Oscar Sosa was convicted of bringing methamphetamine from Mexico into the United States. Sosa argues that three errors he did not object to during his five-day trial rendered the proceeding unfair. Two of his claims-the prosecution's use of drug profiling evidence and bolstering of witnesses' credibility-are errors that we have repeatedly warned the government about. Troubled as we are by the continued use of these improper tactics, we do not find that Sosa has met his burden of showing that the errors substantially affected the outcome of the trial. United States v. Olano , 507 U.S. 725 , 736, 113 S.Ct. 1770 , 123 L.Ed.2d 508 (1993). We also find no clear or obvious Confrontation Clause violation.

*618 I.

The case against Sosa began with the arrest of two drug couriers, Juan Sarmiento and Jose Galvan, in Harlingen, Texas. A DEA task force received a tip about two suspicious subjects heading to a gas station that also served as a bus stop. Two sheriff's deputies and a DEA agent stopped Sarmiento and Galvan before they could board a bus and obtained consent to search their luggage and persons. The officers found six bundles of crystal meth sewn into the lining of Sarmiento's jacket and four bundles in Galvan's pockets. Galvan and Sarmiento were arrested and interviewed.

The two suspects gave conflicting statements about the origin of the drugs, but both said they were planning on taking the meth to a man named Oscar in Plant City, Florida. They also identified two women, "Betty" and "Patti," as the "owners or managers" of the narcotics. Police later identified those individuals as Patricia Sosa and Bertha Sosa, Oscar Sosa's mother and aunt, respectively, who lived in Mexico and supplied the drugs. The attempt to identify "Oscar" led investigators to Sosa. The investigation further revealed that Genaro Luera was connected to Patricia and Bertha because they were his wife's aunts.

The government argued that the conspiracy worked like this: (1) Patricia and Bertha obtained the drugs in Mexico; (2) Sosa and Luera hired couriers such as Galvan and Sarmiento to pick up the drugs from the U.S. side of the Texas-Mexico border and transport them to Florida; and (3) Sosa and Luera received the meth in Florida where they sold it.

Galvan, Sarmiento, and Luera all pleaded guilty and agreed to cooperate, which included testifying against Sosa. All three identified Oscar (both in-court and through a photo array) as a member of their drug trafficking organization. In addition to the testimony of these three cooperators, the government called DEA Agent Jason Bradford who testified as an expert on drug trafficking. A Customs and Border Patrol officer also testified about the various familial relationships. The jury convicted Sosa on both of the charged counts: possession with intent to distribute 50 grams or more of meth and conspiracy to commit that offense.

Sosa now challenges his conviction based on three alleged errors that he did not identify in the district court: first, that the government presented impermissible profiling evidence when the DEA agent linked the profile of a drug dealer to Sosa's behavior; second, that the government improperly bolstered the credibility of all three eyewitnesses; and third, that the government violated the Confrontation Clause when a DEA agent testified about a tipster's statements that inculpated Sosa's mother. Sosa also argues the conviction should be reversed based on the cumulative effect of the three errors, even if each error alone does not warrant reversal.

II.

Because Sosa failed to object to this testimony, which might have eliminated the errors, he must surmount the significant hurdles of plain error review to receive a new trial. There are four of them: (1) there must be an "error or defect," (2) the error must be "clear or obvious, rather than subject to reasonable dispute," (3) the error "must have affected the appellant's substantial rights," and (4) the error must have "seriously affect[ed] the fairness, integrity or public reputation of judicial proceedings." Puckett v. United States , 556 U.S. 129 , 135, 129 S.Ct. 1423 , 173 L.Ed.2d 266 (2009).

*619 A.

The first alleged error is that the government introduced impermissible profiling testimony by having the expert witness not only describe the typical aspects and behavior of a drug trafficking organization but also tell the jury where Sosa fit into that structure. An expert witness may explain to a jury the mechanics of a drug trafficking organization. See United States v. Gonzalez-Rodriguez , 621 F.3d 354 , 364 (5th Cir. 2010). When stated in general terms, such testimony may help the jury "understand the significance and implications of" certain conduct. Id. ; United States v. Medeles-Cab , 754 F.3d 316 , 321 (5th Cir. 2014). But the ultimate responsibility of linking a defendant's conduct with the typical characteristics of drug trafficking must be left to the jurors. Gonzalez-Rodriguez , 621 F.3d at 364 . If the profile testimony itself makes that connection, then it crosses into the forbidden territory in which testimony with the "expert" imprimatur is allowed to opine on the ultimate issue of guilt which is for the "trier of fact alone." FED. R. EVID. 704(b) ; Medeles-Cab , 754 F.3d at 321 .

Agent Bradford's testimony stepped well past the "fine but critical line" between "expert testimony concerning methods of operation unique to the drug business, and testimony comparing a defendant's conduct to the generic profile of a drug courier." Gonzalez-Rodriguez , 621 F.3d at 364 . Bradford began with acceptable testimony describing the typical roles within a drug trafficking organization, such as the couriers, the wholesalers, and the distributors. But the testimony invaded the province of the jury when Bradford began matching those roles to individuals in the case, including the defendant, in this manner:

[the Prosecution] - ... Can you kind of draw this drug trafficking organization with the players that you know from your investigation.

[Agent Bradford] - ... All right. So, again, we're going to start off with the source of supply or transportation coordinator. Sometimes they're the same, sometimes they're different, okay?

In this case, we recognize Patti and Bertha Sosa as filling those roles, okay? The investigation showed that an unknown man named Freddie LNU, meaning last name unknown, filled the role of the courier because he would take the drugs into the United States. Okay?

Once it got to the United States, we had more couriers, okay? And you got to see them live and in person. That was Mr. Galvan and Mr. Sarmiento, okay? They went to a wholesaler, then the distributor who we will recognize as Mr. Oscar Sosa.

Oscar had clients. That was Mr. Genaro Luera. And then Mr. Luera sold to, and this is unique, back to Mr. Galvan and other people. Okay?

Mr. Luera also spoke about unknown other clients and customers that Mr. Sosa had, but he didn't have extensive knowledge of.

That alone would be impermissible. But the government went even further in eliciting testimony (via a leading question no less) that Sosa's specific behavior was "common of drug traffickers":

[the Prosecution] - When you're looking at [Sosa's records] and you're not finding any assets-well, really, three things-not finding any assets in Mr. Sosa's name, is that somehow in fact indicative of-of some things that say, okay, you know what, it's kind of strange he doesn't have any assets in his name, that tells me what?

[Agent Bradford] - Yes, we consider that conduct common of drug traffickers.

*620 [the Prosecution] - And why is that?

[Agent Bradford] - Because they don't want to leave a trail for their assets.

Given the number of times we have found quite similar profile testimony improper, the above testimony was obvious error. See, e.g., Gonzalez-Rodriguez , 621 F.3d at 366 (finding obvious error when a law enforcement agent expressly linked a defendant's behavior with general profiling evidence to prove knowledge of the crime); United States v. Vedia , 288 F. App'x 941 , 947-48 (5th Cir. 2008) (explaining that it was obvious error when an expert witness used "drug profiling" to imply that most drug couriers know that they are carrying drugs in their vehicle and thus the defendant likely knew he had drugs in his vehicle). The government, while not conceding error, at least ultimately recognized that the testimony was problematic in stating during oral argument that it "would not advise [its] prosecutors to do this again." Today's opinion, and the body of precedent it relies on that is filled with similar warnings, should add an exclamation point to that advice.

Even with egregious error, however, a defendant's failure to seek correction at trial means he must show "a reasonable probability that his trial would have come out differently." Gonzalez-Rodriguez , 621 F.3d at 367 . This third requirement of plain-error review has prevented defendants from obtaining relief in most of the other cases involving improper drug profiling testimony. See, e.g., Gonzalez-Rodriguez , 621 F.3d at 367-68 (refusing to reverse improper use of profiling evidence despite obvious error because the witness's "ultimate opinion, although improper, was unlikely to have swayed the jury's conclusion."); Vedia , 288 F. App'x at 947-48 (finding a witness's drug profiling testimony an "error that is clear under current law" but refusing to reverse because "the jury had sufficient evidence to convict Vedia of possession with intent to distribute cocaine"); United States v. Ramirez-Velasquez , 322 F.3d 868 , 879 (5th Cir. 2003) (holding a DEA Agent's profiling evidence was "improper" but not sufficient to vacate the conviction on plain error review because there were curative jury instructions and sufficient other evidence "from which the jury could infer [the defendant's] guilt.").

It does so once again. Although the DEA agent exceeded the bounds of opinion testimony in identifying Sosa as a drug trafficker, three co-conspirators also made that identification, and they are allowed to. None of the documentary evidence, which included money transfers and phone records, pinpointed Sosa's involvement, but it did provide some general corroboration of the conspiracy the cooperating witnesses described. Perhaps recognizing that the testimony of three co-conspirators is difficult to overcome especially when he bears the burden of showing prejudice, Sosa seeks to highlight inconsistencies among the three accounts. There are certainly some. But they are not so significant to convince us that the inadmissible profile testimony was likely the difference maker at trial. Nor did the prosecutor remind the jury of the agent's improper testimony during closing argument. Because Sosa is unable to prove prejudice, he is not entitled to a new trial.

B.

That same obstacle to plain error correction also prevents us from granting Sosa relief on his second claim-that the government improperly bolstered the cooperators' credibility-even though it too has merit. The government contends there was no improper bolstering, emphasizing that it can introduce evidence of a cooperator's plea agreement and let the jury know that truthful testimony is a requirement of *621 that agreement. See United States v. Edelman , 873 F.2d 791 , 795 (5th Cir. 1989) ("Admission of a plea agreement wherein the witness has agreed to testify truthfully or face prosecution for perjury is not impermissible bolstering of the witness."). But that is not the extent of Sosa's complaint. He invokes the principle that prosecutors are not permitted to bolster a witness's credibility by implying that the prosecutor, or even worse the neutral judge, has determined the testimony to be truthful. See United States v. Gracia, 522 F.3d 597 , 601 (5th Cir. 2008) ("[A] personal assertion by a prosecutor of a government witness's credibility is impermissible.").

That is what happened here. The most glaring example involved Sarmiento's testimony. The prosecutor did not just mention that the plea deal required Sarmiento to testify truthfully, but also elicited testimony that both the prosecutor and the judge had already determined that his testimony was truthful. This serious error occurred because unlike the usual situation when a cooperating codefendant is still pending sentencing when he testifies, Sarmiento already had been sentenced. This is how the prosecutor used the reduction for substantial assistance the court had already awarded Sarmiento:

Q. Okay. So the deal between you and I was tell the truth and you'll get a favorable sentencing recommendation; is that right?

A. Yes.

Q. And you told the truth and I recommended to Judge Hanen that you get a third off your sentence?

A. Yes.

Q. And Judge Hanen gave you 80 months?

A. Yes.

Q. And so you did get a reduction on your sentence?

A. Yes.

Q. And so you're now here in front of the jury and you're keeping up your end of the deal. And what was your end of the deal?

A. To cooperate.

Q. And to what?

A. Say the truth.

Q. Tell the truth. And no other promises, no other deals, no other hopes, other than please don't get me killed in prison?

A. Yes.

(emphasis added).

This problem repeated itself when Galvan testified. The prosecutor again highlighted that the same judge presiding over Sosa's trial had already given the witness a one-third reduction from his Guidelines range when the only basis for doing so was Galvan's honoring his agreement to "testify truthfully."

The final cooperating witness to testify, Luera, had not yet been sentenced. So although the prosecutor discussed with him ad nauseam the need to tell the truth, any bolstering did not involve this greater danger of telling the jury that the one neutral party in the trial-the judge-had endorsed the credibility of a key government witness.

Because the questioning of at least Sarmiento and Galvan put the prosecutor's and judge's stamp of approval on their credibility, improper bolstering occurred. See United States v. Young, 470 U.S. 1 , 18-19, 105 S.Ct. 1038 , 84 L.Ed.2d 1 (1985) (explaining that a problem with bolstering is that a prosecutor's opinion of a witness's credibility "carries with it the imprimatur of the Government and may induce the jury to trust the Government's judgment rather than its own view of the evidence"); United States v. Gallardo-Trapero , 185 F.3d 307 , 320 (5th Cir. 1999) (noting it is improper for the government to "invoke[ ] the aegis of governmental imprimatur");

*622 United States v. Brooks, 508 F.3d 1205 , 1210 (9th Cir. 2007) (recognizing improper vouching when a witness "agreed with the prosecutor's suggestion that the prosecutors and the judge in other trials believed he testified truthfully, and that if he had 'lied, given false testimony at those trials,' he would not have been given a reduction in his sentence"); United States v. Francis , 170 F.3d 546 , 550-51 (6th Cir. 1999) (finding improper bolstering when the prosecutor elicited testimony suggesting that a plea agreement only materialized after the government determined the witness's statements to be truthful).

But we need not decide whether this error was an obvious one under our case law, which addresses improper bolstering of law enforcement witnesses (not co-conspirators) during closing argument (not witness examination). 2 See, e.g. , United States v. Aguilar , 645 F.3d 319 , 323-26 (5th Cir. 2011) ; United States v. Raney , 633 F.3d 385 , 395 (5th Cir. 2011) (per curiam); Gracia , 522 F.3d at 600 . Even assuming Sosa clears the first two hurdles of plain error review, he again cannot show that the improper bolstering affected his substantial rights. As we have discussed, the testimony of three cooperating witnesses presented a strong case of guilt even in light of some inconsistencies in their testimony. In one sense this may make the bolstering more prejudicial because it resulted in the endorsed credibility of the key testimony. On the other hand, the reciprocal corroboration provided by three conspirators identifying Sosa as a fellow participant in their crime was likely a much greater factor in the jury's acceptance of that testimony than was the bolstering. The prejudice inquiry thus is a close call, but that does not get Sosa where he needs to be as he has the burden of demonstrating a reasonable likelihood that the bolstering influenced the verdict. Molina-Martinez v. United States , --- U.S. ----, 136 S.Ct. 1338 , 1348-49, 194 L.Ed.2d 444 (2016) (recognizing that the defendant bears the burden of establishing prejudice on plain error review and noting that a "remand for retrial" poses greater "difficulties" than a remand for resentencing). Additional obstacles to clearing that hurdle are that the bolstering was not repeated during closing argument when it can be most potent and the jury was instructed that it "alone must evaluate witness credibility." Ramirez-Velasquez , 322 F.3d at 875 (citing this jury instruction and whether the bolstering "permeate[d] the entire atmosphere of the trial" as factors in assessing the impact of bolstering (quoting United States v. Iredia , 866 F.2d 114 , 117 (5th Cir. 1989) ) ). We conclude Sosa has not established that the bolstering substantially affected the outcome.

C.

Sosa also argues that a Confrontation Clause violation occurred when Agent Bradford mentioned a tip implicating Patricia Sosa. Bradford was asked how the DEA determined that Patricia Sosa was involved in dealing drugs from Mexico. He explained that he was contacted by the Houston DEA office, which had received an automated alert that other agents were investigating the name "Patricia Sosa." Bradford stated that the two offices determined they were investigating the same Patricia Sosa and that an undercover agent had confirmed that Patricia was looking for couriers to transport drugs from Mexico into the Houston area.

The Confrontation Clause bars testimonial statements offered against the defendant when there has been no opportunity *623 for cross-examination. Crawford v. Washington , 541 U.S. 36 , 68-69, 124 S.Ct. 1354 , 158 L.Ed.2d 177 (2004). But as every student taking Evidence quickly learns, there is not a hearsay or confrontation problem when the evidence is not being used for the truth of the matter asserted. Williams v. Illinois , 567 U.S. 50 , 70, 132 S.Ct. 2221 , 183 L.Ed.2d 89 (2012). One example of that is when a law enforcement tip is introduced to explain why an officer took investigatory steps. See United States v. Smith , 822 F.3d 755 , 761-62 (5th Cir 2016) ; United States v. Carrillo , 20 F.3d 617 , 619-20 (5th Cir. 1994). As we recently recognized, courts must be vigilant in ensuring that these attempts to "explain the officer's actions" with out-of-court statements do not allow the backdoor introduction of highly inculpatory statements that the jury may also consider for their truth. United States v. Kizzee , 877 F.3d 650 , 659 (5th Cir. 2017) (explaining that absent a defense challenge to the adequacy of the investigation, "there is a questionable need for presenting [these] out-of-court statements because the additional context is often unnecessary, and such statements can be highly prejudicial"). That danger is greatest for statements that implicate the defendant. Id . at 659 (citing Taylor v. Cain , 545 F.3d 327 , 335 (5th Cir. 2008) ; United States v. Johnston , 127 F.3d 380 , 394 (5th Cir. 1997) ). The tip Bradford recounted did not. It only mentioned Patricia Sosa, who was an acknowledged participant in the drug trafficking. Because the jury's considering the tip for its truth rather than as an explanation of the agent's actions would not have told it anything it did not already know, at a minimum it was not obvious that this statement was offered for its truth. As a result, there was no clear Confrontation Clause violation.

D.

Sosa argues that even if he does not prevail on his individual claims, a new trial is warranted due to cumulative error. This doctrine provides that "an aggregation of non-reversible errors (i.e., plain errors failing to necessitate reversal and harmless errors)" can merit a new trial. United States v. Delgado , 672 F.3d 320 , 343-44 (5th Cir. 2012) (en banc). We do not find that this case is one of the "rare instances" when reversal for cumulative error is appropriate. Id. at 344 .

* * *

Today's outcome is the same as many of our prior decisions addressing drug profiling testimony and bolstering of witnesses: we find that the government engaged in misconduct but nonetheless conclude the defendant cannot meet the heavy burden of obtaining reversal for error he did not object to during trial. If the ultimate end of prosecution is securing convictions, it may not be surprising that this trend has not deterred these improper trial tactics. Of course, winning is not supposed to be a prosecutor's lodestar. Striking "hard blows" but not "foul ones" in pursuit of justice is. Berger v. United States , 295 U.S. 78 , 88, 55 S.Ct. 629 , 79 L.Ed. 1314 (1935). Fidelity to that higher calling would prevent us from seeing these errors yet again.

JENNIFER WALKER ELROD, Circuit Judge, concurring:

I write separately to emphasize that I do not condone the prosecutorial misconduct here and, as the Supreme Court has suggested we should, continue to discourage it. See Donnelly v. DeChristoforo , 416 U.S. 637 , 648 n.23, 94 S.Ct. 1868 , 40 L.Ed.2d 431 (1974) (Rehnquist, J.) ("We do not, by this decision, in any way condone prosecutorial misconduct, and we believe that trial courts, by admonition and instruction, and appellate courts, by proper exercise of their supervisory power, will continue to discourage it.").

*624 The prosecutor's decision to invoke the imprimatur of the court in its efforts to persuade was no doubt improper. We have repeatedly admonished the government about the impropriety of the type of prosecutorial comments and profiling testimony used here. Yet it appears the government has again ignored our previous admonishments. See United States v. Bowen, 818 F.3d 179 , 191 (5th Cir. 2016) ("We have repeatedly chastised federal prosecutors for making improper remarks in closing arguments ...." (quoting United States v. Rodriguez-Lopez , 756 F.3d 422 , 433 (5th Cir. 2014) ) ); United States v. Aguilar , 645 F.3d 319 , 324 (5th Cir. 2011) (recounting that "the government has been cautioned repeatedly by this court against making such arguments, yet we continue to face them on appeal" (quoting United States v. Raney , 633 F.3d 385 , 395 (5th Cir. 2011) ) ); United States v. McCann , 613 F.3d 486 , 496 (5th Cir. 2010) (holding that it was improper to make a "largely emotional appeal" to jurors that the officers should be believed because of their status as officers); United States v. Pittman , 401 F. App'x 895 , 899 (5th Cir. 2010) (concluding that there was improper bolstering where the prosecutor stated that the agents were "just doing their job" and had no reason to lie); United States v. Gonzalez-Rodriguez , 621 F.3d 354 , 366 (5th Cir. 2010) (describing improper profile testimony as "over the line"); United States v. Ramirez-Velasquez , 322 F.3d 868 , 879 (5th Cir. 2003) ("The government goes too far in soliciting the functional equivalent of an opinion ...."); United States v. Gallardo-Trapero , 185 F.3d 307 , 319-21 (5th Cir. 1999) (stating that it is improper to invoke "the aegis of a governmental imprimatur" to bolster witness credibility).

I am reminded of Justice Jackson's famous charge to the federal prosecutor that "[o]nly by extreme care can we protect the spirit as well as the letter of our civil liberties, and to do so is a responsibility of the federal prosecutor." Robert H. Jackson, U.S. Att'y Gen., The Federal Prosecutor, Address at the Second Annual Conference of United States Attorneys in Washington, D.C. (Apr. 1, 1940), in 24 J. Am. Jud. Soc'y 18, June 1940, at 18-20. "[T]he United States Attorney is 'the representative not of an ordinary party to a controversy, but of a sovereignty whose obligation to govern impartially is as compelling as its obligation to govern at all; and whose interest, therefore, in a criminal prosecution is not that it shall win a case, but that justice shall be done.' " Strickler v. Greene , 527 U.S. 263 , 281, 119 S.Ct. 1936 , 144 L.Ed.2d 286 (1999) (quoting Berger v. United States , 295 U.S. 78 , 88, 55 S.Ct. 629 , 79 L.Ed. 1314 (1935) ).