Takeaway: A product label reaches every buyer, but a webpage reaches only the buyers who visit it. That distinction decided LoSavio v. Tesla, Inc., No. 25-7927, 2026 WL 3093826 (9th Cir. Oct. 16, 2026), in which the Ninth Circuit vacated the certification of two Rule 23(b)(3) damages classes and a Rule 23(b)(2) injunctive-relief class of California purchasers of Tesla’s “Full Self-Driving” (FSD) package. The panel’s finding of legal error in the district court’s inference of exposure to the challenged statement provides a framework for contesting classwide-exposure presumptions in online-marketing cases.

Plaintiff Thomas LoSavio alleged that Tesla violated California’s consumer-protection statutes and common law by misrepresenting its cars’ hardware capabilities in connection with the FSD packages it sold. The centerpiece was the “Hardware Statement” on Tesla’s Autopilot webpage, which, according to the district court’s certification order, represented that Tesla vehicles had the hardware needed for full self-driving capability at a safety level substantially greater than a human driver’s capability. The trial court certified two (b)(3) classes and a (b)(2) class, but the Ninth Circuit granted Tesla’s Rule 23(f) petition to review the certification ruling.

To satisfy predominance for his California fraud-based claims, LoSavio had to show classwide exposure to the Hardware Statement. LoSavio, 2026 WL 309826, at *1 (citing Mazza v. American Honda Motor Co., 666 F.3d 581 (9th Cir. 2012)). Plaintiffs often try to carry that burden through a presumption, and under Walker v. Life Insurance Co. of the Southwest, 953 F.3d 624 (9th Cir. 2020), the question is whether the defendant “so pervasively disseminated material misrepresentations that all plaintiffs must have been exposed to them.” Id. at *1. The Ninth Circuit identified only two scenarios in which it has recognized that presumption: (1) an “extensive and long-term [fraudulent] advertising campaign”; and (2) a representation on a product label or another medium typically provided to all customers, such as the pre-application policy illustrations in Walker. Id.

The district court had placed this case in the label category. It emphasized Tesla’s unusual direct-to-consumer marketing and the website’s central role, finding it “highly likely” that buyers of an expensive package like FSD would visit the site. Id. at *2. The Court of Appeals disagreed, explaining that a label supports an inference of exposure because every purchaser “necessarily has the allegedly misleading statement in front of him.” Id. Here, the district court had recognized there was “no specific requirement forcing [FSD purchasers] through the Autopilot page to complete their purchase.” Id.

That recognition proved decisive. The inference required two assumptions: that buyers reached the Autopilot page, and that they saw the Hardware Statement once there. Even if the statement’s prominence supported the second assumption, the first defeated classwide exposure. The Ninth Circuit relied on Sevidal v. Target Corp., 117 Cal. Rptr. 3d 66 (Ct. App. 2010), which refused to infer exposure where customers had to click an optional tab to see the statement. Id.

The Court of Appeals rejected the plaintiff’s supporting evidence on two grounds. First, website-traffic data merely “consistent with” classwide exposure does not show that all class members “must have been” exposed to that data. Id. Second, other statements reinforcing the Hardware Statement’s message neither appeared on a label nor reached every prospective buyer, and they did not, alone or together, amount to an “extensive and long-term advertising campaign” under Mazza. Id.

On a related issue, the district court concluded that Tesla’s statute-of-limitations defenses turned on an objective question: whether Tesla’s misrepresentations stopped consumers from discovering their claims. But the Ninth Circuit disagreed, finding that inquiry turns on each buyer’s exposure to the Autopilot website—where Tesla attributed FSD delays to regulatory hurdles—and potentially to other statements. Id. While individualized limitations questions will not always bar certification, the individualized exposure questions “pervade[d]” the limitations analysis here. Id. (citing Cameron v. E.M. Adams & Co., 547 F.2d 473 (9th Cir. 1976)).

Finally, because the injunctive-relief class had been defined as a subset of the (b)(3) classes, the Ninth Circuit held that it “rises and falls” with them and vacated it as well. Id. at *3.

The Losario decision highlights the importance of developing a record showing how customers actually buy a product; which webpages are optional; and whether the challenged statement appears in any required step. And it supports limiting inferred classwide exposure to long-term advertising campaigns or other systematic disclosures to all class members, undercutting arguments for implying such exposure from a company’s marketing model or website. Evidence that exposure was “highly likely,” or that website traffic data was “consistent with” exposure, should not suffice to show that every class member necessarily saw the statement.