Petitioner Mamadou Aliou Diallo, a native and citizen of Guinea, seeks review of a July 25, 2007 order of the Board of Immigration Appeals (“BIA”) affirming the November 14, 2005 decision of Immigration Judge (“IJ”) William Van Wyke denying Petioner’s application for asylum, withholding of removal, and relief under the Convention Against Torture (“CAT”). Diallo’s claims were based on alleged political persecution, including multiple arrests and torture, that he suffered as a result of his affiliation with the Rally of the People of Guinea Party (“RPG”), a political party that opposes the government of President Lansana Conté.

The IJ found Petitioner’s testimony non-credible. On appeal to the BIA, Diallo argued that he was credible, particularly emphasizing that his testimony was consistent, responsive, and sufficiently detailed. He further argued that the State Department Country Reports (“Country Reports”) corroborated his claims, and that the IJ overemphasized small disparities between Diallo’s testimony and what was, or was not, included in those Country Reports. The BIA summarily adopted the IJ’s decision without rejecting any of the IJ’s reasoning nor specifying which of the IJ’s particular findings supported its decision.

I. Standard of Review

Where, as here, the BIA summarily affirms an IJ’s adverse credibility finding, we review the factual and legal findings contained in the IJ’s opinion, including those aspects not discussed by the BIA. See Yun-Zui Guam, v. Gonzales, 432 F.3d 391, 394 (2d Cir.2005) (per curiam). We review the IJ’s adverse credibility finding under the substantial evidence standard, which requires that the decision be supported by “reasonable, substantial and probative evidence in the record.” Lin Zhong v. U.S. Dep’t of Justice, 480 F.3d 104, 116 (2d Cir.2007) (internal quotation marks omitted). While this standard is highly deferential to the IJ, we must determine whether the IJ has provided “specific, cogent” reasons for the adverse credibility finding and whether those reasons bear a legitimate nexus to the finding. Id. at 117. If the testimony provided is “generally consistent, rational, and believable, the presence of some inconsistent testimony need not necessarily be fatal to a petitioner’s claims if the disparities are relatively minor and isolated and do not concern material facts.” Xiao Ji Chen v. U.S. Dep’t of Justice, 471 F.3d 315, 335 (2d Cir.2006) (internal quotation marks omitted). ^[1]

[*235] In cases of this sort, therefore, we look to see whether there are significant issues that were raised before the BIA on which the BIA made errors. ^[2] In Chenery 1. the Supreme Court held that in reviewing a determination or judgment which an administrative agency is alone authorized to make, the reviewing court must evaluate such a decision “solely by the grounds invoked by the agency.” S.E. C. v. Chenery Corp., 332 U.S. 194, 196, 67 S.Ct. 1575, 91 L.Ed. 1995 (1947) (“Chenery II”) (referring to its decision in S.E.C. v. Chenery Corp., 318 U.S. 80, 63 S.Ct. 454, 87 L.Ed. 626 (1943) (“Chenery I"). That is, we cannot substitute our judgment of what would be a more adequate or proper ground for a decision if the agency’s decision did not rest on those grounds. “To do so would propel the court into the domain which Congress has set aside exclusively for the administrative agency.” Id. Consequently, if we find an error in the BIA’s decision on an issue that was exhausted before it, because we cannot substitute our judgement for that of the agency, Chenery I would seem to require us always to remand.

We have, however, held that consistent with Chenery I we may affirm an adverse credibility determination even when the IJ’s reasoning is deficient, provided that we can confidently predict that upon a reconsideration cleansed of errors, the agency would reach the same result. Lin Zhong, 480 F.3d at 117; Cao He Lin v. U.S. Dep’t of Justice, 428 F.3d 391, 395, 401-02 (2d Cir.2005). ^[3] If, instead, there are reasons for doubting that the agency would come out the same way, then we cannot say that remand would be futile. Such reasons can be any number of sorts. For example, the agency may have changed its position on a critical issue since the appealed decision, or the law may have otherwise changed. Another instance that could defeat futility is if a petitioner could, on remand, proffer materials to the agency that were not brought to the agency before and that are of a strength sufficient possibly to convince the agency to come out differently. Such materials might be newly found and therefore not before the agency earlier, but they might also be matters that had not previously been adequately considered because they were not brought to the BIA’s attention on the first go-around. In all such instances, we cannot be confident that remand on the basis of the error the agency did commit would be an “empty and unnecessary formality,” Lin Zhong, 480 F.3d at 117, and hence remand is necessary.

It is important to note that such a remand does not give a petitioner an unwarranted second bite at the apple. What materials and matters may be raised on a remand is always subject to the relevant statutes and BIA regulations. As a Court, we only look to what could be considered by the agency on remand, and decide whether, given Chenery I and the errors made by the agency the first time, remand is useless and thus futile.

[*236] Moreover, there is nothing in this approach that is inconsistent with the reasons for our refusal to review unexhausted issues. When we held in Theodoropoulos v. INS, 358 F.3d 162 (2d Cir.2004), that a district court lacked jurisdiction to entertain a habeas claim where the petitioner did not exhaust his remedies before the BIA, we emphasized the importance of exhaustion when reviewing decisions of administrative agencies. Id. at 165. We noted that because we are not in a position to make the ultimate decision on facts that can be read in multiple ways until we have the benefit of the BIA’s judgment on them, it was essential that each issue be exhausted before the BIA. Only when we have the agency’s judgment to review can we adequately consider the relevant issues. But when we remand on the basis of exhausted issues, because we are not confident of what the BIA would do on remand with respect to other questions, we do not ultimately consider any issues that the BIA has not reviewed. Our decision to remand asks only whether, in addition to the errors that the BIA has made, there are other issues that are plausible enough so that, on reconsideration, the agency might change its decision. What we do is simply give the BIA an opportunity to make its decision cleansed of errors, which only then becomes subject to our review.

With that in mind, we turn to the case at hand.

II. Application

In this case, we find errors in the BIA’s decision based on arguments Petitioner raised before it. On appeal to the BIA, Diallo argued that he was credible, particularly emphasizing that his testimony was consistent, responsive, and sufficiently detailed. He further argued that the Country Reports corroborated his claims, and that the IJ overemphasized small disparities between Diallo’s testimony and what was, or was not, included in those Country Reports. The Government concedes that Diallo fully exhausted this latter argument, Respondent’s Br. at 14 n. 8, and a review of the administrative record reveals that the IJ erred in his reading of the background material on conditions in Guinea and in his conclusion that the Country Reports did not corroborate Dial-lo’s testimony. For example, the IJ found that there was no corroboration for Dial-lo’s testimony that the Guinean government arrested “more or less like sixty to seventy people” in November 2001, at the time of Diallo’s second arrest. But the 2001 Country Report stated that “over fifty protesters were arrested prior to the November 11, 2001 referendum, and more than sixty students were arrested in late November of that year.” The Government argues that the Country Report actually undermines Petitioner’s testimony because it only refers to arrests prior to or after November 11, whereas Diallo testified to arrests on that date. But Diallo only testified that he was arrested on November 11, and there is no particular reason to think that he was referring only to that date when discussing the arrests of the others. Instead, he mentioned the figures of sixty to seventy people in response to the IJ’s question of “[h]ow many people were arrested in November 2001?” which did not specify a particular day.

Additionally, in contrast to the IJ’s conclusion, the Country Report’s mention of two-week detentions does not undermine Petitioner’s testimony about his own prolonged incarceration or those of his uncle and father. While the Country Report specifically noted groups being released two weeks after the November 2001 protests, it also said that security forces “committed abuses with impunity,” “used arbitrary arrest and detention,” and “continued to hold an unknown number of political prisoners” as of the end of 2001. Moreover, the Report described a “parallel and covert system of justice run by unidentified uniformed personnel who conduct midnight arrests, detain suspects, and use torture in secret prisons to obtain confessions .... ”

[*237] The IJ reasoned that because the Country Reports were particular as to many events, and because anti-government groups were vigilant about reporting abuses to human rights groups, the failure of human rights groups to mention any given arrest or abuse called into question its existence. This approach, however, places undue weight on the Country Reports, which, no matter how well-researched, cannot be expected to capture all of the details of every abuse in a given country. See Tian-Yong Chen v. INS, 359 F.3d 121, 130 (2d Cir.2004). It also exaggerates the ability of such anti-government groups to monitor all abuses. We therefore conclude that the IJ erroneously found that the Country Reports contradicted Diallo’s testimony, and this error undermines its ultimate decision that Diallo was not credible.

Given these errors in an issue that Diallo exhausted below, we must next consider whether it would be futile to remand this case to the BIA. See Cao He Lin, 428 F.3d at 401-02. Petitioner raises many other issues to us with respect to the IJ’s original ruling. A fair number of these were not expressly raised before the BIA, and at least some of them could properly be argued to that agency on remand. As to these questions, there is no BIA judgment before us to review, and thus we are not suggesting what findings would be appropriate. We note only that these criticisms of the IJ’s decision are plausible enough that were we to remand the case, it might be that the BIA would, on the basis of these arguments, change its adverse credibility determination. As a result, we cannot confidently predict that the BIA, having erred in its reading of the Country Reports, and having these other questions before it, would come to the same result upon reconsideration. It follows that remand is not futile, and is required.

III. Conclusion

We hold that — given the presence of some errors in the BIA’s decision as to issues that were properly exhausted — we can, in determining whether remand to the BIA would be futile, consider the plausibility of other newly claimed errors, i.e., errors that were not adequately raised before the agency. Because there were sufficient exhausted flaws in the reasoning of the IJ’s adverse credibility determination to warrant a remand and because we cannot confidently predict the same outcome if there were further agency reconsideration, we hereby GRANT the petition for review, and REMAND to the BIA for further proceedings consistent with this decision. Should the BIA determine that additional development of the record is necessary, it should remand to the IJ for that purpose. Accordingly, the order of removal is VACATED and Petitioner’s motion for a stay of removal is DISMISSED as moot.