Dogfood record
How AgenticSEO used AgenticSEO to improve its own AI visibility
Last updated:
Short answer
AgenticSEO ran its own audit-to-publish loop on agenticseo.live: fix crawler access, make every marketing route server-rendered, publish entity-clear JSON-LD and an llms.txt index, then write answer-ready pages for the questions buyers actually ask. This page records what was changed, in what order, and which signals moved.
1. Audit crawler access and answer readiness
The first run scored every public route for crawler access (GPTBot, OAI-SearchBot, Google-Extended, ClaudeBot, PerplexityBot), title and description quality, schema coverage, and whether each page answered one question in a self-contained passage. Marketing routes passed on access but failed on answer readiness: the pages sold the product without answering the buyer's question on the page.
2. Make the marketing site server-rendered
The app is a single-page React build, so early crawls of the marketing routes returned an empty shell. A build-time prerender step now renders each public route to static HTML, so assistants and search crawlers that do not execute JavaScript read the same body copy a visitor sees. This was the single largest change and it applies to every public route, not just the homepage.
3. Publish entity-clear structured data
One Organization and one WebSite node sitewide, with AgenticSEO and AgenticSEO.live declared as alternate names so the brand is not parsed as the generic category phrase. Per-route Article, FAQPage, Dataset, and BreadcrumbList nodes were added where the page type warranted them, each self-referencing its own canonical URL.
4. Add the machine-readable index — with realistic expectations
An llms.txt index at the site root lists the pages worth reading and the buyer prompts the site is built to answer. Adoption of llms.txt by major assistants is still unproven, so it is treated as the least proven layer of the stack: cheap to publish, never the plan on its own.
5. Write one page per buyer question
Instead of more product copy, the loop drafted answer pages and guides that resolve a single question in the first paragraph — what agentic SEO is, how to rank in ChatGPT, how to measure Share of AI Voice — plus commercial-intent comparison pages for the vendors buyers evaluate against us, each with vendor source links and a visible last-updated date.
6. Re-measure, then repeat
Tracked prompts run on a schedule against ChatGPT and Google Gemini, and Google Search Console is connected so classical and AI signals sit in one queue. Search Console impressions grew roughly tenfold over the period covered here (from the high-70s to the low-800s in a comparable window), and several glossary and guide pages now rank on page one for their exact terms. Citation share on broad category prompts remains the hardest metric to move and is still a work in progress.
What we did not claim
No ranking guarantees, and no attribution of every impression gain to a single change — the category itself grew over the same period. The honest reading is that the technical layers removed avoidable losses (unreadable pages, ambiguous entity signals, missing answers), and the content layer earned the gains that followed.
Questions
Did AgenticSEO use its own product to do this?
Yes. The audit, the prompt tracking, the generated JSON-LD and llms.txt, the FAQ and page drafts, and the publish step are the same features customers use. The only work done by hand was editorial review before publishing.
What made the biggest difference?
Server-rendered HTML for every public route. Structured data, an llms.txt index, and answer-ready copy only help if the crawler can read the page at all — a JavaScript-only marketing site fails that test first.
How long did it take to see movement?
Search Console impressions began moving within a few weeks of the prerender and content changes. Assistant citation share moves more slowly, because assistants need to recrawl and re-embed the pages before answers change.
Can the same loop run on my site?
Yes — start with the free three-page check. It audits crawler access and answer readiness, and returns ready-to-publish assets so you can compare the before and after on your own domain.
Run the same loop on your site
Three pages, about a minute, no sign-up and no card.
Try the free check