AEO, GEO and SEO for a Developer Site: The Checklist I Actually Applied

I run a developer site, and I wanted it to be found by two audiences: people using Google, and people asking an AI assistant. This is the checklist I applied. It separates the changes that are well understood from the ones that are bets, and it says how to verify each one.

I have no traffic results to report yet. Anyone who promises specific ranking or citation outcomes from this kind of work is guessing.

What should the first paragraph of every page do?

It should answer the question the page is about, in plain sentences, with the key fact included. Search features and AI assistants both prefer to lift a self-contained answer. Put the answer first, then the evidence, then the caveats.

On my benchmark pages, the first block is the question as a heading and a two-sentence answer containing the numbers. Example: "Is HashSet.contains faster than ArrayList.contains in Java? Yes: 10,000 lookups in 100,000 items took 1,762 ms versus 0.25 ms on JDK 17."

Which structured data is worth adding?

Use JSON-LD that matches what is on the page. These are the ones I use:

• Article or TechArticle with headline, dates, and an author that points to one shared Person entity by @id, so every page ties back to the same author.
• FAQPage only when the page visibly contains those questions and answers.
• BreadcrumbList for the page hierarchy.
• CollectionPage with an ItemList for index pages.

Do not mark up content that is not on the page. Check with Google's Rich Results Test and the Schema.org validator after each deploy.

Should I let AI crawlers in?

If you want to be cited, do not block them. My robots.txt explicitly allows Googlebot and Bingbot and the AI crawlers I know of (GPTBot, OAI-SearchBot, ChatGPT-User, ClaudeBot, Claude-SearchBot, PerplexityBot). That is a business decision: allowing training crawlers is not the same as allowing search crawlers, and you may want different rules for them.

Is llms.txt worth it?

It is a cheap bet. I publish /llms.txt with a short description of me, my work, and links to my best pages. I do not know of evidence that it changes rankings or citations, so I treat it as a convenience for agents that choose to read it.

What gives a page a reason to be cited?

Original information. Assistants and search engines have little reason to cite a page that restates what ten other pages say. A page that contains numbers nobody else published, with the method and the caveats, is citable.

That is why I turned my own measurements into pages: ArrayList vs HashSet in Java, CSV vs Parquet in DuckDB, nginx proxy_cache throughput and others. Every page states the environment, the setup and when the result does not apply. The whole set is on the benchmarks index.

How do I do programmatic SEO without making spam?

Generate pages only when each one has unique data, and keep the template honest. My benchmark pages come from one JSON file per measurement: question, answer, results, setup, code, reason, limits, environment, FAQ and related pages. A page is only created when a real measurement exists. If I cannot fill the "when this does not apply" field truthfully, the page does not ship.

Programmatic pages that only swap a keyword into a template are what search engines learned to demote.

How do I verify each change?

1. View source or run curl to confirm the answer text, canonical tag and JSON-LD are in the HTML a crawler receives, not only in client-side rendering.
2. Run the Rich Results Test on one page per template.
3. Submit the sitemap in Search Console and watch the indexing report.
4. After a few weeks, compare impressions and clicks per page group.
5. Ask two or three assistants your target questions and note whether your pages appear as sources. Repeat monthly.

What would I do next?

Keep publishing only what I can measure, link related pages to each other, update pages when the numbers change, and read the Search Console data before deciding what to write. The checklist is the easy part. The hard part is having something worth citing.