OPINION

PER CURIAM.

The State of California appeals a conditional grant of a writ of habeas corpus. The district court held that petitioner Rhonda Jean Dyas had been unconstitutionally shackled during her trial and that the shackling prejudiced her. The State does not dispute that the shackling was unconstitutional; it challenges only the holding that Dyas was prejudiced.

Dyas was convicted in California state court of first degree murder and robbery in 1991. The murder was carried out by Dyas’s husband and another man. Dyas prevented another person, Tina Boyd, from going into the room during the murder and may have been otherwise involved. Testimony at trial was contradictory and the jury took 3 days to deliberate following Dyas’s 5-day trial.

During trial, Dyas was kept in leg shackles in the courtroom. Her attorney had asked that she not be shackled in the courtroom. The trial judge denied the request, stating only that he did not believe the shackles would be “so visible that they come to the attention of the jury at all unless it’s brought to their attention” and that the nature of the case was such that he preferred the defendants to wear leg restraints. Dyas was also shackled while being led to and from the courtroom. Dyas’s attorney asked the jurors on voir dire if they would be able to disregard seeing the defendants led into the courtroom in shackles, to which the jurors responded that they would.

*588 The California Court of Appeal held that keeping Dyas shackled during trial was constitutional error. The appellate court found, however, that the error was harmless because there was no evidence to support Dyas’s argument that the jurors could see the shackles. Dyas subsequently filed a petition for habeas corpus in federal district court, and a magistrate judge ordered an evidentiary hearing to determine what the jurors actually saw.

Three jurors, one prospective juror who had been excused, members of Dyas’s defense team, and Dyas testified at the hearing. One of the jurors and the prospective juror had been able to see Dyas’s shackles from the jury box. Another juror recalled seeing Dyas in shackles in the hallway outside the courtroom. Dyas testified that the shackles caused her pain and inhibited her communication with her attorney, although she had not complained of these problems during the trial. The magistrate judge recommended that the district court grant a writ of habeas corpus, finding that there was prejudice because at least one juror was able to see the shackles and the evidence was not overwhelming. The district court adopted the recommendation and granted the writ.

When a defendant has been unconstitutionally shackled, the court must determine whether the defendant was prejudiced. See Ghent v. Woodford, 279 F.3d 1121, 1132(9th Cir.2002). Our conclusion that Dyas was prejudiced is virtually compelled by our decision in Rhoden v. Rowland, 172 F.3d 633 (9th Cir.1999). There we pointed out that shackling during trial carries a high risk of prejudice because it indicates that the court believes there is a “need to separate the defendant from the community at large, creating an inherent danger that a jury may form the impression that the defendant is dangerous or untrustworthy.” Rhoden, 172 F.3d at 636(citing Holbrook v. Flynn, 475 U.S. 560, 568-69, 106 S.Ct. 1340, 89 L.Ed.2d 525 (1986)). Prejudice is particularly likely here because at least one juror saw Dyas’s shackles during the trial from the jury box. Id. (“[Wjhen the defendant’s erroneous shackling has been visible to the jurors in the courtroom, we have found the shackling warranted habeas relief.”). It is likely that other jurors saw the shackles, but if even one juror is biased by the sight of the shackles, prejudice can result. See Parker v. Gladden, 385 U.S. 363, 366, 87 S.Ct. 468, 17 L.Ed.2d 420 (1966) (a defendant is “entitled to be tried by 12, not 9 or even 10, impartial and unprejudiced jurors”).

Two other factors increased the risk that Dyas was prejudiced by the juror or jurors having seen the shackles. Dyas was charged with a violent crime, increasing the risk that “the shackles essentially branded [her] as having a violent nature.” Rhoden, 172 F.3d at 637. Moreover, the evidence against Dyas was not overwhelming, a fact reflected in the length of the jury’s deliberations. Because the case was close, an otherwise marginal bias created by the shackles may have played a significant role in the jury’s decision. Id.

The State contends that there are, nonetheless, several aspects of Dyas’s trial that eliminated any reversible prejudice from the unconstitutional shackling. These arguments are not persuasive.

First, the State relies on the jurors’ own statements during voir dire that they would not be affected by seeing the defendant in shackles on the way to and from court. The questions and answers on voir dire, however, dealt with Dyas’s being shackled on her way to and from the courtroom. When the shackling continues during trial, it conveys a more continuous and stronger message to the jury that the defendant is dangerous. See Rhoden, 172 F.3d at 636. Moreover, the Supreme *589 Court has recognized that “little stock need be placed in jurors’ claims” that they will not be prejudiced. Holbrook v. Flynn, 475 U.S. 560, 570, 106 S.Ct. 1340, 89 L.Ed.2d 525 (1986). Where, as with visible shackling,

a practice may be inherently prejudicial, jurors will not necessarily be fully conscious of the effect it will have on their attitude toward the accused. This will be especially true when jurors are questioned at the very beginning of proceedings .... [Tjherefore, the question must be not whether jurors actually articulated a consciousness of some prejudicial effect, but rather whether an unacceptable risk is presented of impermissible factors coming into play.

Id. (quotation marks omitted). The analysis thus must focus on whether the risk was there, not whether the jurors could recognize the risk. In this case, the risk was present.

Second, the State argues that Dyas should not be able to claim that the jurors’ awareness of the shackles was prejudicial, because it was the defense attorney that brought the shackles to the jurors’ attention. Again, however, the defense attorney’s questions concerned the fact that Dyas would be transported to and from the courtroom in shackles, not that she would be shackled during trial. The defense attorney correctly anticipated that some jurors would become aware that Dyas was in custody and was being transported in shackles; the voir dire questions sought to minimize the effect of such awareness. But it was the shackling during trial that caused the prejudice, and Dyas’s attorney did not call the jury’s attention to that fact.

Lastly, the State argues that the sightings of Dyas in shackles in the hallway on the way to and from the courtroom were inconsequential. We do not reach this question because we find that the use of shackles in the courtroom itself carried a high enough risk of prejudice that the unconstitutional shackling cannot be considered harmless.

Because Dyas was unconstitutionally shackled during her trial, and the shackling was seen by one or more jurors to her prejudice, the shackles had “substantial and injurious effect” on the verdict and were not harmless error. Brecht v. Abrahamson, 507 U.S. 619, 637, 113 S.Ct. 1710, 123 L.Ed.2d 353 (1993). The judgment of the district court conditionally granting a writ of habeas corpus is

AFFIRMED.