AI-native SEO: why your traffic is shifting (and what to ship instead).

For two decades, SEO was a single game: rank on a page of ten blue links. That game is over. AI assistants, ChatGPT, Claude, Perplexity, Google’s AI Overview, answer the question before the user clicks. Your traffic chart is not lying to you. The funnel moved. Many small-cap teams have seen meaningful drops in organic traffic over the last two years as AI-overview surfaces take share. The pages still rank. The clicks still arrive on the first page of Google. But the user no longer scrolls past the AI overview. They get their answer, they make their decision, they never see the brand. Worse, the traffic that does land often converts at a lower rate than it used to, because the assistant has already told the visitor whether they should buy. This is not a temporary dislocation. The shift is structural. Google’s AI Overview now surfaces on a meaningful share of informational queries (multiple third-party studies put it in the same range) and that share has trended upward year over year, while Perplexity, ChatGPT, and Claude eat most of the long tail. By the time a user ends up on your page, they have already been counseled. Your job is no longer to rank. Your job is to be the source the AI quotes. Here is what changes when the assistant becomes the funnel. Keyword density and hero images stop mattering, and internal linking matters less than it used to. What matters is structure, citation, and entity recognition. The assistants are looking for pages that read like reference material, clear thesis up top, structured headings, schema markup, named entities, sourced claims. They quote individual sentences, not paragraphs, so each sentence has to be self-contained. What we ship for clients on this is a specific series of moves, in order. First, we rewrite the top ten landing pages so each opens with a one-sentence thesis that an AI can quote. Second, we add JSON-LD Article and Organization schema to every template, with sameAs links to the LinkedIn, X, and Crunchbase profiles that anchor entity recognition. Third, we open robots.txt to the AI crawlers that matter, GPTBot, ClaudeBot, and PerplexityBot, and we allow Google-Extended so your content stays eligible for Google’s AI features. Yes, you want them reading you, not blocked. Fourth, we audit alt text on every image and rewrite it to be descriptive, not decorative. Fifth, we restructure your pages to cite primary sources rather than only linking internally, because the assistants favor content that is sourced and citation-dense. The hardest part is mindset. For two decades, the SEO playbook treated outbound links as leakage. Citing your sources is now an asset, not a risk. The pages the assistants quote most are the pages that read like research notes, high citation density, clear authorship, named entities, transparent sources. If your page reads like a brochure, the assistants will skip it. There is a counter-move that matters specifically for small-cap public companies. Your IR pages, your news releases, your management bios, these are the entity surfaces the AI uses to identify your company. Make them clean. Use consistent name spelling across every page. Use the same management bio everywhere. Add Person and Organization schema to your team page. Ensure your ticker, exchange, and CUSIP appear in structured data, not just in copy. The assistants are doing entity resolution on you whether you participate or not, and the easier you make their job, the more accurate the answer they give about your company when an investor asks. Two more moves that matter and almost no one is doing yet. Add a /llms.txt file to your root, a plain-text manifest that lists your most important pages, in the llms.txt format (proposed by Answer.AI in 2024) that a growing list of AI vendors now read. And version your major pages with a visible “last updated” date in structured data; the assistants weight recency heavily, and a stale 2022 page is invisible to them no matter how well it ranks. Six months in, the dashboard tells a different story than it used to, and a better one. Your overall traffic is lower than it was two years ago, but the traffic you do receive tends to convert at a higher rate, because the visitors who arrive have already been pre-qualified by the assistant. Your name appears in AI-generated summaries when potential clients ask about your category. Your publicly available pages and on-record materials are citable in the major models, so when an investor asks, the assistant has clean source material to point at. Investors who run the AI lookup before a meeting come in informed. This is not a content marketing problem. It is an information architecture problem. The teams that fix it first will spend the next decade owning the answer in their category. The teams that wait will spend the next decade explaining to their boards why traffic keeps falling. What we ship. An AI-SEO audit, a rewrite of your highest-traffic pages in citation-friendly format, structured data on every template, robots.txt and sitemap aligned to current assistant policies, and a content cadence built for assistants. Fixed-scope engagement. Documented and owned by your team.