Grindr says 80% of its code is now AI-written while it price-tests a $350-a-month AI matchmaker, and investors still discount the stock

Grindr is the cleanest public test yet of whether an AI-native codebase can re-rate a stock, and the market's early answer is a shrug. The company says 80% of its code is now AI-written and that its roughly 95 technical staff do "roughly 350 people's worth of work" 1. Its Q2 2026 filing with the SEC, by contrast, shows revenue up 33% to $138 million at a 42% Adjusted EBITDA margin, with full-year guidance raised to roughly $540 million 2. As of the August 30 TechCrunch piece, the stock trades at roughly 11 times 2027 EBITDA, about a 35% discount to peers 1. That combination is the story: Grindr's filed numbers check out, its AI-productivity claims appear in no Grindr filing we reviewed, and the discount sits precisely on the unauditable half of the ledger.

What the filing proves, and what no filing can

The split is easy to see once you sort each number by whether an outsider could check it.

• Filed and checkable, per the Q2 2026 8-K: revenue of $138.1 million (+33% year over year), Adjusted EBITDA of $57.6 million at a 41.7% margin, net income of $17.7 million, half-year revenue of $268.1 million, and raised full-year guidance of about $540 million revenue and $232 million Adjusted EBITDA 2.
• Claimed and uncheckable, per CEO George Arison in the interview: about 80% of code AI-written, a 2.5x gain in engineering productivity over the past year, 94 or 95 people across technical roles doing roughly 350 people's worth of work, and retention on the unreleased EDGE tier "higher than anything we've had before" 1.

Run our audit math on the filed side and the guidance looks closer to banked than heroic. The $268.1 million first half is just under half of the $540 million guide, so the raised outlook implies a second half of about $272 million, only 1.4% above the first. The tripling from $195 million in 2022 1 to $540 million guided in 2026 works out to 2.8x, and the guided EBITDA margin is 43%. Divide the guided revenue across the roughly 95 technical roles the CEO cites and you get about $5.7 million of revenue per technical employee. One cross-check also reconciles: TechCrunch's 1.4 million paying users against the 15 million average monthly active users in the filing is 9.3%, matching the "9% of its user base" figure in the piece's framing.

None of that diligence touches the AI numbers, because no required disclosure standard covers "share of code written by AI." The market is being asked to price a productivity claim it cannot audit.

The $350 tier is the part analysts can actually count

EDGE, Grindr's AI-powered flagship tier, is the mechanism that converts unverifiable matching claims into a countable number. It sits above the existing XTRA tier at $23.99 a month and Unlimited at $44.99, and a Canadian test price worked out to roughly $350 to $375 a month in US dollars 1. That is 7.8x to 8.3x the Unlimited price, and the CEO says it was one elasticity test point among several, not a final price, with the launch set for late this year or early 2027.

Run the tier through the subscription math: each 1% of the 1.4 million paying base that converts at $350 a month is about 14,000 subscribers paying roughly $4,200 a year, or about $59 million annualized, which is close to 11% of guided 2026 revenue. Half a percent is about $29 million. TechCrunch frames the growth so far as coming almost entirely from getting existing users to pay more rather than user growth; on Arison's own figures, pay conversion has risen from under 6% to over 9% since 2022 and ARPU has nearly doubled 1. Subscriber counts get disclosed. Code provenance does not. TechCrunch reports Morgan Stanley's July upgrade to overweight citing EDGE and the telehealth push; the contrast with the AI-codebase number is our reading of that account 1.

Bias, verification risk, or both

Arison's bias case rests partly on anecdote: one investor once showed him a model with a line item that knocked 25% off fair value, a single anonymous account, not a market-wide finding 1. He also cites a consulting firm that declined the business and a bank that refused Grindr's money during the Silicon Valley Bank crisis. The same coverage complicates the story: Morgan Stanley, Goldman Sachs, and Raymond James raised price targets this year, the stock has climbed roughly a third over the six months through the August 30 piece, and the piece's own framing has Grindr trading at a premium multiple to Match Group even while carrying a 35% discount to "peers," which on our reading means whatever set prices the discount, the dating-app comp is not in it 1.

Our read is that both explanations survive the evidence. Category bias is documented anecdotally, and separately the market declines to pay for productivity claims with no audit path. The arithmetic of the gap is the stake: at roughly 11 times 2027 EBITDA against an implied peer multiple near 17x, closing the 35% discount at constant EBITDA is a 54% re-rating, larger than most product cycles deliver.

What would settle it is countable: EDGE's attach rate after launch, and whether Grindr's own filing language about "terraforming" into an AI-native organization 2 ever gets a number behind it. On this record, the discount is not only a verdict on a gay dating app. It is the price the market charges for the one AI claim nobody can check.