Chapter 4

Search Disruption

The tailwind and the threat are the same force. Financial shopping keeps migrating online — NerdWallet's opportunity — but the free organic-search traffic that built its audience is eroding as Google's AI overviews and large language models answer money questions without a click. Management's language moved from "not much of an impact" in mid-2024 to "steep organic search declines" for full-year 2025 [1]. LLM referrals are a higher-converting but still tiny replacement channel.

The online shift, and who else is in the lane

The secular tailwind is genuine: consumers and small businesses increasingly research and buy financial products online, and NerdWallet is a scaled comparison marketplace positioned to capture that shift. Management frames the runway as long — the company is "still at an early stage" and has "only scratched the surface of our addressable market" [2].

The same migration draws well-capitalized competition. The FY2025 10-K names online rivals — Bankrate, Credit Karma, LendingTree, Zillow — and, pointedly, warns that "major search engines and content aggregators, may be able to leverage their existing products and services or access to data to our disadvantage" [3]. The tailwind is real, but it is shared — and the largest party in the lane, Google, is also the single largest source of NerdWallet's traffic.

One gatekeeper, and a risk factor that keeps hardening

Since its 2021 listing, NerdWallet has disclosed that it is "dependent on internet search engines, particularly Google, to direct traffic to our websites and refer new users to our platform" [4]. What has changed is how the filing describes the AI threat to that channel, tightened almost every year.

The FY2021 10-K flagged Google dependence with no mention of AI [5]. The FY2022 10-K added a conditional line: "The introduction of AI-assisted technologies may impact the relevance of search engines… if we are not able to adapt our content strategy to a changing web search landscape" [6]. By FY2025 the same passage reads as an event in progress: "The introduction and acceptance of AI-assisted technologies could further impact search engine relevance, causing declines in our ranking and decreased platform traffic," alongside the flat statement that "Past changes have caused declines in traffic and user growth" [7]. A distinct FY2025 risk factor now treats "generative and agentic AI technologies" as their own category of exposure [8]. The word added between FY2022 and FY2025 — "acceptance" — is the tell: the risk moved from possible to adopted.

From "not much of an impact" to "steep declines"

The earnings calls trace the same shift in real time, and faster than the annual filing. In mid-2024 Chen was dismissive of AI overviews, arguing they were "not much of an impact," that ChatGPT web visits had "actually fell 12% month-over-month," and that Google was if anything gaining share [9]. One quarter later the tone changed: shopping traffic "took a turn for the worse" in August and September, and "the renewed push by search engines to incorporate their own answers directly into search results, like the AI overviews mentioned, is affecting us" [10].

Sources: quarterly earnings-call transcripts — Q2 2024 [11], Q3 2024 [12], Q4 2024 [13], Q1 2025 [14], Q2 2025 [15], Q3 2025 [16], Q4 2025 [17], Q1 2026 [18].

By the middle of 2025 the framing had settled. Organic search "remains quite challenging," AI overviews had rolled out to "a much larger range of queries," and — because those overviews answer educational questions on the results page without a click — monthly active users "have been affected more than revenue" [19]. That split matters: NerdWallet's less-commercial "learn" traffic is more exposed to AI answers than its high-intent "shop" traffic, which is why revenue has held up better than the audience metric. Summing the year, Chen described "steep organic search declines" driven by consumers turning to AI overviews and LLMs, and guided that organic search would "remain under pressure" [20].

The substitution, and where it leads

This erosion is the reason performance marketing — bought traffic — climbed to roughly half of revenue over the same window; the arithmetic sits in Bought Traffic. The free top-of-funnel that once fed the marketplace at near-zero marginal cost is being replaced by traffic the company pays for click by click. Whether that substitution is a temporary bridge or a permanent margin drag turns on one thing the filings cannot yet answer: whether a new organic channel grows fast enough to refill the funnel the old one is losing.

A new organic channel, still small

There is a candidate, and management leans on it. NerdWallet says its trusted content has made it "the most cited source in our competitive set" in LLM answers [21] and that it is "dominant when it comes to LLM share in financial services or money questions" on third-party data [22]. The referral quality looks better than legacy search: "the conversion rates on that LLM referral traffic are much higher and growing rapidly" [23], and those users arrive with "significantly higher intent to transact" [24].

The honest limit is that the channel is not yet material. On the most recent call Chen placed it plainly: LLM referral traffic is "just a very small piece of our overall pie right now from a revenue perspective" [25]. The company discloses no figure to size it, so the replacement channel is real and qualitatively superior per click, but too small to weigh against the organic declines it is meant to offset.

Whether the marketplace layer is defensible

The deeper question is whether AI reaches past the traffic layer and disintermediates the marketplace itself. Management's answer rests on two frictions specific to financial services. The first is licensing and compliance: quoting insurance, credit, mortgages or investments requires licensed, deterministic, compliant outputs "not probabilistic answers," which any intermediary — human or agentic — must satisfy. The second is institutional buy-in: a two-sided marketplace only works if lenders and insurers choose to participate, and "an insurance company can refuse to quote an AI agent that is shopping around by inserting a multifactor authentication step" [26].

That is a coherent defense of the transaction layer, and it is management's, not an independent finding. The counter is that it protects the wrong step. The licensing moat guards the point of purchase, while the value NerdWallet actually captures begins upstream, at discovery — the same top-of-funnel that AI overviews are already thinning. If large model providers strike direct data or referral arrangements with financial institutions, the discovery step could be intermediated even where the regulated transaction cannot. The read that fits the evidence: the organic-search headwind is structural and confirmed by the company's own escalating language, the paid channel is covering it today at a real but bounded cost, and the LLM offset is promising but unproven at any scale that shows in revenue. What would move that read is a disclosed, material LLM-referral contribution — or a genuine stabilization in organic search, which management hoped for in early 2025 [27] and had walked back by year-end.