Set up technical and on-page SEO before a site goes live
Search engines and AI tools should read every public page, each page should say what it is, and the markup should validate. This skill makes the changes it can reach and lists the rest for you.
Example
You ask
We launch our new site next week. Check the technical setup, and tell me which decisions are mine, such as the AI crawler rules.
The assistant
Your assistant checks the host, HTTPS, canonical links, titles, sitemap, and structured data, and makes the fixes it can reach. It asks the three AI crawler questions before writing any rule, and labels each item verified, implemented but unverified, or blocked.
Use it when
- A new site is about to go live, so its pages can be read and measured from day one.
- A site is being rebuilt, moved, or repaired, and its basics need a check.
- You ask what is missing from a site's technical setup or on-page basics.
- The robots.txt file, the sitemap, or the structured data needs a review.
You give it
- Access to the site files or builder. Without it, your assistant writes out each change for you to make.
- The stack or builder the site runs on, if you know it. Your assistant asks once if it does not.
- Your answers to three AI crawler questions: AI search answers, AI training, and AI reading a page on request.
It returns
- Each setup area checked, with what is done, what is not, and what you must do yourself
- A robots.txt file, a sitemap that lists only canonical pages returning 200, and JSON-LD that matches the visible page
- Each item labeled verified, implemented but unverified, or blocked, plus a reference scan in priority order
How the skill runs
Set the address and rendering
Pick one host and one URL form, and redirect the others with a 301. Serve HTTPS only. Put each page's text, title, description, and links in the HTML the server sends, then check a plain fetch of the page.
Check HTML and accessible names
Set lang on the html element, one H1 per page, and headings in order. Use native buttons and links, label every form field, and give each icon-only control a name. Every informative image gets alt text.
Set the on-page basics
Each page gets a unique title under 60 characters, a description of 140 to 160 characters, an H1 that matches the intent, and a canonical link in the chosen URL form. Set Open Graph title, description, and a 1200 by 630 image.
Set robots.txt and AI crawler rules
Ask the three AI questions first, and do not guess the answers. Then apply each answer with the control that does it: a robots.txt line blocks only the crawler it names. Keep Googlebot allowed, and add a Sitemap line.
Add sitemap and structured data
List only canonical pages that return 200, keep lastmod accurate, and check that the XML parses. Use JSON-LD that matches the visible page, with no FAQPage or HowTo for a search appearance. Validate every template.
Check speed and measurement
Use lab results to find fixes and field data to confirm them. Target LCP of 2.5 seconds or less, INP of 200 milliseconds or less, and CLS of 0.1 or less at the 75th percentile. Set up Search Console and one GA4 tag on every page.
Run the reference scan and report
After the build, scan the site and list the failing checks in priority order. Label each item verified, implemented but unverified, or blocked, and list what you must do yourself, such as DNS, the consent banner, and the AI crawler answers.
Where it stops
- It does not guess AI crawler rules. Blocking the wrong name can remove the site from AI search answers, so the owner answers first.
- Googlebot stays allowed. Blocking it to stop AI use removes the site from Google Search, including AI Overviews.
- A failing llms.txt check concerns agent discovery. Google Search does not use llms.txt.
- Without access to the site files or builder, each change is written out for you to apply. DNS and the consent banner need your own access either way.
Questions before you install
Do I need to know how my site is built?
No. If the stack is unknown, your assistant asks once. With access to the files or builder, it makes the changes itself. Otherwise it writes each change out with its old and new value.
Do I have to allow AI crawlers?
No. You decide three things: whether your pages may appear in AI search answers, whether AI companies may train on your content, and whether AI tools may read a page when someone asks about it. The rules follow your answers.
Does an llms.txt file help rankings?
No. The file serves agent discovery, and Google Search does not use it. It does not promise rankings or AI citations, so add it when you want agent discovery and say that is the reason.