Making React sites discoverable by Google and AI

To make a React site discoverable, start by giving each page a real URL and readable content. A crawler needs to find the article, receive its text and understand the topic. The framework’s name does not do that work for you.

Look at what the server returns

On this portfolio, TanStack Start renders articles on the server. The title, paragraphs, code and references are in the initial HTML. The demos become interactive afterwards. You can read the substance of the article without running JavaScript.

1# After starting the site locally
2curl -sS http://127.0.0.1:3000/en/writing/working-better-with-codex
3
4# Also check the status of a missing page
5curl -I http://127.0.0.1:3000/en/writing/unknown-article

Server rendering guarantees no ranking. Here, it makes the content directly available. Also check links from the homepage, redirects and pages that return errors.

Keep the information consistent

A canonical URL identifies a page’s preferred address. The sitemap lists public URLs to discover. Structured data describes the author and article. These should match the visible page, without artificially updated dates or invented expertise.

1{
2  "@context": "https://schema.org",
3  "@type": "BlogPosting",
4  "headline": "Working better with Codex",
5  "author": {
6    "@type": "Person",
7    "name": "Killian Jeuffroy"
8  },
9  "inLanguage": "en",
10  "datePublished": "2026-08-31",
11  "mainEntityOfPage": "https://portfolio.example/en/writing/working-better-with-codex"
12}

A preview site should not declare an invented address. Configure the real domain before publishing, then check the deployed HTML. A correct tag in the source is not enough if the host returns another page or asks you to sign in.

What about AI answers?

Google says the basics of SEO still apply to its AI search features. It does not require special markup. Its guide also says Google Search does not use llms.txt. Adding that file is therefore not a shortcut to being cited.

An llms.txt index can offer links to other readers or tools. Support varies. For my pages, I focus on a clear answer at the start, examples you can inspect, an identifiable author and references to check the technical points.

Measure after publishing

Choosing a current topic gives you a direction, not a traffic promise. After publishing, look at queries, impressions and clicks in Search Console. A title can attract and disappoint if the article does not answer the question. I would rather improve that answer than multiply variations of the same text.