Will fixing this hurt my human-facing site?

No. Server-rendering primary content and adding structured data improve human-facing UX and SEO too. Allowing AI agents in robots.txt is independent. Publishing /llms.txt is invisible to human users.

How do I verify it worked?

Run a scan against your domain. AgentSpeed reports each check pass/warn/fail with citations. Or use the three free tools (Agent UA tester, JSON-LD validator, robots.txt checker) for targeted audits.

How to make your website readable to AI agents

The single biggest improvement most sites can make is server- rendering primary content. AI agents read raw HTML before they render. A page that hides its hero, pricing, or product copy behind a JavaScript bundle gets summarised from the empty shell: usually nav links, a footer, and skeleton placeholders. Agents cite “this site has nothing to do with X” while X is in fact the entire homepage rendered after hydration.

Practical checklist

SSR or static-generate marketing pages. Hero copy, value props, pricing, CTAs all in the initial HTML response.
Treat JavaScript as enhancement. The page should be useful with JS disabled. View-source on critical pages should reveal the content.
Add structured data. JSON-LD Organization and WebSite on every page; per-page types (Product, Article, FAQPage, BreadcrumbList) where they apply.
Set unique meta. <title> and <meta name="description"> per page. <link rel="canonical"> resolves duplicate variants.
Allow AI agents in robots.txt. Explicit User-agent: GPTBot / Allow: / blocks override inherited wildcards.
Publish /llms.txt. Curated index of your canonical pages: docs, pricing, API, support, policies.

How to verify

Three free tools. First, fetch your URL with the Agent UA tester. It diffs how ChatGPT, Claude, Perplexity, and a default browser see your page. If the body similarity drops below 0.80 between bots and the browser, that’s cloaking or a JS-only render. Second, validate your structured data with the JSON-LD validator. Third, run a full AgentSpeed scan for an end-to- end report.