The SEO playbook that built entire companies from 2010 to 2023 — publish at volume, target long-tail keywords, accumulate backlinks — is no longer a growth strategy. It is a content treadmill with a collapsing floor.
The Numbers Are Brutal
Google AI Overviews launched at scale in mid-2024 and by Q1 2026 they appear on over 65% of all search queries in the U.S. The impact on organic traffic has been immediate and compounding. Semrush and Ahrefs both published data showing that informational content pages — the backbone of most startup content strategies — saw click-through rates drop between 30% and 60% year-over-year once AI Overviews claimed the top of the page.
For pure-play content publishers, the damage is existential. HubSpot reported a 50% decline in organic traffic in 2025. Dotdash Meredith, one of the largest content networks in the world, saw search-driven traffic fall more than 40%. These are not SEO-as-a-strategy failures — they are structural collapses caused by the search interface itself changing.
I've watched founders in my portfolio spend 18 months building content moats that are now worth a fraction of what they were. The keyword-first content strategy is not evolving — it is being made obsolete from the top down.
Why the Traditional Model Breaks Under AI Search
The ten-blue-links model was built on a simple value exchange: Google indexes your content, users click through, you get traffic. AI search breaks both sides of that exchange. The model now reads your content, synthesizes an answer, and serves it without the user ever visiting your site.
What makes this worse for startups is that AI Overviews tend to cite established sources — Forbes, NerdWallet, Wikipedia — for their sourcing signals. A domain-authority arms race that already disadvantaged smaller players has become even more lopsided. You can write a technically superior article and lose to a lower-quality piece from a domain that has been around since 2004.
The zero-click problem also compounds over time. Every query AI resolves perfectly is a query that no longer generates any organic traffic to anyone. The addressable market for traditional SEO shrinks with every model improvement. And the models are improving every quarter.
What Still Works: The Surviving Traffic Categories
- •Transactional intent: "Buy X near me," product comparison pages, and checkout-adjacent queries still drive clicks because AI Overviews cannot complete a purchase. These pages are safe — for now.
- •Branded search: When someone types your company name, AI does not intercept that query. Brand-building is now an SEO strategy, not just a marketing nice-to-have.
- •Proprietary data and original research: AI Overviews frequently cite primary sources for statistics. If you publish original surveys, proprietary benchmarks, or unique datasets, AI answers may link to you as the source — driving referral traffic from inside AI results.
- •Local and hyper-specific queries: Queries that require real-time, local, or highly contextual information remain click-through territory. AI Overviews perform poorly on "best plumber in [city]" and similar geo-specific searches.
- •High-complexity professional content: Nuanced legal, financial, or technical analysis where AI disclaims accuracy still sends users to human-authored sources. The deeper and more specialized the content, the lower the AI-answer substitution rate.
- •Community and UGC platforms: Reddit, Quora, and Stack Overflow are being cited inside AI Overviews at high rates because AI treats user-generated content as authentic social proof. Counterintuitively, community-driven content is outperforming polished editorial in AI search results.
The New Playbook: Optimize for AI Citation, Not AI Ranking
The goal is no longer to rank at position one. The goal is to be the source AI cites when it answers at position zero. These are meaningfully different objectives that require different strategies.
Getting cited by AI Overviews requires high domain authority, structured content with clear factual claims, and being the originating source for statistics and data rather than a secondary aggregator. Schema markup, E-E-A-T signals (experience, expertise, authoritativeness, trustworthiness), and HTTPS all remain relevant — but they feed the citation algorithm rather than the ranking algorithm.
Founders should also think about search diversification more aggressively. In 2026, Google is one of several AI-powered discovery surfaces. Perplexity, ChatGPT Search, Microsoft Copilot, and Meta AI are all processing search-style queries and surfacing content. Optimizing for one engine is now a concentration risk. Syndication, direct newsletter distribution, and podcast presence are all forms of search-engine diversification that are undervalued in this environment.
The startups that survive the AI search transition are not the ones who wrote more content — they are the ones who own proprietary data, strong brand signals, and direct audience relationships that no algorithm can intermediate.
Stay current with VC and startup trends at Value Add VC. Originally published in the Trace Cohen newsletter.