Opinion by JUDGE WEBB
¶ 1 A jury convicted Sandra L. Jacobson of vehicular homicide, driving under the influence (DUI), and other related charges, all arising from a collision between her truck and a taxi cab. Two passengers in the taxi were killed. On appeal, Jacobson primarily contends the trial court erred when it refused her counsel's request to poll the jury concerning exposure to mid-trial publicity that included inadmissible, prejudicial information.
¶ 2 We conclude that the court abused its discretion when it declined to poll the jury, instead relying solely on its assumption that the jurors had followed repeated admonitions to avoid news reports. Because we also conclude that this error was not harmless beyond a reasonable doubt, we reverse Jacobson's convictions and remand the case for a new trial on all charges.
I. Background
¶ 3 On the fourth day of trial, while the prosecution was still presenting its case, defense counsel alerted the court to the possibility that Channel 4, the local affiliate of a national television network, would air on its evening news program a report containing information prejudicial to the defense. When the day's proceedings ended, the court gave the jury an extended admonition to avoid "newscasts" and "newspaper sites."
*1134¶ 4 The next morning, defense counsel told the court that a report about the case had appeared the preceding night on Channel 4 and the station's website. The court and defense counsel agreed that the broadcast and the text which appeared alongside video content on the website included the following:
• references to Jacobson's prior convictions, including a DUI;
• references to her prior traffic accidents resulting in injury;
• an allegation that she had been driving with a suspended license when her truck struck the taxi cab;
• a suggestion that she was facing at least two additional trials on other matters;
• an interview with someone involved in a previous accident with her who said that she had been driving with a suspended license in that accident and offered that she should be incarcerated forever; and
• reference to her other driving violations.
¶ 5 The record lacks any direct information about the telecast or the website. But the prosecutor did not take issue with anything said between the court and defense counsel about the content, nor does the Attorney General do so on appeal. When defense counsel asked the court to inquire of the jury regarding exposure to this material, the prosecution did not take a position.
¶ 6 The court responded that the information was "absolutely not anything [that was] going to come up in [the] trial" and that the "timing was unfortunate." The court referred to the expanded admonition given when the prior day's proceedings had ended, and mentioned that several jurors had looked serious and nodded affirmatively during that admonition. Then the court "decline[d] to inquire of [the jury] any further about any specific program" because it decided to "take them at their word," saying that telling the jury about the broadcast would be like "dangling a cookie under a 2-year old's nose." Finally, it noted that defense counsel had made the necessary record.
II. Preservation and Standard of Review
¶ 7 The Attorney General contends Jacobson did not preserve this issue because her counsel acquiesced in the trial court's decision to rely on the admonitions, thereby inviting any error. The record belies this contention.
¶ 8 A defendant preserves an issue when the defendant timely requested relief at trial on the same ground raised on appeal. See People v. Ujaama, 2012 COA 36, ¶ 37, 302 P.3d 296. Jacobson's counsel timely asked the court to poll the jury based on publicity of extraneous, prejudicial information. This request is the basis for her argument on appeal. As well, the court acknowledged that the necessary record had been made. See People v. Coughlin, 304 P.3d 575, 582 (Colo. App. 2011) (objection "sufficient to preserve an issue for appeal if the objecting attorney presents arguments or utilizes language that alerts the trial court to the impending error").
¶ 9 Still, citing People v. Cook, 22 P.3d 947 (Colo. App. 2000), the Attorney General argues that because defense counsel implicitly agreed with the court's reliance on its admonitions, Jacobson cannot now argue that the court erred in refusing to poll the jury. In Cook, the division held that because defense counsel told the trial court additional jury instructions would be "fine," the defendant had explicitly agreed with the trial court's decision to rely on these instructions rather than polling. Id. at 949. But Jacobson's counsel never explicitly agreed with the trial court's decision. And counsel's failure to take exception to the ruling, after having urged polling, is not implicit agreement. Cf . C.R.C.P. 46.
¶ 10 Because a trial court has broad discretion to decide whether a media report prejudiced the defendant's right to a fair trial, review is for an abuse of discretion. Harper v. People, 817 P.2d 77, 83-84 (Colo. 1991). A trial court abuses its discretion "when it misconstrues or misapplies the law," People v. Sieck, 2014 COA 23, ¶ 5, 351 P.3d 502 ; "fail[s] to exercise discretion," People v. Darlington, 105 P.3d 230, 232 (Colo. 2005) ; or rules in a manner "manifestly against the *1135weight of evidence," Hytken v. Wake, 68 P.3d 508, 510 (Colo. App. 2002).
¶ 11 When a trial court abuses its discretion by declining to question the jury about exposure to extraneous, prejudicial information, the error is subject to constitutional harmless error review. Dunlap v. People, 173 P.3d 1054, 1091 (Colo. 2007) (concluding that the trial court's failure to poll the jury was harmless beyond a reasonable doubt); see also United States v. Thompson, 908 F.2d 648, 652-53 (10th Cir. 1990) (applying constitutional harmless error review).
III. Law
¶ 12 "The due process clauses of the United States and Colorado Constitutions guarantee every criminal defendant the right to a trial by an impartial jury." People ex rel. Faulk v. Dist. Court, 673 P.2d 998, 1000 (Colo. 1983). This guarantee is satisfied by " 'a jury capable and willing to decide the case solely on the evidence before it, and a trial judge ever watchful to prevent prejudicial occurrences and to determine the effect of such occurrences when they happen.' " People v. Dahl, 160 P.3d 301, 304 (Colo. App. 2007) (quoting Smith v. Phillips, 455 U.S. 209, 217, 102 S.Ct. 940, 71 L.Ed.2d 78 (1982) ); see also State v. Bey, 112 N.J. 45, 548 A.2d 846, 861 (1988) (Impartiality requires a jury to decide the case " 'based on evidence received in open court, not from outside sources.' " (quoting Sheppard v. Maxwell, 384 U.S. 333, 351, 86 S.Ct. 1507, 16 L.Ed.2d 600 (1966) )).
¶ 13 If jurors may have been exposed to extraneous, inherently prejudicial material, to protect this constitutional right an appellate court must be able to assess actual exposure from the record. See Gov't of Virgin Islands v. Dowling, 814 F.2d 134, 140 (3d Cir. 1987) ("Because the trial court did not conduct a voir dire, and because we cannot speculate what the jurors' responses would have been to an appropriate inquiry," the trial judge "erred when he failed to develop a record sufficient to permit evaluation of the potential prejudice to the defendant.") (cited with approval in Harper, 817 P.2d at 84 ); Thompson, 908 F.2d at 651-52 ("[T]he reason for the lack of specific evidence concerning the juror's connection with the newspaper article was the trial court's failure to question the jurors about their exposure.") (cited with approval in Harper, 817 P.2d at 84 ). Polling the jury may show that "no exposure to the publicity ha[s] occurred at all." Bey, 548 A.2d at 870 (cited with approval in Harper, 817 P.2d at 84 ). But absent polling, "we cannot assume that this was the case, or, if some exposure did occur, that it had no effect on the juror's impartiality." Id .
¶ 14 Harper is the most recent Colorado Supreme Court decision on possible jury bias from mid-trial exposure to extraneous, prejudicial information. The supreme court adopted a test to "determine the appropriate manner of evaluating the prejudicial effect of media reports when the issue is presented during trial." 817 P.2d at 82.
¶ 15 In Harper, the defendant faced charges of sexual assault on a child. After the jury had been empanelled, the court instructed the jurors to avoid any outside reading on the case. On the second day of trial, the local paper published an article on an interior page about the defendant's prior conviction of sexual assault on a different child victim. Although unrelated, this assault and the charged offense had allegedly occurred during the same time period. The next morning, defense counsel asked the court to question the jury about the article. The court refused, explaining that the defendant had not shown any juror had read the article.
¶ 16 The supreme court acknowledged the trial court's reliance on People v. Holmes, 191 Colo. 477, 481, 553 P.2d 786, 789 (1976), holding that because of the presumption that the jury had followed the trial court's instructions, polling the jury was unnecessary. Harper, 817 P.2d at 82. But the Harper court identified two flaws with this presumption: jurors may inadvertently learn about newspaper articles, and rules prohibiting counsel's communication with jurors during trial hinder the ability to acquire evidence of prejudice. Id. The court concluded that deciding prejudice on the basis of such a presumption does not adequately protect a defendant's right to an impartial jury. Id. at 83.
*1136¶ 17 Departing from Holmes, the supreme court adopted the following test to address possible prejudice from mid-trial publicity:
"The simple three-step process is, first, to determine whether the coverage has a potential for unfair prejudice, second, to canvass the jury to find out if they have learned of the potentially prejudicial publicity and, third, to examine individually exposed jurors-outside the presence of the other jurors-to ascertain how much they know of the distracting publicity and what effect, if any, it has had on the juror's ability to decide the case fairly."
Id . at 83 (quoting United States v. Gaggi , 811 F.2d 47, 51 (2d Cir. 1987) ). Still, the court cautioned that this test must be "conducted in the framework of the facts of each case, and the trial court has broad discretion in deciding the ultimate issue of whether the media reports prejudiced the defendant's right to a fair trial." Id . at 83-84.
¶ 18 When implementing the first step, the trial court should "focus principally upon whether the content of the media report is inherently prejudicial." Id. at 84. Then the court should consider "[o]ther relevant factors" in weighing the potential for unfair prejudice: whether the material includes information inadmissible at trial or not presented to the jury; how closely this information relates to the principal issue in the trial; the timing of the publicity; and the likelihood of jury exposure. Id.
¶ 19 As to the last factor, the court may consider the effectiveness of any admonitions to avoid news of the defendant or trial "in light of the nature and manner of dissemination of the news reports." Id. Yet, "admonitions, alone, do[ ] not sufficiently neutralize news reports" that may reasonably have been discovered by jurors. Id. And, in any event, "[d]oubt about the existence of prejudice should be resolved by proceeding to step two and polling the jurors as a group." Id.
¶ 20 Applying this new test, the supreme court first noted the article's "highly prejudicial contents." Id . at 86. The court pointed out that the article discussed a prior conviction for the same type of offense that was being tried, and both crimes allegedly occurred during the same time period. Id. at 85. Also, because the evidence against the defendant was equivocal, the court recognized a greater potential that the information about his prior conviction could lead the jurors to convict because he had a propensity to sexually assault children. Id. (noting that evidence of prior criminal convictions is generally inadmissible due to its highly prejudicial effect).
¶ 21 In weighing the remaining factors, the court explained that the defendant did not "assert mere vague speculation of possible prejudice." Id. Instead, he "identified a particular newspaper account having a specific and certain prejudicial content," which "appeared during the second day of trial in the local newspaper of a relatively small city." Id. The court acknowledged the trial court's admonitions, but noted that because the article was short and located on an interior page, the "obscurity of the article may have led a juror to read it inadvertently." Id. at 86.
¶ 22 The supreme court explained that the " 'more serious the potential jury contamination, especially where alleged extrinsic evidence is involved, the heavier the burden to investigate.' " Id. (quoting United States v. Harris, 908 F.2d 728, 734 (11th Cir. 1990) ). Because the record established the potentially prejudicial effect of the newspaper article, the court concluded that "some direct inquiry of the jury was necessary." Harper, 817 P.2d at 86. It explained that "[f]raming this inquiry in general terms would minimize any implication that an article containing information adverse to the defendant had been published." Id. And such "collective questioning, in combination with the court's admonition, could have sufficed to establish absence of prior or prospective juror exposure to the article." Id.
¶ 23 Divisions of this court and courts in other jurisdictions applying this three-step test to potentially prejudicial mid-trial publicity further illuminate the respective functions of trial and appellate courts. See People v. Bondurant, 2012 COA 50, ¶¶ 81-85, 296 P.3d 200 ; People v. Ferrero, 874 P.2d 468, 473-74 (Colo. App. 1993) ; see also United States v. Aragon, 962 F.2d 439, 443-47 (5th Cir. 1992) ;
*1137State v. Holly, 145 N.M. 513, 201 P.3d 844, 848-51 (2009).
¶ 24 First, the trial court must determine the potential for unfair prejudice from the publicity, starting with whether the extraneous information is inherently prejudicial and would not be admissible at trial. Bondurant, ¶ 82 ; Ferrero, 874 P.2d at 473. In each of the other Colorado cases upholding refusal to poll the jury, the extraneous information was found not to have been inherently prejudicial. People v. Gardenhire, 903 P.2d 1165, 1171 (Colo. App. 1995) ("As there is no indication that [the media report] was inherently prejudicial, the court's inquiry of the jury panel as a group was sufficient."); Ferrero, 874 P.2d at 473 (publication of jurors' names in the local newspaper was not inherently prejudicial); see also People v. Harlan, 8 P.3d 448, 505 (Colo. 2000) (newspaper article about death penalty case not connected with the defendant's trial was not prejudicial), overruled on other grounds by People v. Miller, 113 P.3d 743 (Colo. 2005). Neither Harper nor any of these cases suggests that when possible jury exposure to extraneous information has been called to the trial court's attention, the court may decline to address inherently prejudicial content and potential for unfair prejudice.
¶ 25 Second, the trial court must decide whether to poll the jury. Compare Bondurant, ¶¶ 83-85 (concluding that because trial court found news story irrelevant to the defendant's case, refusing to poll was not an abuse of discretion), with People v. Harrison, 58 P.3d 1103, 1110 (Colo. App. 2002) (applying Harper in the context of jury misconduct during deliberations and holding that because trial court "implicitly found" the conduct "had the potential to be prejudicial," it properly followed the Harper steps), abrogation recognized by Dunlap, 173 P.3d at 1091.
¶ 26 But can an appellate court review a trial court's step-two decision not to poll, if at step one the trial court failed to make express findings about whether the extraneous information was inherently prejudicial or, looking at other factors, created a potential for unfair prejudice? Harper shows how the appellate court can do so.
¶ 27 In Harper, the supreme court made its own determination that the newspaper article "had great potential for unfair prejudice," which the trial court had not addressed. 817 P.2d at 85. Likewise, appellate courts in other jurisdictions have reviewed the extraneous material and made their own determinations of inherent prejudice. See Aragon, 962 F.2d at 444 (an "after-the-fact analysis" was performed to "determine whether the news material was innately prejudicial"); United States v. Herring, 568 F.2d 1099, 1105 (5th Cir. 1978) (considering the jury's "inferences which would naturally flow" from publicity about death threats against the prosecution's key witness) (cited with approval in Harper, 817 P.2d at 84 ); Brown v. State, 601 P.2d 221, 232 (Alaska 1979) (where publicity was not part of the record, relying on descriptions in transcripts to determine "what was prejudicial about" the material) (cited with approval in Harper, 817 P.2d at 84 ).
¶ 28 Also consistent with Harper, 817 P.2d at 86, several cases in other jurisdictions have recognized that because admonitions alone do not neutralize the potential for unfair prejudice, a trial court should not rely on admonitions as the sole basis for declining to poll the jury about its exposure to extraneous, inherently prejudicial material. See, e.g., Aragon, 962 F.2d at 445 (because of the "conspicuousness of the news account and its prejudicial content, notwithstanding the court's general instruction to the jury, ... there was a substantial probability that the publicity reached the jurors present"); United States v. Lord, 565 F.2d 831, 838 (2d Cir. 1977) (trial court "erred in relying solely on repetitive admonitions" when the circumstances created a strong likelihood of jury exposure to prejudicial material); Mares v. United States, 383 F.2d 805, 807 (10th Cir. 1967) (trial court should have polled the jury, despite a "strong and comprehensive" admonishment).
¶ 29 In sum, when a trial court fails to determine whether extraneous material is inherently prejudicial or the jury's possible exposure to this material created a potential for unfair prejudice, an appellate court may *1138do so. If the appellate court deems the content inherently prejudicial, the appellate court should consider other factors and decide whether possible exposure to the material created a potential for unfair prejudice. In that event, if the trial court's only basis for refusing to poll the jury involved the presumed efficacy of its admonitions, the appellate court should conclude that the trial court has abused its discretion. See Harper, 817 P.2d at 84.
IV. Application
¶ 30 After Jacobson's counsel alerted the trial court to potentially prejudicial publicity, the court and counsel generally agreed on the information that had been published. At that point, under Harper, id., the court should have made a finding whether the material was inherently prejudicial. Although the court agreed that the publicity included information inadmissible at trial, it did not make such a finding. Nor did it make an express finding on the potential for unfair prejudice. Even if such findings are matters of discretion, these failures constituted an abuse of discretion. Darlington, 105 P.3d at 232.
¶ 31 Without these findings, we are hampered in deciding the ultimate question whether the trial court abused its discretion in refusing to poll the jury. To do so, we follow Harper and start by considering whether the extraneous material was inherently prejudicial. Next, we consider whether the material and the context created a potential for unfair prejudice. Then we address whether the trial court should have polled the jury.
¶ 32 We conclude that the material was inherently prejudicial. As in Harper, it included not only information about Jacobson's criminal record, but also the prior DUI offenses related closely to the charges being tried and Jacobson's affirmative defense. See generally Stull v. People, 140 Colo. 278, 284, 344 P.2d 455, 458 (1959) ("[E]vidence of similar acts has inhering in it damning innuendo likely to beget prejudice in the minds of jurors, and ... tends to inject collateral issues into a criminal case which are not unlikely to confuse and lead astray the jury."), superseded by rule as stated in Yusem v. People, 210 P.3d 458 (Colo. 2009).
¶ 33 Based on the remaining Harper factors, we also conclude that the material had a potential for unfair prejudice:
• The publicity included information that had not been and would not be admitted at trial.
• Because this information closely related to Jacobson's affirmative defense that she became intoxicated after the accident-the primary contested issue at trial-it could undermine this defense by "caus[ing] jurors to believe that [she] had a propensity to commit such a crime." Harper, 817 P.2d at 85.
• As in Harper, the evidence was equivocal because, among other reasons, Jacobson was prepared to, and ultimately did, testify that she had consumed an alcoholic beverage only after the accident, and the officers who first contacted her did not detect intoxication.
• Because the publicity occurred during the prosecution's case-after defense counsel had referred to the affirmative defense in opening statement but before Jacobson testified-the timing increased the danger that some jurors might prejudge her defense.
¶ 34 And we conclude that the circumstances created a "reasonable possibility" of jury exposure. Id. at 85. To begin, the information was aired by the local affiliate of a national broadcast television network and also published online. True, the trial court instructed the jurors to avoid all news sources, including those on the Internet. The Attorney General correctly points out that this instruction was broader than the admonition given in Harper . Yet, the trial court did not instruct the jury to entirely avoid the Internet. The ubiquitous nature of news on the Internet carries at least as great a risk of inadvertent exposure as publication on an interior page of the newspaper article in Harper . See United States v. Fumo, 655 F.3d 288, 331 (3d Cir. 2011) (Nygaard, J., concurring in part and dissenting in part) ("The availability of the Internet and the abiding presence of social networking now *1139dwarf the previously held concern that a juror may be exposed to a newspaper article or television program."). As noted in Harper, juror exposure to extraneous material can be inadvertent. But even if these circumstances balance each other out, Harper commands that doubt be resolved by polling the jury. 817 P.2d at 84.
¶ 35 For these reasons, we conclude that the trial court abused its discretion in declining to poll the jury, notwithstanding its diligence in admonishing the jury during voir dire and at the end of each day's proceedings to avoid all news sources.See R.J.Z. v. People, 104 P.3d 278, 280 (Colo. App. 2004) (trial court's "determination that these factors warranted denial of the petition reflects a misapplication of certain of the factors and, as to other factors, is manifestly against the weight of evidence").
¶ 36 The Attorney General's citation to United States v. Elfgeeh, 515 F.3d 100, 128-31 (2d Cir. 2008), does not require a different result. In that case as here, neither defense attorney expressly mentioned the three-step test, and one defense attorney voiced concern about calling the jurors' attention to the extraneous information. But unlike here, the Elfgeeh court emphasized that one of the two defendants did not join in the request to poll the jury. Id. at 131.
¶ 37 Citing Dunlap, 173 P.3d at 1091-92, the Attorney General also argues that because the record does not show jury exposure, polling was unnecessary. Although in Dunlap the court applied the Harper test, Dunlap involved allegations of juror misconduct during deliberations, not exposure to mid-trial publicity. And because the alleged misconduct occurred among alternates while the regular jurors deliberated in a separate room, the court concluded that "[n]othing in the record indicate[d] that the regular jurors were exposed to information regarding the [prior] conviction or sentence," and there was "no reasonable probability that [the defendant] was prejudiced." Id. at 1091. But here, we have discerned a reasonable possibility that the jurors may have been exposed to the extraneous information.
¶ 38 Finally, the Attorney General argues that because the prosecution presented overwhelming evidence of guilt, any error was harmless beyond a reasonable doubt. This argument is unpersuasive, for three reasons.
¶ 39 First, the Attorney General bears the burden of proving that the error was harmless beyond a reasonable doubt, and has not done so here. See, e.g., People v. Harris, 43 P.3d 221, 230 (Colo. 2002). Second, through her testimony, Jacobson raised the affirmative defense recognized in section 42-4-1301(2)(a), C.R.S. 2014 (it is an affirmative defense "if a defendant presents some credible evidence, that the defendant consumed alcohol between the time that the defendant stopped driving and the time that testing occurred"). Third, "[w]e do not reweigh the evidence or assess witnesses' credibility on appeal because the jury is the sole judge of witness credibility." People v. Griffiths, 251 P.3d 462, 465 (Colo. App. 2010).
V. Conclusion
¶ 40 The judgment of conviction is reversed, and the case is remanded for a new trial on all charges. Having so concluded, we need not address Jacobson's other contentions, which are unlikely to arise on retrial.
Navarro and Kapelke
Sitting by assignment of the Chief Justice under provisions of Colo. Const. art. VI, § 5 (3), and § 24-51-1105, C.R.S. 2014.