A class action tests whether AppLovin's AI improved itself on the schedule it sold

The class action filed against AppLovin▲+0.49% — AppLovin, up 0.49 percent today this month does not ask whether its AI works. It asks whether the AI improved on the schedule the shares were priced on. In *Talbot v. AppLovin*, No. 26-cv-10584, in the Northern District of California↗, buyers between February 12 and August 5, 2026, allege the ad-technology company overstated how consistently its AI models improve, oversold the "virtuous cycle" that improvement was meant to compound, and hid delays in the generative AI video tool built to feed them ¹²³.

The pitch was a schedule of improvement

AppLovin's software decides which ads appear inside mobile-phone games and, more and more, for online shops ⁴. Its AXON models pair the ad with the user, and the outcome of each placement trains the next decision ⁵. The resulting gain in advertisers' return on spend, what the industry calls uplift, is what pulls budget in ⁵.

When the stock slid in February, management called it "a real disconnect between market sentiment and the reality of our business↗," and the class period opened that morning ¹. On May 6, chief executive Adam Foroughi called a recent model update "quite substantial," with April "bigger than any quarter that we had" ⁵.

The bottleneck moved from the model to the ads themselves

The loop has two feeds, and only one is automatic. Outcomes collect themselves; the creative the model places had to scale beyond human hands, which is what the generative video tool was for. Management has called creative "the biggest hurdle" for e-commerce ⁴, and on August 5 the tool was "still [a] work in progress" ¹.

The stall went public across two days

On July 13 a Bank of America note called the June 22 opening of AppLovin Ads to all advertisers a "muted GA start"; the stock fell 12.6% ². On August 5, revenue of $1.92 billion missed the $1.94 billion consensus, and management said its "pace of meaningful model improvement was lighter than normal during the quarter" ¹². The shares fell a further 19.66% the next day ¹². Between the two drops the company lost more than $44 billion of market value ⁵. Growth is still running above 50% a year, and it has slowed every quarter since mid-2025 ⁶.

Advertisers stayed; the story did not

Jefferies' October advertiser survey↗ expects AppLovin's share of e-commerce advertising budgets to reach 8.1% this quarter ⁷. Respondents also expect to give it 8.3% of budgets in 2027, up from 7.8% this year, the largest expected gain of any platform polled ⁷. The same survey found generative AI used for video creative by only 37% of advertisers, down from half ⁷. Jefferies still cut its 2027 e-commerce revenue forecast to $1.7 billion from $2.8 billion ⁷.

Citi reads the same adoption data as acceleration, with United States store growth quickening to 2.7% a week, and kept a $600 target on October 12↗ ⁸. Bank of America, whose July note the suit counts as its first corrective disclosure, cut its target to $325 on October 9 ¹²⁹.

The next hard fact arrives on November 4

The complaint identifies more than $109.1 million of insider share-sale proceeds↗ during the class period ¹. The question underneath is timing: when during the spring did management know the uplift had thinned? No public figure measures model uplift, so the answer has to come from reported results. The next set arrives on November 4, guided to just over $2 billion of revenue ¹⁰, with the November 16 deadline to seek lead-plaintiff status two weeks behind it ².

Management says the next step up in model performance "landed just after quarter end," and the business is "back on the trajectory we expect"↗ ⁴. If the step-up is real, that report is where it shows. Advertisers' budgets already show the capability is real ⁷; the improvement schedule is the part only reported numbers can settle.