SourceCited / Method & sources
Method & sources
This isn't a repackaged blog post. It's the working method behind a network of content and tools sites that get cited by AI answer engines — written down so you can run it with Claude.
Everything here comes from repeatedly auditing real sites against a single evolving SEO/GEO methodology, shipping the fixes, and watching what actually moved citations. The five phases are the parts that held up. Treat every statistic as directional and re-verify quarterly — this field churns monthly.
The workflow is derived from first-hand practice, not theory: building AI-crawler
robots.txt and llms.txt files, emitting the full JSON-LD vocabulary
(Organization, Article, FAQPage, HowTo, Speakable, Dataset), front-loading stat-bearing answers,
hunting orphan pages and stray noindex, cleaning up legacy URLs with hard 410s to force
deindexing, and shipping dated "recent developments" sections the same week news breaks. The parts
that reliably moved the needle became the five phases.
The method aligns with the strongest published evidence. The founding Princeton/Georgia Tech GEO study (arXiv:2311.09735, KDD 2024) measured visibility lifts up to 40% from adding statistics, quotations, and cited sources. A University of Toronto analysis (arXiv:2509.08919, 2025) found AI search overwhelmingly favors earned, third-party media over brand-owned content. St. Gallen's measurement study (arXiv:2604.07585, 2026) showed ~65% of cited sources change day to day — visibility is a distribution, not a rank. And an MIT Sloan experiment (arXiv:2504.06435, 2025) found citations raise user trust even when they're wrong, which is why accuracy monitoring sits inside Phase 4. Vendor datasets (Ahrefs, Muck Rack, Semrush, Profound, Conductor, Moz, Seer) corroborate the earned-media and per-platform findings; treat their exact percentages as directional.
robots.txt.Statistics here are directional, often vendor-published, and date-stamped to mid-2026. They exist to size decisions, not to be quoted as universal benchmarks. Re-verify anything that will drive a budget or a roadmap.
SourceCited is written and maintained by Ned Walsch, an open-source intelligence researcher who builds and runs a small network of free, browser-based content and tools sites. He started SourceCited to write down the SEO/AEO/GEO method he actually uses to get those sites cited by AI answer engines — no sign-ups, no paywalls — and keeps it current as engines, crawlers, and the research change.
Everything here reflects hands-on practice on real sites rather than repackaged theory. When a tactic stops working or a figure goes stale, the relevant page gets updated instead of left to rot. If the site has helped you, you can buy him a coffee.