Optimizing SaaS for AI Discovery: A Four-Phase Overhaul

A founder claims that half of their traffic will originate from AI channels by 2026. This prompts a detailed technical overhaul of public-facing SaaS surfaces for AI discovery.

A founder reports that by 2026, half of their organic traffic will originate from AI discovery channels like ChatGPT. This claim drove a weekend-long technical overhaul of a SaaS product's public surfaces, moving from a 2018-era SEO baseline to a stack designed for AI crawlers and large language models (LLMs). The project, detailed in a dev.to post, outlines a four-phase process, though only the initial phase and its core technical changes are fully described.

The premise is that traditional SEO, while a solid baseline, no longer suffices for a landscape where AI models increasingly mediate user discovery. The founder's approach focuses on explicit signals for AI crawlers, structured data, and performance metrics, aiming to ensure content visibility in a post-search-engine world.

Structured Data and Localization

The initial SEO setup relied on a PageMeta wrapper for basic title, description, canonical, OG, and Twitter Card tags, covering 52 of 139 pages. This pattern remained unchanged. The significant addition was comprehensive structured data using JSON-LD from schema.org. The founder implemented Organization and WebSite schemas on the homepage, Article and Breadcrumb for blog posts, Person for profiles, and FAQPage for FAQ sections. This provides explicit, machine-readable context about the content, which is critical for LLMs synthesizing information.

Simultaneously, the founder added hreflang="es-mx" and x-default attributes to every page render. This signals regional targeting and a default language version to crawlers, a feature entirely absent in the prior setup. Accurate localization signals help AI models serve content to the correct linguistic and geographic audiences.

Dynamic Sitemaps and Crawler Policies

The original setup used a static sitemap.xml listing only 13 landing page URLs, maintained manually. This was replaced with a sitemap index that references the static landing page list and three new dynamic sitemaps. These dynamic maps, served by the backend, cover blog posts, user profiles, and salary reports. Dynamic sitemaps ensure that AI crawlers have an up-to-date and comprehensive map of all public content, scaling with the product's growth.

Crucially, the founder introduced an explicit llms.txt file and updated robots.txt policies. While the prior robots.txt had a default Allow for User-agent: *, the new configuration includes explicit Allow blocks for specific AI crawlers: GPTBot, ClaudeBot, anthropic-ai, PerplexityBot, Google-Extended, and CCBot. The llms.txt file specifically points to the sitemap, defines crawling guidelines for LLMs, and marks private surfaces as disallowed. This granular control dictates how AI models interact with and index the SaaS product's content.

Indexing and Performance Monitoring

Previously, new content required manual submission to Search Console for indexing. The updated process integrates IndexNow notifications, automatically alerting search engines like Bing, Yandex, Naver, Seznam, and Cloudflare when a new post is published. This accelerates content discovery by AI models and traditional search engines alike, reducing the time between publication and indexing.

Real User Monitoring (RUM) was also implemented, a feature entirely missing before the overhaul. Web Vitals metrics (CLS, INP, LCP, FCP, TTFB) are now sampled at 10% in production and sent to CloudWatch. These metrics provide insights into actual user experience, which increasingly influences how AI models and search engines rank content. Performance is no longer just a user experience factor; it is a discovery factor.

Pre-rendered HTML for JS-less Crawlers

Many AI crawlers and social media previewers do not execute JavaScript. The original setup provided static HTML defaults for social previewers, but lacked specific handling for JS-less bots on dynamic routes. The founder implemented a post-compilation script that emits an index.html file for each of the eight static landing page routes. This ensures that AI bots without a JavaScript engine can still access and parse the specific meta-information for each route, without resorting to complex tools like Puppeteer in the build process.

What We'd Change

The founder's

The investor read

The shift towards AI-mediated discovery represents a significant evolution in organic acquisition channels, impacting how early-stage SaaS companies acquire customers. This founder's tactical overhaul signals a growing recognition among operators that traditional SEO alone is insufficient. Investors should assess a company's 'AI surface' readiness, looking for explicit strategies around structured data, LLM-specific crawler policies, and performance metrics. Companies that proactively optimize for these channels may gain a competitive edge in customer acquisition costs and brand visibility. This also indicates a potential for new tooling and services in the 'AI SEO' space, creating opportunities for investment in infrastructure and analytics platforms.