Reversed, vacated, and remanded by published opinion. Judge GREGORY wrote the majority opinion, in which Senior Judge ALARCÓN joined. Judge SHEDD wrote a dissenting opinion.
OPINION
GREGORY, Circuit Judge:Carlton Luck (“Luck”) appeals the district court’s partial denial of his motion to set aside his conviction under 28 U.S.C. § 2255. Luck argues that his conviction for conspiracy to distribute and possess with intent to distribute fifty grams or more of cocaine base should be reversed because his trial counsel was ineffective for failing to request an informant instruction, declining to call impeachment witnesses, putting Luck’s prior incarceration before the jury, and not sharing Jencks material turned over by the government with Luck. We find that trial counsel was ineffective in failing to request an informant instruction and vacate Luck’s conviction and sentence.1
I.
In 2005, Luck was indicted for conspiracy to distribute and to possess with intent to distribute fifty grams or more of cocaine base (Count One); using, brandishing, or carrying a firearm during and in relation to a drug trafficking crime (Count Two); and distribution of cocaine base (Counts Three and Four). On March 21, 2005, he proceeded to trial on all charges before a jury.
At trial, the government presented evidence from three main witnesses: the police officer investigating the case and two paid informants.2 Atrina Johnson (“Johnson”) and Jesse Thompkins (“Thompkins”) were the government’s main witnesses. Both were informants paid for providing information to the police. Johnson approached the police on her own initiative to provide information about Luck because she wanted to reduce her exposure for a robbery charge she was facing. Thomp*185kins also came forward without prompting by the police. He provided information on Luck to benefit Johnson, who was the mother of his child, and to earn money for himself.
Johnson testified that she met Luck through Thompkins, who was regularly purchasing cocaine base from Luck. Johnson also knew Luck because she regularly stored drags in her attic for another man known as JaB. She testified that when JaB was arrested, Luck came to her house and retrieved the drugs JaB stored there. Thereafter, while facing a state robbery conviction, Johnson asked the police if she could work as an informant in return for a favorable report to her sentencing judge. Johnson then conducted two controlled buys for the police. On March 18, 2004, she called Luck on his mobile and asked him to sell her cocaine base. She then went to Luck’s house and bought cocaine base from his brother. Johnson testified that she again called Luck on March 24, 2004, and arranged to buy cocaine base, then went to his house and purchased it from him. Charlottesville Police Detective Brian O’Donnell (“O’Donnell”) wired Johnson with video and audio recording equipment for the buys, but the image and sound quality were so poor that essentially nothing was recorded. O’Donnell was not present at the buys, so he had no direct knowledge of the dealings. Additionally, while he patted Johnson down before the buys, he did not search any of her inner garments for narcotics.
Thompkins also testified as to his knowledge of Luck’s drug dealing. Thompkins frequently interacted with Luck, buying cocaine base from Luck for personal use two to three times a day for a year. He also testified that he saw Luck sell drags to other people. Additionally, Thompkins once sold Luck a pit bull, for which Luck paid $100 in cash and $100 in cocaine base. On April 29, 2004, a warrant was executed on Luck’s house. The search recovered scales and baggies commonly used to package narcotics; however, no drugs were recovered. The police also seized Luck’s mobile phone, which was the number Johnson had called to set up the buys, and a phone bill for the mobile phone showing that Luck was the account holder.
After the government presented its case in chief, Luck called several family witnesses in rebuttal. Both his mother and his aunt testified that he came to Charlottesville after a period of incarceration to turn his life around and that they had no knowledge of his involvement with drugs. Luck’s uncle, who lived with Luck, testified that he never saw him involved in any drug transaction.
After deliberating a short while, the jury returned a guilty verdict on all counts, and Luck was subsequently sentenced to 444 months imprisonment. Luck’s direct appeal was denied by this Court. United States v. Luck, 200 Fed.Appx. 263 (4th Cir.2006). On June 23, 2008, Luck filed a pro se motion to set aside his sentence under 28 U.S.C. § 2255 alleging that both his trial and appellate counsel provided ineffective assistance. In particular, Luck argued that trial counsel was ineffective for not requesting an informant instruction, not calling impeachment witnesses, inadequately preparing his mother and aunt to testify, not preventing the admission of two photographs, and failing to give him access to discovery materials. He also argued that appellate counsel was ineffective by failing to argue that Luck’s indictment was constructively amended by the trial judge’s instructions.
The district court granted Luck’s motion with respect to the issue of constructive amendment and ineffective assistance by appellate counsel. Therefore, the court reversed Luck’s convictions on Counts *186Two, Three and Four. The district court however denied Luck’s motion as to the rest of his claims and affirmed his conviction on Count One. Luck timely appealed.3
II.
In an appeal of the denial of a motion under § 2255, this Court reviews the district court’s decisions of law de novo. United States v. Poindexter, 492 F.3d 263, 267 (4th Cir.2007). The question of whether trial counsel provided ineffective assistance is a mixed question of fact and law which this Court reviews de novo. Smith v. Moore, 137 F.3d 808, 817 (4th Cir.1998).
III.
The issue before this Court on appeal is whether Luck’s trial counsel provided ineffective assistance on a number of grounds. We address only his first argument: whether trial counsel was unconstitutionally deficient for failing to request an informant instruction. We find that counsel was ineffective and vacate Luck’s conviction and sentence as to Count One of the indictment.
A.
Questions of ineffective assistance of counsel are governed by the Supreme Court’s decision in Strickland v. Washington, 466 U.S. 668, 104 S.Ct. 2052, 80 L.Ed.2d 674 (1984). There, the court detailed the two-prong approach courts must use when deciding ineffective assistance questions. The defendant bears the burden of proof as to both prongs of the standard. First, the defendant must show that counsel’s representation “fell below an objective standard of reasonableness” as measured by “prevailing professional norms.” Id. at 688, 104 S.Ct. 2052. Courts should be deferential in this inquiry, and have “a strong presumption that counsel’s conduct falls within the wide range of reasonable professional assistance.” Id. at 689, 104 S.Ct. 2052. The defendant must therefore overcome the presumption that the representation “might be considered sound trial strategy.” Id. (citation and internal quotation marks omitted).
Second, the defendant must demonstrate that counsel’s inadequate performance prejudiced him. Id. at 687, 104 S.Ct. 2052. Thus, the defendant must show “a reasonable probability that, but for counsel’s unprofessional errors, the result of the proceeding would have been different.” Id. at 694, 104 S.Ct. 2052. A reasonable probability, in turn, is defined as “a probability sufficient to undermine confidence in the outcome.” Id. In cases where a conviction has been the result of a trial, the defendant must demonstrate that but for counsel’s errors, there is a reasonable probability that he would not have been convicted.
B.
Luck argues that his trial counsel was ineffective because he failed to request an “informant instruction” consisting of the following:
The testimony of an informer who provides evidence against a defendant for pay, or for immunity from punishment, or for personal advantage or vindication, must be examined and weighed by the jury with greater care than the testimony of an ordinary witness. The jury must determine whether the informer’s *187testimony has been affected by interest or by prejudice against a defendant.
United States v. Brooks, 928 F.2d 1403, 1409 (4th Cir.1991) (quoting Devitt & Blackmar, Jury Practice and Instructions § 17.02 (3d ed.1977)). He argues that this Court should follow several circuits in holding that this instruction is always mandatory, or at the least, that in this case counsel was ineffective for failing to request the instruction.
In Brooks, this Court faced the issue of when an informant instruction is required, but did not answer that question one way or the other because we found that the witnesses were not “informers” within the meaning contained in the Devitt and Blackmar treatise. 928 F.2d at 1409. Since then, the issue has not again been raised before this Court, though we have approved of the district court giving such an instruction in several published cases. See, e.g., United States v. Sullivan, 455 F.3d 248, 258-59 (4th Cir.2006) (stating that the jury was properly given an informant instruction when requested by counsel); United States v. Levenite, 211 F.3d 454, 463 (4th Cir.2002) (same); United States v. Anty, 203 F.3d 305, 310 (4th Cir.2000) (same). The only discussion of the issue occurred in an unpublished case prior to Brooks. There, Judge Sprouse in dissent argued that because there was little corroborating evidence for informant’s testimony on the charge, the court should have given an informant instruction. United States v. Worthington, 911 F.2d 726, at *5 (4th Cir.1990) (table) (Sprouse, J., dissenting). Therefore, whether an informant instruction is required or advisable in a case involving paid informants is an open question in this Circuit.
Among the other circuits that have considered this question, there is a consensus that an informant instruction is necessary when the informant’s testimony is uncorroborated by other evidence. See United States v. Bosch, 914 F.2d 1239, 1247 (9th Cir.1990); United States v. Hill, 627 F.2d 1052, 1054-55 (10th Cir.1980); United States v. Garcia, 528 F.2d 580, 587-88 (5th Cir.1976); United States v. Griffin, 382 F.2d 823, 828 (6th Cir.1967). These courts have explained that an informant instruction is necessary because a general witness credibility instruction is not sufficiently cautionary for informants because of special concerns about the incentive that they have to fabricate information for their own benefit. See United States v. Williams, 59 F.3d 1180, 1183-84 (11th Cir.1995) (stating that sole function of an informant instruction is to make jury aware that an informant’s testimony is to be viewed with caution); Garcia, 528 F.2d at 588 (“When the case is close and the witness particularly unreliable ... this Court has declared that the failure to give a cautionary instruction amounts to plain error.”); see also On Lee v. United States, 343 U.S. 747, 757-58, 72 S.Ct. 967, 96 L.Ed. 1270 (1952) (“The use of informers, accessories, accomplices, false friends, or any of the other betrayals which are ‘dirty business’ may raise serious questions of credibility. To the extent that they do, a defendant is entitled to broad latitude to probe credibility by cross-examination and to have the issues submitted to the jury with careful instructions.”). In other words, the jury needs to be instructed to scrutinize informant testimony more carefully than other witnesses, even biased witnesses, because of the potential for perjury born out of self-interest. See Alexandra Natapoff, Snitching 11 (2009) (“[W]hen defendants do go to trial, numerous exonerations reveal just how often juries believe lying criminal informants, even when juries know that the informant is being compensated and has the incentive to lie. A report by the Center on Wrongful Convictions at Northwestern School of Law describes fifty-one *188wrongful capital convictions, each one involving perjured informant testimony accepted by jurors as true.”).
C.
While the case law regarding paid informants generally is persuasive, we need not decide the question of whether and when an informant instruction is required, however, because on these facts it is clear that Luck’s counsel erred in failing to request an informant instruction. This case presents the classic case of a professional informant paid for his services, which in turn makes it the obvious case for an informant instruction.4 See, e.g., Garcia, 528 F.2d at 588 (remarking, in an analogous case where the government’s main witness was a professional, paid witness, that there was “more than the usual need for a cautionary instruction”). Both Johnson and Thompkins presented themselves to police, without any prompting, as sources of evidence for hire. They sought out the police and offered to gather evidence and testify in return for certain benefits: time off of Johnson’s robbery sentence and financial compensation.
Further, there was little corroborating evidence beyond the informants’ testimony. There was literally no corroborating evidence for Thompkins’ claims that he was buying cocaine base from Luck several times a day for a year. Additionally, while Johnson was equipped with recording equipment for her buys, the quality of the recordings was so poor that they were not even used at trial. Detective O’Donnell somewhat corroborated her testimony by observing that she came back from the buys with cocaine base, but he also admitted that he did not conduct a thorough search before Johnson interacted with Luck. Therefore a reasonable lawyer would have been concerned that the uncorroborated testimony of paid informants could have been “manufactured ... out of whole cloth” for the benefit of the informant alone. Griffin, 382 F.2d at 829 (internal quotation marks omitted).
From the facts above, a reasonable attorney would have requested an informant instruction. If there was ever a time to ask to have the jury instructed that paid informants raise special issues about credibility, this was the case. See Robert S. Hunter, Fed. Trial Handbook: Criminal § 75.22 (2009) (“While the testimony of an informer is competent evidence, it should be accompanied by instructions designed to call the attention of the jury to the character of the informer, leaving to the jury the question of the value and credibility of his testimony.”). Furthermore, there is no indication that failing to request the instruction had any root in trial strategy. Luck’s counsel cross-examined both Thompkins and Johnson about the consideration they received from the government for their investigation and testimony. In general, Luck’s counsel’s defense strategy seems to be focused on discrediting the government’s witnesses, rather than offering his own explanation of the events. Thus, were we to hold that failing to complete the strategy and request an informant instruction was a tactical decision by counsel, we would “conjure up [a] tactical decision[] an attorney could have made, but plainly did not.” Griffin v. Warden, 970 F.2d 1355, 1358 (4th Cir.1992). Therefore, Luck has established that his attorney’s conduct met the performance prong of Strickland.
*189D.
Having determined that Luck has established the performance prong, we now consider whether he has shown prejudice. Our inquiry regarding the prejudice prong is twofold: (1) whether the instruction, if requested, should have been given; and (2) if the instruction had been given, was there a reasonable probability that the outcome of the proceedings would have been different. As to the first issue, if requested, the trial court should have given the informant instruction. There is no debate that in a case like this where the government produces very little evidence beyond the uncorroborated testimony of paid informants, an informant instruction should have been given. See Sullivan, 455 F.3d at 258-59 (noting, in a case which involved testimony by a co-defendant, that the district court properly warned the jury to especially consider the credibility of testimony by a biased witness); Garcia, 528 F.2d at 588 (“When the case is close and the witness particularly unreliable ... this Court has declared that the failure to give a cautionary instruction amounts to plain error.”).
The second issue then concerns whether giving the informant instruction could have reasonably changed the probability of the outcome of the trial when the district court gave a general credibility instruction. In this case, the district court gave the jury general instructions regarding witness credibility including:
You are the sole judges of credibility or believability of each witness and the weight to be given the witness’s testimony. In weighing the testimony of a witness, you should consider his relationship to the government or to the defendant; his or her interest, if any, in the outcome of the case....
Therefore, whatever punishment ... another defendant may or may not have— there’s no other defendant in here. We’re talking about any other conspirator in this case — may or may not have received in the past or any punishment that he or she may or may not receive in the future should not enter into your deliberations unless you feel the witness is somehow motivated to make the statements they did on the stand in an effort to help them.
Of course, this does not prevent you from considering the motivation or credibility of a particular witness who may have chosen to testify in hopes of receiving a lighter or reduced sentence. However, while you should carefully consider the motivation and credibility of each witness, the specific punishment he or she may or may not receive is a matter strictly for the court.
J.A. 457, 459.5 It is true that the district court’s general instructions on witness credibility contained all of the elements of the informant instruction. However, the informant instruction is sui generis; it alerts jurors to the potentially unique problems that inhere where an individual is paid to inculpate a defendant. Cf. Anty, 203 F.3d at 310 (citing Edward H. Devitt et al., Federal Jury Practice and Instructions § 15.02 (4th ed.1992)) (discussing the creation of the special credibility instruction as one of the reasons why paid informant testimony may be considered). By summarizing the tools that the jury must use to evaluate credibility and applying it specifically to the case of the informant, the instruction more effectively cautions the jurors to think closely about the testimony. See Natapoff at 198 (“Jury instructions are a classic and crucial vehicle for shaping verdicts. Because jurors are the ultimate fact finders in criminal trials, *190charged with the task of evaluating witness credibility and figuring out what “really” happened, their evaluation of the informant testimony is central to the criminal process.”). Indeed, we have, in several cases, approvingly noted that the jury was given an informant instruction as one of the reasons why there was no prejudice to the defendant when the witnesses against him were paid informants. See, e.g., Levenite, 277 F.3d at 463 (acknowledging the dangers of working with a paid witness but stating that because, among other things, the informant instruction was given “the necessary safeguards were fully employed”). Our case law also suggests that corroboration is an important factor as to the reliability and trustworthiness of paid informant testimony. Id. at 462.
Therefore, our precedent counsels that had the informant instruction been given, there is a reasonable probability that the outcome of the proceeding would have been different. The government’s case was built entirely on a foundation of paid informant testimony. There was minimal physical evidence. No government agent had ever observed Luck engaged in drug activity. Merely giving general instructions as to witness credibility is not sufficient to give confidence that the outcome was not tainted by prejudice. In this case in particular, there was a significant incentive for the government’s two main witnesses to give testimony for their own benefit, both in terms of financial compensation and a reduction in sentence. Therefore, Luck’s trial counsel was ineffective when he failed to request an informant instruction when it would have been reversible error for the court to refuse to give it if requested, and that ineffectiveness prejudiced the outcome of Luck’s trial because the jury was not cautioned to consider the special problems of credibility posed by the government’s paid informants.
IV.
Because we find that Luck’s counsel was unconstitutionally defective in a way that prejudiced Luck, we reverse the decision of the district court and grant his Section 2255 motion. We therefore vacate his conviction and sentence on Count One.
REVERSED, VACATED, AND REMANDED
. Because we conclude that Luck's conviction should be vacated as to the issue of an informant instruction, we do not address the questions of the impeachment witnesses, prior incarceration, and Jencks material.
. The government also presented testimony from a neighbor of Luck’s. As his testimony went to the firearm count, which is not before this court, it is not recounted here.
. Luck was resentenced to 288-months imprisonment on August 11, 2009. His appeal of that sentence is currently pending before this court in case 09-4744, and this Court denied his motion to consolidate with the current case. As today we vacate his conviction, his pending case is now moot and review will be denied.
. Furthermore, as discussed in Part III.D below, if requested, the trial judge should have given the informant instruction in this case, thus rendering Luck’s counsel’s failure to request the instruction even more deficient.
. "J.A. -" refers to the joint appendix filed by the parties upon appeal.