# AEO applied to our own site: what the audit found *AEO · The Litebox team · Last updated: September, 2026 · 6 min read* We ran an AEO audit on litebox.ai itself and held the results to the bar we'd hold a client to, which surfaced a rendering bug we've fixed, a live mismatch in llms.txt, and one fact about the search environment needing no fix. ## Key Facts - litebox.ai's stats module used to render as 0x to crawlers, because the figures loaded client-side and never reached the raw HTML. Re-checked September 15, 2026: raw HTML now carries a real number. - litebox.ai/llms.txt opens with "the full-stack go-to-market (GTM) agency for DevTools startups," which differs from the canonical description in the Brand Voice Guide, "GTM agency for B2B software." Confirmed live, September 15, 2026. - 378 of litebox.ai's 537 target keywords returned a Google AI Overview, counted during the Keyword Universe v2 rebuild, August 5, 2026. ## Why we audited our own site We hold every finding to the same method-and-date standard a client's claim gets, so we applied it to litebox.ai first, following the same structural approach [How to structure a page so AI engines cite it](https://litebox.ai/blog/answer-engine-optimization-website-structure-h1) sets out for a citable page. We ran our AEO checklist and kept what qualified against our own internal claims registry, the same one that governs every client-facing figure we publish. A claim needs a dated method behind it or it doesn't ship, and that rule applied to our own findings the same way it applies to a client's. Three findings cleared that bar. One had already closed by the time we re-checked it. One is still open. One needed no fix at all. ## We found our stats rendering as "0x" to crawlers, and we fixed it We first checked what a crawler receives from /work/copilotkit, a case-study page that carries a stat callout, on July 29, 2026. The raw HTML returned 0x where a real number belongs, because the figure loaded through client-side JavaScript. A fetcher that reads raw HTML without executing scripts never saw it. [Google's crawling docs](https://developers.google.com/search/docs/crawling-indexing/javascript/javascript-seo-basics) confirm the mechanism. Content generated by client-side JavaScript can be absent from the initial HTML response, appearing only once a renderer executes the script. Google eventually runs that renderer for its own crawler, on its own schedule. No AI platform's documentation we found confirms the same for GPTBot, ClaudeBot, or the others, so we treat a client-side number as invisible to them by default. We re-ran the identical check today, curl with a GPTBot user agent against /work/copilotkit. The raw HTML now carries a real number, at the same byte count as a normal browser fetch. The fix shipped sometime between our last check and today. Nobody on the team logged the exact ship date, so the next fix on our own site gets a date attached the moment it goes out. ## llms.txt still describes us differently from how we describe ourselves elsewhere A live check of litebox.ai/llms.txt, confirmed September 15, 2026, shows it opening on "the full-stack go-to-market (GTM) agency for DevTools startups," narrower than the Brand Voice Guide's canonical description, "GTM agency for B2B software." The finding narrowed since we first caught it. In early August, Litebox described itself three separate ways across the site and llms.txt combined: "agency" on the site, plus "go-to-market engineering team" and "high-tech startups" inside llms.txt. Today llms.txt is down to one description, and that one description still doesn't match the canonical wording. It leads with DevTools alone, and the site also serves B2B SaaS demand that framing leaves out, the exact segment a query like "GTM agency for a B2B SaaS company" would need llms.txt to name. Aligning llms.txt to the canonical description is an open item on the Website Brief's Workstream 2, item 5. The [llms.txt format](https://llmtxt.info/llms-txt-format/) has exactly one required element, an H1 with the site's name. The summary right after it is where a site states what it is, in the compact form [the original llms.txt proposal](https://llmstxt.org/) exists to serve. An AI system can't process a whole site the way a reader can, so that opening summary carries more weight than one line usually would. ## Google already shows an AI Overview for most of our target keywords A rebuild of our own Keyword Universe, counted on August 5, 2026, found a Google AI Overview on 378 of our 537 target keywords. That's the SERP we're already competing on. This one carries no fix, closed or open. An AI Overview appearing above a result still leaves that result on the page, competing with a synthesized answer for the same attention it used to have alone. ## What we've done, and what's still open The rendering fix already shipped. A case-study page that used to hand a crawler 0x now hands it a real number. We can confirm it's live today; we can't confirm exactly when it shipped. The llms.txt mismatch has a next step we can name precisely. We'll rewrite the opening summary to match the canonical description, then run the same kind of raw-fetch check to confirm a crawler receives that exact text. This one hasn't shipped yet. The AI Overview measurement stays as recorded, with no action attached to it. ## What this means for your own site We hadn't checked either of these two things on litebox.ai before this audit: what a crawler's raw fetch actually contains, separate from what a browser renders for a person, and whether our own self-description agrees with itself everywhere it appears. Neither shows up in a routine site review, because both require looking at the site the way a fetcher actually reads it, which a routine review doesn't do. The other half of the picture is what AI systems actually say about you once they can read you correctly. [AI visibility audit](https://litebox.ai/blog/ai-visibility-audit) covers that side: building a prompt list, running it across the platforms your buyers use, and logging what comes back. ## FAQ ### What is an AEO audit? An AEO audit checks whether AI systems can access, extract, and cite your content accurately. It covers crawler access, whether your content survives client-side rendering, and whether your site describes itself the same way everywhere. We ran one on litebox.ai and found a rendering bug, an entity mismatch, and one fact worth knowing about the SERP. ### Why isn't my site showing up in ChatGPT? A crawler blocked in robots.txt is one common technical cause. Content that loads through client-side JavaScript and never reaches the raw HTML a fetcher requests is another. Our own stats module hit exactly that second case, and a human visitor looking at the same page had no way to notice. ### How to get your content cited by AI? Confirm an AI system can actually read the content first. Check what your raw HTML returns to a crawler user agent, separate from what a browser shows you, and keep your site's self-description consistent everywhere it appears, including llms.txt. ### How to improve brand visibility in AI search? Fix the structural issues first, rendering gaps and crawler blocks, and handle messaging after. A well-worded page that never reaches the crawler as text earns nothing, and an inconsistent self-description undercuts a page that does reach it. The llms.txt mismatch is still open, and we said so anyway. If your own site needs this same check, running it and fixing what it finds is work we partner on with DevTools teams through our [growth and AEO program](https://litebox.ai/services/growth). [All articles](https://litebox.ai/blog) · [Talk to the team](https://litebox.ai/contact-us) --- HTML version: https://litebox.ai/blog/aeo-on-our-own-site · All pages: [sitemap.xml](https://litebox.ai/sitemap.xml) · Agent guide: [llms.txt](https://litebox.ai/llms.txt)