Concurring opinion filed by Senior Circuit Judge SENTELLE, with whom Senior Circuit Judge EDWARDS joins.
SRINIVASAN, Circuit Judge:
The National Association of Criminal Defense Lawyers submitted a request under the Freedom of Information Act to obtain an internal Department of Justice publication known as the Federal Criminal Discovery Blue Book. The Blue Book is a manual created by the Department to guide federal prosecutors in the practice of discovery in criminal prosecutions. It contains information and advice for prosecutors about conducting' discovery in their cases, including guidance about the government’s various obligations to provide discovery to defendants.
The Department refused to disclose the Blue Book, invoking the Freedom of Information Act’s Exemption 5, which exempts from disclosure certain agency records that would be privileged from discovery in a lawsuit with the agency. The Department maintained that the Blue Book fell within the attorney work-product privilege, and therefore Exemption 5, because it was prepared by (and for) attorneys in anticipation of litigation. The district court agreed that the Blue Book is privileged attorney work product and thus is exempt from disclosure. We likewise conclude that the Blue Book consists of exempt attorney work product, but we remand to the district court for an assessment of whether the Blue Book also contains non-exempt policy statements amenable to reasonable segregation from the privileged work product.
I.
A.
The Freedom of Information Act (FOIA) provides that government agencies must make agency records available to citizens upon request, subject to nine enumerated exemptions. 5 U.S.C. § 552. “Congress intended FOIA to permit access to official information long shielded unnecessarily from public view.” Milner v. Dep’t of Navy, 562 U.S. 562, 565, 181 S.Ct. 1259, 179 L.Ed.2d 268 (2011) (internal citations and quotation marks omitted). The statutory exemptions, accordingly, “are explicitly made exclusive and must be narrowly construed.” Id: (internal citations and quotation marks omitted).
Under Exemption 5, agencies may withhold “inter-agency or intra-agency memorandums or letters which would not be available by law to a party other than an agency in litigation with the agency.” 5 U.S.C. § 552(b)(5). Exemption' 5 covers records that would be “normally privileged in the civil discovery context.” NLRB v. Sears, Roebuck & Co., 421 U.S. 132, 149, 95 S.Ct. 1504, 44 L.Ed.2d 29 (1975). The exemption allows the government to withhold records from FOIA disclosure under at least three privileges: the deliberative-process privilege, the attorney-client privilege, and the attorney work-product privilege. Coastal States Gas Corp. v. Dep’t of Energy, 617 F.2d 854, 862 (D.C. Cir. 1980). This case solely involves the last of those privileges, the attorney work-product privilege.
[250]*250B.
On December 20, 2012, the National Association of Criminal Defense' Lawyers (NACDL) sent a FOIA request to the Department of Justice (DOJ) seeking disclosure of the Blue Book. DOJ processed the request under the direction of Susan Gerson, Assistant Director in the FOIA/Privacy Act Staff of the Executive Office for U.S. Attorneys. After reviewing the Blue Book and consulting with DOJ staff familiar with the Book’s inception and drafting, Gerson determined that it should be withheld in full pursuant to FOIA Exemptions 5 and 7(E). (The latter, exemption, which we have no occasion to reach, exempts certain types of “records or information compiled for law enforcement purposes.” 5 U.S.C. § 552(b)(7)(E).) On February 28, 2013, Gerson sent NACDL a form letter invoking both exemptions and denying its request.
After unsuccessfully appealing the denial, NACDL brought this FOIA suit in the district court seeking to compel- DOJ to release the Blue Book. The parties filed cross-motions for summary judgment, and DOJ again invoked Exemptions 5 and 7(E) in support of its decision to withhold the Blue Book in full. Because the parties disagreed about how to characterize the Blue Book’s contents, the district court reviewed the Book in camera before rendering its decision on the merits of DOJ’s decision to withhold.
Following its in camera review of the Blue Book, the district court granted DOJ’s motion for summary judgment. Nat’l Ass’n of Criminal Def. Lawyers v. Exec. Office for U.S. Attorneys, 75 F.Supp.3d 552 (D.D.C. 2014). The court found that DOJ was not required to disclose-any portion of the Blue Book, holding that the Book in its entirety is protected as attorney work product. Id. at 557, 561. Because the court found that the Blue Book could.be withheld in full under Exemption 5, it did not address the applicability of Exemption 7(E). Id. at 556.
II.
NACDL appeals, arguing that the claimed FOIA exemptions are inapplicable and the Blue Book therefore should be disclosed in full. We review the district court’s decision de novo. See Judicial Watch, Inc. v. Dep’t of Justice, 432 F.3d 366, 369 (D.C. Cir. 2005). In doing so, we must “ascertain whether the agency has sustained its burden of demonstrating that the documents requested are ... exempt from disclosure under the FOIA.” Assassination Archives & Research Ctr. v. CIA, 334 F.3d 55, 57 (D.C. Cir. 2003) (quoting Summers v. Dep’t of Justice, 140 F.3d 1077, 1080 (D.C. Cir. 1998)) (quotation marks omitted). When making that determination, we rely centrally on the agency’s descriptions of the content of the relevant documents as set forth in its Vaughn index and accompanying affidavits. See Mead Data Cent., Inc. v. U.S. Dep’t of Air Force, 566 F.2d 242, 251 (D.C. Cir. 1977).
We conclude that the Blue Book consists of attorney work product which is exempt from disclosure under FOIA’s Exemption 5.
Courts have long recognized that materials prepared by one’s attorney in anticipation of litigation are generally privileged from discovery by one’s adversary. See Hickman v. Taylor, 329 U.S. 495, 510-12, 67 S.Ct. 385, 91 L.Ed. 451 (1947); In re Sealed Case, 146 F.3d 881, 884 (D.C.Cir. 1998). The attorney work-product privilege applies in both civil and criminal cases. See United States v. Nobles, 422 U.S. 225, 236-38, 95 S.Ct. 2160, 45 L.Ed.2d 141 (1975); [251]*251Fed. R. Civ. P. 26(b)(3); Fed. R. Crim. P. 16(a)(2), 16(b)(2)(A).
The privilege aims primarily to protect “the integrity of the adversary trial process itself.” Jordan v. Dep’t of Justice, 591 F.2d 753, 775 (D.C. Cir. 1978) (en bane). It does so by “provid[ing] a working attorney with a ‘zone of privacy’ within which to think, plan, weigh facts and evidence, candidly evaluate a client’s” ease, and prepare legal theories.” Coastal States, 617 F.2d at 864. Without the privilege, “much of what is now put down in writing would remain unwritten” because “[a]n attorney’s thoughts, heretofore inviolate, would not be his own.” Hickman, 329 U.S. at 511, 67 S.Ct. 385. Protecting attorney work product from disclosure prevents attorneys from litigating “on wits borrowed from the adversary.” Id. at 516, 67 S.Ct. 385 (Jackson, J., concurring).
“Congress had the attorney’s work-product privilege specifically in mind when it adopted Exemption 5” to the FOIA. Sears, 421 U.S. at 154, 95 S.Ct. 1504. Not every document created by a government lawyer, however, qualifies for the privilege (and thus, the exemption). “[I]f an agency were entitled to withhold any document prepared by any person in the Government with a law degree simply because litigation might someday occur, the policies of the FOIA would be largely defeated.” Coastal States, 617 F.2d at 865. To avoid that result, we have long required a case-specific determination that a particular document in fact was prepared in anticipation of litigation before applying the privilege to government records. See, e.g., Senate of Puerto Rico v. U.S. Dep’t of Justice, 823 F.2d 574, 586-87 (D.C. Cir. 1987).
In ascertaining whether a document was prepared in anticipation of litigation, we have applied a “ ‘because of test, asking whether, in light of the nature of the document and the factual situation in the particular case, the document can fairly be said to have been prepared or obtained because of the prospect of litigation.” United States v. Deloitte LLP, 610 F.3d 129, 137 (D.C. Cir. 2010) (quoting Sealed Case, 146 F.3d at 884). For that standard to be met, the attorney who created the document must have “had a subjective belief that litigation was a real possibility,” and that subjective belief must have been “objectively reasonable.” Sealed Case, 146 F.3d at 884.
B.
We find those standards satisfied with, regard to the Blue Book. DOJ explains that “[t]he Blue Book was designed to provide advice regarding the law and practice of federal prosecutors’ discovery disclosure obligations and to serve as a litigation manual to be used by all DOJ prosecutors and paralegals” in their cases. Goldsmith First Decl. ¶ 5 (J.A. 93). As a result, DOJ says, the Blue Book was “created in anticipation of reasonably foreseeable litigation,” namely, federal criminal prosecutions. Gerson Decl. ¶ 17 (J.A. 84); see id. ¶ 20 (J.A. 85), We agree.
The Blue Book “describes] the nature and scope of [federal prosecutors’] discovery obligations under applicable constitutional provisions, caselaw, and the Federal Rules- of Criminal Procedure.” Gerson Deck ¶ 20 (J.A. 85). It consists of “nine chapters, written by DOJ prosecutors with expertise in a wide range of discovery-related topics,” addressing subjects including: Federal Rule of Criminal Procedure 16, regarding discovery; the government’s obligations to disclose exculpatory information under Brady v. Maryland, 373 U.S. 83, 83 S.Ct. 1194, 10 L.Ed.2d 215 (1963) and Giglio v. United States, 405 U.S. 150, 92 S.Ct. 763, 31 L.Ed.2d 104 [252]*252(1972); disclosure duties arising from the Jencks Act, 18 U.S.C. § 3500; items protected from disclosure; and the use of protective orders and ex parte and in camera submissions in discovery. Goldsmith First Decl. ¶ 5 (J.A. 93).
According to DOJ, the Blue Book is not a “neutral analysis of the law” but rather “contain[s] confidential legal analysis and strategies' to support the Government’s investigations and prosecutions.” Gerson Deck ¶ 21 (J.A. 86). In contrast with publicly-available documents such as the United States Attorneys’ Manual, which set out statements of agency policy, the Blue Book is an internal manual containing litigation strategies. Goldsmith Second Deck ¶ 7 (J.A. 103). It gives “practical ‘how-to’ advice,” Goldsmith First Deck ¶ 5 (J.A. 94), to federal prosecutors about “how to handle different scenarios and problems,” Ger-son Deck ¶ 20 (J.A. 85). It discusses “the types of challenges [prosecutors] may encounter in the course of prosecutions and potential responses and approaches.” Goldsmith Second Deck ¶ 8 (J.A. 103); accord Goldsmith First Deck ¶ 14 (J.A. 99). The Book “contemplates facts that may arise in judicial proceedings” and evaluates “how a court would likely consider those facts.” Goldsmith First Deck ¶ 6 (J.A. 94).
DOJ thus argues that disclosing the Blue Book would “essentially provide a road map to the strategies federal prosecutors employ in criminal cases.” Id. It contends that disclosure would afford anyone who wanted to read the Blue Book (including opposing counsel) “unprecedented insight into the thought processes of federal prosecutors.” Id. ¶ 12 (J.A. 98). Disclosure thus would “undermine the criminal trial process by revealing the internal legal decision-making, strategies, procedures, and opinions critical to the Department’s handling of federal prosecutions.” Id. ¶ 13 (J.A. 98). In addition, it would “severely hamper the adversarial process[,] as DOJ attorneys would no longer feel free to memorialize critical thoughts on litigation strategies for fear that the information might be disclosed to their adversaries to the detriment [of] the government’s current and future litigating positions.” Id.
Taking into account the nature, content, and function of the Blue Book as described in DOJ’s affidavits, we believe it “can fairly be said to have been prepared ... because of the prospect of litigation.” Deloitte, 610 F.3d at 137 (quoting Sealed Case, 146 F.3d at 884). Our in camera review of the Blue Book confirms that the affidavits accurately describe the Book and its contents. The Book therefore consists of protected attorney work-product.
C.
NACDL does not dispute that the Blue Book was prepared for use in litigation. It claims that the Blue Book nonetheless falls outside the work-product privilege for three reasons: (i) the Blue Book was not prepared in anticipation of litigating a specific claim or case; (ii) the Blue Book principally serves a non-adversarial function; and (iii) the Blue Book’s content resembles that of a neutral treatise. We find each of those arguments unpersuasive.
1.
NACDL’s principal contention is that the Blue Book cannot qualify for the work-product privilege because, even if it was created in contemplation of litigation generally, it was not prepared in anticipation of litigating a specific claim or case. NACDL reads our. decisions to establish a specific-claim requirement for government documents to qualify for work-product protection, at least in the context of government-initiated litigation such as criminal prosecutions. NACDL misunderstands our decisions.
[253]*253As an initial matter, we have long held that there is no general, overarching requirement that a governmental document can fall within the work-product privilege only if prepared in anticipation of litigating a specific claim. See Sealed Case, 146 F.3d at 885; Schiller v. NLRB, 964 F.2d 1205, 1208 (D.C. Cir. 1992); Delaney, Migdail & Young, Chartered v. IRS, 826 F.2d 124, 126-28 (D.C. Cir. 1987). In Schiller, for instance, we held that the privilege covered NLRB documents providing direction and advice to agency lawyers on the litigation of cases under the Equal Access to Justice Act. 964 F.2d at 1208. We specifically rejected the contention “that the work-product doctrine requires that the documents be created- in anticipation of litigation over a specific claim.”. Id. “Exemption 5 extends to documents prepared in anticipation of foreseeable litigation,” we explained, “even if no specific claim is contemplated.” Id. (citing Delaney, 826 F.2d at 127).
That is not to say that anticipation (or non-anticipation) of a specific claim can never have any relevance when assessing the applicability of the work-product privilege. The contemplation of specific claims can help differentiate situations in which lawyers have litigation adequately in mind from those in which lawyers are not (yet) sufficiently anticipating litigation.
Our decision in Coastal States Gas Corporation v. Department of Energy, 617 F.2d 854, is illustrative. Coastal States involved memoranda drafted by Department of Energy lawyers to assist Department auditors in interpreting agency regulations. The auditors used the memoranda when auditing firms for compliance with agency regulations. Those audits were not considered “investigations,” because, “at that point, no charge had been made mor was a violation necessarily suspected.” Id. at 858. But if an auditor subsequently determined that a firm had committed a violation, the audit could turn into a more targeted investigation and, in some cases, give rise to litigation. See id. at 858-60.
We denied application of the work-product privilege to the memoranda in Coastal States, finding “no indication ... that there was even the dimmest expectation of litigation when the[ ] documents were drafted.” Id. at 865. We said that “[t]o argue that every audit is potentially the subject of litigation is to go too far.” Id. The Department thus had “failed to carry its burden of establishing that litigation was fairly foreseeable at the time the memoranda were prepared.” Id. We distinguished the memoranda from documents “prepared with a specific claim supported by concrete facts which would likely lead to litigation in mind.” Id. Accordingly, in a later case involving documents prepared “in the course of an investigation” that “had reached the stage ... at which [the ageney] was comparing the accumulated facts to the caselaw and evaluating” specific legal theories to pursue, we held that litigation was “sufficiently ‘in mind’ for [the] documents] to qualify as attorney work product.” SafeCard Servs. Inc. v. SEC, 926 F.2d 1197, 1203 (D.C. Cir. 1991).
In both Coastal States and SafeCard, we used language suggesting that, for the documents in question to-be prepared with an enforcement action sufficiently on the horizon to implicate the work-product privilege, the agency’s investigation must have advanced to the point that the documents’ authors contemplated bringing “a specific claim supported by concrete facts.” SafeCard, 926 F.2d at 1202 (quoting Coastal States, 617 F.2d at 865). In neither case, however, did we establish any across-the-board specific-claim prerequisite for application of the privilege. Indeed, as noted, we later specifically disavowed any such [254]*254specific-claim requirement in Schiller, 964 F.2d at 1208.
Rather, the point of the specific-claim inquiry in Coastal States and SafeCard was to differentiate between audits as to which enforcement litigation might well never take place, Coastal States, 617 F.2d at 865, and active investigations with an enforcement action foreseeably at hand, SafeCard, 926 F.2d at 1202-03. In those cases, looking at whether agency attorneys were contemplating a specific claim proved useful in assessing the likelihood that litigation would ever come to pass. At an early stage, such as a neutral compliance audit with no specific claim (or even any violation) in mind, there is insufficient reason to anticipate litigation. But for documents created at a later stage in which the agency contemplates bringing a specific action, disclosure might reveal the government’s “legal theories,” “weighting of] facts and evidence,” or “candid[] evaluation]” of a case. See Coastal States, 617 F.2d at 864.
We face a very different situation here. A specific-claim requirement would make little sense in the context of the Blue Book. Unlike the audit documents at issue in Coastal States—which might well never be used (or be of use) in litigation—the Blue Book is entirely about the conduct of litigation. It is aimed directly for use in (and will inevitably be used in) litigating cases. Its disclosure therefore risks revealing DOJ’s litigation strategies and legal theories regardless of whether it was prepared with a specific claim in mind. It was prepared with the litigation of all charges and all cases in mind. The presence or absence of a specific claim or transaction might be a helpful consideration in the context of an agency compliance inquiry with no enforcement action or litigation necessarily on the horizon. But it is an unhelpful consideration here given that the Blue Book undoubtedly was created in anticipation of—and for use in—foreseeable litigation, i.e., federal criminal prosecutions.
NACDL relies heavily on our decision in Sealed Case, 146 F.3d 881. There, we considered the applicability of the work-product privilege to a lawyer’s notes and other documents prepared in anticipation of a possible action brought against the lawyer’s client. We found the documents were covered by the privilege even though no specific claim against the client had yet emerged. See id. at 885-86. In discussing the significance of the fact that no specific claim had arisen by the time of the documents’ creation, we drew a distinction between two types of cases. First, we pointed to Coastal States and SafeCard in observing that, in prior eases in which “the documents at issue had been prepared by government lawyers in connection with active investigations of potential wrongdoing,” we had required anticipation of a specific claim in order to invoke the work-product privilege. Id. at 885. By contrast, in cases like Schiller, in which lawyers acted “not as prosecutors or investigators of suspected wrongdoers, but as legal advisors protecting their agency clients from the possibility of. future litigation,” we rejected the need for a specific claim to implicate the privilege. Id. at 885-86. The facts in Sealed Case fell into the latter category.
This case, according to NACDL, fits within the former category because it involves lawyers acting “as prosecutors.” While that may be so, Sealed Case did not hold that, in any case involving documents prepared by or for prosecutors, the work-product privilege could apply only if the documents had been created in anticipation of a specific claim. Instead, because Sealed Case did not involve documents created by prosecutors, we expressly declined to address whether “the Coastal States/SafeCard specific claim test has any contin[255]*255ued vitality when government lawyers act as prosecutors or investigators of suspected wrongdoers.” Id. at 885. And as we have explained, the existence (or non-existence) of a specific claim proved salient in those cases as a means of identifying whether documents had been prepared at a time when litigation was sufficiently in mind—i.e., whether “litigation was a real possibility.” Id. at 884. But in the case of a document like the Blue Book, prepared entirely for use in wholly foreseeable (even inevitable) litigation, there is no need to apply any. specific-claim test to conclude that litigation is sufficiently likely to warrant application of the work-product privilege.
2.
NACDL next argues that the Blue Book falls outside the work-product privilege because it aims to advance a non-adversarial function—namely, education and training of prosecutors. NACDL advocates drawing a line between (unprivileged) documents that convey agency policy and (potentially privileged) documents that help the agency prevail in court. Whatever the validity of such a line, it would not advance NACDL’s cause because the Blue Book was designed to help federal prosecutors-prevail in court on behalf of the government.
We have long recognized that the applicability of the work-product privilege can turn in significant measure on a document’s function. See Delaney, 826 F.2d at 127 (citing Coastal States, 617 F,2d at 858). And we agree with NACDL that materials serving no cognizable adversarial function, such as policy manuals, generally would not constitute work product. See Coastal States, 617 F.2d at 863. The Blue Book, however, does serve an adversarial function.
The Book does not merely pertain to the subject of litigation in the abstract. Instead it addresses how attorneys on one side of an adversarial dispute—federal prosecutors—should conduct litigation. It describes how to respond to the other side’s arguments, which cases to cite, and what material to turn over and when to do so, among numerous othér practical and strategic considerations. The Blue Book, for instance, “describes the types of claims defense counsel have raised and could raise regarding different discovery issues, or the tactics they could employ in-litigation ,.. and the arguments prosecutors can make to respond to these' claims.” Gerson Decl. ¶ 21 (J.A. 85).
In any event, insofar as the Blue Book might also serve non-adversarial functions, “a document can contain protected work-product material even though it serves multiple purposes, so long as the protected material was prepared because of the prospect of litigation.” Deloitte, 610 F.3d at 138. As a result, “material generated in anticipation of litigation may also be used for ordinary'business purposes without losing its protected status.” Id. In that light, any educational or training function the Blue Book might serve would not negate the document’s adversarial use in (and its preparation in anticipation of) litigation. The Blue Book therefore qualifies for the work-product privilege.
3.
NACDL also argues that a neutral recitation of legal rules or case law in the manner of a treatise, as opposed to a description of a lawyer’s litigation strategy or theory of the case, fails to qualify as attorney work product. That distinction is of no assistance to NACDL because the latter category more fairly describes the Blue Book than does the former.
In Schiller, we found the work-product privilege applicable to “lawyer-prepared [256]*256documents containing tips and advice for litigating cases under the Equal Access to Justice Act.” Sealed Case, 146 F.3d at 886 (describing Schiller). Much like the documents at issue in Schiller, the Blue Book contains case-handling tips and tactical advice for litigating discovery matters in criminal prosecutions: “in addition to legal analysis,” the Blue Book includes “a comprehensive set of strategic considerations, procedures, and practical advice for conducting. criminal prosecutions,” much of which is “interspersed within the legal analysis.” Goldsmith First Decl. ¶ 9 (J.A. 96).
As a result, unlike “neutral” accounts of government policy like the United States Attorneys’ Manual, the Blue Book imparts litigation strategy to government lawyers: it conveys “advice on criminal discovery practices, potential strategic and logistical concerns, interpretations of law and risk assessments in light of relevant legal authority, ... practice notes, techniques, procedures, and legal strategies that in-the-field prosecutors may and do employ during the course of criminal proceedings.” Id. ¶ 6 (J.A. 94). As with the tips and advice in Schiller, the Blue Book thus consists of exactly the “sort of information— prepared in anticipation of litigation— [which] falls within the attorney work-product privilege and, therefore, within [EJxemption 5.” Schiller, 964 F.2d at 1208.
To be sure, the Blue Book contains certain information—such as “compilations of cases,” Gerson Deck ¶ 21 (J.A. 85)—that may come with a seeming air of neutrality if considered in strict isolation. But disclosure of the publicly-available information a lawyer has decided to include in a litigation guide—such as citations of (or specific quotations from) particular judicial decisions and other legal sources—would tend to reveal the lawyer’s thoughts about which authorities are important and for which purposes. The Blue Book, for instance, does not include lists of cases in a vacuum. It instead “offers compilations of cases that prosecutors can use to support different arguments” in litigation as well as “[c]ases illustrating potential pitfalls that prosecutors should avoid” when conducting discovery. Id. (J.A. 85-86). That sort of information squarely implicates the work-product privilege.
III.
Finally, NACDL argues that, even if certain portions of the Blue Book qualify as work product and thus are exempt from disclosure, DOJ must disclose any non-exempt portions. NACDL relies on the FOIA’s direction to agencies to disclose any non-exempt “portion” of a record containing exempt material if the nonexempt parts are “reasonably segregable.” 5 U.S.C. § 552(b). And as we have long held, “[t]he focus of the FO'IA is information, not documents, and an agency cannot justify withholding an entire document simply by showing that it contains some exempt material.” Mead Data, 566 F.2d at 260.
As the district court noted, however, an agency need not segregate and disclose non-exempt material if a record is “fully protected” as work product. Nat’l Ass’n of Criminal Def. Lawyers, 75 F.Supp.3d at 557 (quoting Judicial Watch, 432 F.3d at 371). In such cases, because the entire record is exempt from disclosure, there are no non-exempt portions left to segregate. The district court found that the Blue Book is fully protected as work product and thus did not undertake a separate segregability analysis. Id. at 561.
In cases involving voluminous or lengthy work-product records—the Blue Book is more than 500 pages in length—we think it generally preferable [257]*257for courts to make at least a preliminary-assessment of the feasibility of segregating nonexempt material. When reviewing such records in camera, courts may look at “what proportion of the information in a document [appears to be] non-exempt and how that material is dispersed throughout the document.” Mead Data, 566 F.2d at 261. Material is more likely to be reasonably segregable in longer documents with “logically divisible sections.” See id. at 261 n.54. In such cases, courts -presumably would examine each section to determine if it might be amenable to segregation and disclosure. Such a determination also may be possible on the basis of the agency’s Vaughn index and affidavits, if those materials suggest that a lengthy work-product record likely contains segregable material.
We' recognize that “FOIA places the burden of justifying nondisclosure on the agency seeking to withhold information, and this burden cannot be shifted to the courts by sweeping, generalized claims of exemption for documents submitted for in camera inspection.” Id. at 260. But when an agency has maintained all along that a record is “fully protected” as work product, its Vaughn index and affidavits may not address segregability. In such cases, it may be that portions of the record which otherwise appear to contain neutral information are encompassed within (and integrated with) protected work product and thus there is no portion that is “reasonably segregable.” But there may also be cases in which a record containing some amount of work product also contains—or at least appears to contain—segregable, non-exempt material subject to disclosure. In that circumstance, a court presumably would require the agency to provide “a description of which parts of the withheld documents are nonexempt ... and either disclose them or offer adequate justification for continuing to withhold them.” Id.
In this case, the government acknowledges that it has already publicized its general policy statements about federal prosecutors’ discovery obligations. In its response to the petition for rehearing filed after we issued our original (and hereby superseded) opinion in this case, the government emphasized that the Department of Justice “has made public the agency policy governing federal prosecutors’ discovery obligation,” and provided website links to pertinent sections of the United States Attorneys’ Manual and three memoranda issued by the Deputy Attorney General addressing the topic. Gov’t Resp. to Pet. for Reh’g 11-12. The government’s submissions in this case also confirm that the Blue Book contains a discussion of those policy statements. See Gov’t Br. 46, 61-62. One of the government’s declarations, for instance, notes that Chapter 1 of the Blue Book is entitled, “Department of Justice Policy, Positions, and Guidance.” Goldsmith Second Decl. ¶ 7 (J.A. 103).
In light of the government’s submissions, we think it appropriate to assess whether the Blue Book contains nonexempt statements of policy that are reasom ably segregable from the protected attorney work product and that therefore should be disclosed. Because the district court did not consider whether the Blue Book contains reasonably segregable statements of the government’s discovery policy, we remand for that court to conduct that analysis in the first instance. Such an analysis, we have explained, does not call for parsing the Blue Book “line-by-line” or segregating material “dispersed throughout the document.” Mead Data, 566 F.2d at 261. Instead, the emphasis is on segregation of non-exempt material found in “logically divisible sections.” Id. at 261 n.54. If the district court concludes that the Blue Book contains non-exempt and reasonably segregable statements of the [258]*258government’s discovery policy, the court could then consider whether Exemption 7(E) of FOIA would protect any of that material from disclosure. The district court to this point has had no occasion to examine the government’s argument that Exemption 7(E) shields the Blue Book from disclosure.
* * * * *
For the foregoing reasons, we affirm the judgment of the district court in part and remand for an assessment of whether the Blue Book contains non-exempt and reasonably segregable statements of discovery policy.
So ordered.