering the agency’s decision to correct its erroneous reading of the transcript.  KALULU V. BONDI 41 rely on this inconsistency to support an adverse credibility determination. See Munyuh, 11 F.4th at 762. Finally, the agency found that Kalulu “provided conflicting testimony related to . . . whether anyone in her family knew that she was a lesbian.” Kalulu testified that she fled to her cousin’s house after she was beaten and kidnapped by her girlfriend’s brothers. The IJ then asked if she told her cousin that she was a lesbian: IJ: Did you tell your cousin what happened? Kalulu: No, I did not tell her at that time because I did not want her to contribute to the feelings I was feeling. And because she did not know that I was a lesbian. IJ: Did any of your family know that you were a lesbian? Kalulu: Nobody did. IJ: And why not? Kalulu: Because I was afraid to tell them. IJ: And why were you afraid? Kalulu: Because it’s a big taboo to be a lesbian in my country. And because it’s . . . illegal. At a later point in the hearing, the IJ asked again if anyone was aware that Kalulu was a lesbian: IJ: And you also stated that no one in Zambia knew that you were a lesbian except for [the girlfriend], correct? 42 KALULU V. BONDI Kalulu: Until I told my cousin. IJ: Okay. But up until that time . . . no one knew, correct? Kalulu: Yes, Your Honor. [. . .] IJ: Okay. And when did you tell your cousin about being a lesbian? Kalulu: About three weeks [after the incident] when she kept asking what happened to me. [. . .] IJ: Okay. So other than your cousin that we just talked about that you told . . . about being a lesbian, have you ever told any of your other family, including siblings, cousins, aunts, uncles, parents, anybody? Kalulu: No, Your Honor. IJ: Nobody at all. Kalulu: Only my auntie, the one who was here in America. Kalulu’s testimony about her family’s knowledge of her sexual orientation does not conflict. The IJ mischaracterized the record when she stated that Kalulu “changed her testimony and said that she told her cousin that she is a lesbian approximately three weeks after the incident[.]” Kalulu’s testimony that she did not tell her cousin she was a lesbian “at that time” is in reference to seeing her cousin just after she was attacked. Her statement immediately follows  KALULU V. BONDI 43 her testimony about fleeing to her cousin’s house after being beaten. Thus, Kalulu did not “change” her testimony when she later testified that she had informed her cousin she was a lesbian three weeks after the attack. And her immediate clarification, “only my auntie,” cannot be deemed an inconsistency. Unprompted, Kalulu clarified that other than her cousin, the only family member who knew she was a lesbian was her aunt living in California. See Ren v. Holder, 648 F.3d 1079, 1087 (9th Cir. 2011) (“[Petitioner]’s initial error . . . was quite clearly a quickly-corrected innocent mistake. As such, it cannot form the basis for an adverse credibility determination.”). C. Remand Is the Appropriate Remedy The main area of disagreement between the majority and dissenting opinions is not whether most of the agency’s adverse credibility factors are supported by substantial evidence in the record. It is clear they are not. But the majority concludes that if even only four or five of the twelve adverse findings are supported by the record, we must affirm that agency’s adverse credibility determination. This conclusion ignores binding circuit precedent and basic principles of administrative law. Although “[t]here is no bright-line rule” under which some specific number of infirm adverse credibility findings requires remand, Alam v. Garland, 11 F.4th 1133, 1137 (9th Cir. 2021), the rejected findings here “all but gut” the agency’s rationale, see Kumar, 18 F.4th at 1156. “Because so little remains in support of the adverse credibility finding,” our precedent requires that “we grant the petition and remand to determine whether the totality of the circumstances continues to support that finding.” Id. at 1153; see Barseghyan, 39 F.4th at 1141 (“We remand on an 44 KALULU V. BONDI open record for the BIA to determine in the first instance whether the remaining inconsistency is sufficient to support the adverse credibility determination.”). We do so because the agency has not had the opportunity to consider whether the small handful of remaining findings—a few minor inconsistencies and an observation about her demeanor during a single line of questioning—are sufficient on their own to discredit Kalulu’s testimony under a totality of the circumstances. Kumar, 18 F.4th at 1156. Fundamental principles of administrative law forbid us from answering that question ourselves. See Smith v. Berryhill, 139 S. Ct. 1765, 1779 (2019) (“Fundamental principles of administrative law, however, teach that a federal court generally goes astray if it decides a question that has been delegated to an agency if that agency has not first had a chance to address the question.”); see also Florida Power & Light Co. v. Lorion, 470 U.S. 729, 744 (1985) (“[I]f the agency has not considered all relevant factors, . . . the proper course, except in rare circumstances, is to remand to the agency for additional investigation or explanation.”). Indeed, both our precedent and the Supreme Court have consistently repudiated the majority’s approach. See, e.g., Coronado v. Holder, 759 F.3d 977, 987 (9th Cir. 2014) (“[U]nder the ordinary remand rule, we are not permitted to decide a claim that the immigration court has not considered in the first instance.” (internal quotation marks omitted)); Tekle v. Mukasey, 533 F.3d 1044, 1055–56 (9th Cir. 2008) (“Under INS v. Ventura, [537 U.S. 12, 16 (2002) (per curiam)], after determining that an adverse credibility finding is not supported by substantial evidence, we ordinarily remand to the BIA to ‘make the basic asylum eligibility decision.’”); Zi Lin Chen v. Ashcroft, 362 F.3d 611, 621 (9th Cir. 2004) (citing Ventura and noting that the  KALULU V. BONDI 45 Ninth Circuit has only recognized two “rare” exceptions to remand with respect to adverse credibility determinations); Barseghyan, 39 F.4th at 1146 (remanding for the BIA to determine whether one remaining inconsistency is sufficient to support an adverse credibility determination). More recently, the Supreme Court reversed the Sixth Circuit for “conduct[ing] a de novo inquiry” and “reach[ing] its own conclusions” in its review of an agency decision imposing sanctions against a former bank executive. Calcutt v. FDIC, 589 U.S. 623, 629 (2023). In that case, the Sixth Circuit found that the FDIC’s sanctions were supported by substantial evidence, “even if some findings . . . were incorrect.” Calcutt v. FDIC, 37 F.4th 293, 334–35 (6th Cir. 2022). The Supreme Court reversed, observing that the agency had “never . . . considered whether the sanctions against [the executive] were warranted on the narrower set of [findings] that the Sixth Circuit identified.” 598 U.S. at 628. Like an adverse credibility finding, the Court noted that an agency’s decision to impose sanctions “is highly fact specific and contextual, given the number of factors relevant to petitioner’s ultimate culpability.” Id. at 630. The majority here repeats the Sixth Circuit’s error by substituting its own judgment for that of the agency. 5 5 The majority falls back on a strawman argument that “because other parts of the record contain unrelated testimony that has not been shown to be inconsistent,” the dissent contends that “those parts should be reweighed against Petitioner’s inconsistencies to reevaluate her credibility.” A careful reading of the dissent shows that no such comparison in testimony was made. Rather, I have reviewed the agency’s findings for substantial evidence and explained how a clear majority of the agency’s adverse credibility factors are infirm, a conclusion my colleagues do not seriously dispute. The REAL ID Act 46 KALULU V. BONDI Accordingly, I would remand not only for the agency to consider whether Kalulu’s supporting documentary evidence independently establishes her claims of past persecution, but also to determine in the first instance whether the remaining factors on their own support an adverse credibility determination. and our precedents therefore require that we remand for the agency to determine in the first instance whether the remaining inconsistencies are sufficient, under a totality of the circumstances, to support an adverse credibility finding. See Barseghyan, 39 F.4th at 1146; Alam, 11 F.4th at 1137; Kumar, 18 F.4th at 1155–56. It is notable that the majority fails to grapple with this caselaw, and what the majority labels a “novel ratio test” is simply an application of our established precedent. At bottom, it is the majority that usurps the agency’s role by concluding that the agency would make the same adverse credibility determination on remand. That is not the province of this court.