EverQuote's AI Era Has a Price Tag—and the Market Is Haggling

Record quarter, muted reflex

EverQuote’s third quarter was, by the numbers, the best in its history. Revenue rose 25% to $218.1 million, adjusted EBITDA rose 41% to $35.5 million, and the home vertical grew 70% to $27.7 million. Joseph Sanborn did not bury the lede: “We grew revenue 25% and Adjusted EBITDA 41% year-over-year.” — Joseph Sanborn, Chief Financial Officer · 2026-11-02 Yet the call was not really about Q3. It was about what EverQuote is willing to spend to remain relevant as shopping moves from web pages to personal AI agents. That is the tension: the income statement is accelerating, but the Q4 guide implies a sharp deceleration to 12% revenue growth and a sequential VMM decline as the company funds agentic traffic channels. Investors are being asked to underwrite a new distribution channel before it has revenue.

AI is not new; the bill is

Jayme Mendal has been talking about an AI era in P&C insurance distribution for several quarters. In February he told investors “we’re not a software business, right? We’re a data-powered 2-sided marketplace.” — Jayme Mendal, Chief Executive Officer · 2026-02-24 On the current call he said the company is “at the forefront, leading P&C insurance distribution into the AI era.” — Jayme Mendal, Chief Executive Officer · 2026-11-02 The consistency is notable: the AI narrative is not a fresh pivot. What is fresh is the artifact. Last week EverQuote launched Luva, an AI-native insurance shopping platform, and signed a multi-year partnership with Waniwani to build and host AI distribution infrastructure. Management also disclosed apps across Meta, OpenAI and Anthropic, subject to platform approval. The keyword agentic commerce has moved from white-paper speculation to a named product with a P&L cost.

The old moat is still the moat

The bull case is that carriers do not want their rates scraped indiscriminately by large language models. Mendal argued that insurers are wary of continuous reshopping, rate transparency and adverse selection. EverQuote’s 15 years of carrier integrations and proprietary shopping data give it a role as a bridge. That is why the stock has not collapsed on the AI threat; it is also why the current margin investment is not obviously reckless. But the AI agent opportunity is not yet a revenue line. Mendal was candid: less than 1% of traffic comes from LLMs, and even less from agentic shoppers.

Right now it’s negligible. It’s immaterial in terms of its scale. We get less than 1% of our traffic from LLMs and even less from agentic shoppers. But it is likely to expect it will -- reasonable to expect it will grow over time.

Soft market, hard choice

The carrier backdrop remains the quieter engine. Sanborn said insurers are in the early stages of a soft market cycle that analysts expect to last into 2029, with combined ratios low enough to keep growth budgets open. The home vertical is the clearest proof: 70% growth in Q3 after 30%-35% in the first half. Sanborn wisely warned not to model 70%: “I would say 70% was very impressive, and I would not put that into your model, Ralph.” — Joseph Sanborn, Chief Financial Officer · 2026-11-02 Home is still about 13% of revenue, up from roughly 10%-11% a year ago, so the mix shift is real but not yet a second leg.

The harder choice is VMM. In August, management said new traffic channels were “progressing as planned” — Jayme Mendal, Chief Executive Officer · 2026-08-03 and did not expect them to “materially impact one way or the other, the kind of medium-term VMM operating point.” — Jayme Mendal, Chief Executive Officer · 2026-08-03 On this call, Q4 VMM is guided down sequentially because of investment in agentic LLM channels and Luva. Sanborn called Q4 “a reasonable proxy for next year as sort of a starting point.” — Joseph Sanborn, Chief Financial Officer · 2026-11-02 That is a subtle but important shift: the company is no longer treating agentic traffic as an experiment that sits outside the model. It is now inside the model, and it is dilutive near term.

The valuation is the hinge

This is where the story becomes more than an AI narrative. EverQuote generated record revenue and EBITDA, has $229.1 million of cash and no debt, and just authorized a new $50 million buyback. Yet the stock trades at roughly 7x trailing net income, down more than 60% year over year, and about 1.1x revenue. Price to net income was 7.3x, down 61% year over year. That multiple says the market does not believe the current earnings stream is durable, or does not believe the AI investment will earn a return. The fundamentals through June support the former worry less than the latter: Total revenue reached $195 million in the June quarter, up 25% year over year, before Q3’s $218.1 million. Operating margin had recovered to 12.6%, and free cash flow margin was 8.8% in the June quarter. The cash engine is real; the question is whether it is being spent on a moat or on a science project.

What to watch

Three things will determine whether this quarter was a launchpad or a peak. First, Luva adoption: management is watching “arrival specifically through those channels” — Jayme Mendal, Chief Executive Officer · 2026-11-02 and carrier panel build-out, not just app downloads. Second, VMM stabilization in 2027; if agentic channels remain a drag beyond the first part of next year, the Q4 proxy comment becomes a guidance cut in slow motion. Third, carrier response: if large insurers impose guardrails or refuse to expose rates to AI agents, EverQuote’s bridge becomes less valuable. The unit economics of agentic referrals are still unproven; management hopes greater conversion offsets lower traffic quality, but that is a hypothesis, not a reported metric.

For now, EverQuote is doing what few small-cap marketplaces can: growing 25%, generating cash, and buying back stock while investing in a channel that could redefine its industry. The risk is that the market prices the near-term margin dip before the long-term option value shows up. That is the bet.