SEO & AI Search Checklist – What to Check on Your Site

Crawler and bot access

robots.txt returns 200 and allows Googlebot and Bingbot on every page you want found.
OAI-SearchBot (ChatGPT search answers) and Claude-SearchBot are allowed if you want to appear in those answers.
GPTBot and ClaudeBot, which are used for model training only, are allowed or blocked as a deliberate choice, separate from search.
Google-Extended is set on purpose; it is a robots.txt token that controls Gemini training and grounding and does not affect Google Search.
Claude-User and ChatGPT-User, which fetch a page when a person asks, are allowed unless you decided otherwise; robots.txt is not guaranteed to bind ChatGPT-User.
A named bot group in robots.txt repeats the generic rules, because a group for a specific bot replaces the * group instead of adding to it.
Your CDN or firewall lets verified bots through (checked by reverse DNS or published IP lists, not by user-agent string) and does not serve them a CAPTCHA or login wall.

Every page you want in search returns 200 and has no noindex.
Pages you want out of search use noindex on a crawlable page, not a robots.txt Disallow, which would hide the noindex from the crawler.
Each indexable page has one absolute self-referencing rel="canonical" in the head, and internal links point at the canonical URL.
Moved or deleted URLs return a 301 or a real 404 or 410, never a 200 page that says ‘not found’.
The XML sitemap lists only canonical, indexable 200 URLs and is referenced from robots.txt.
Sitemap lastmod changes only when the main content changes, not on every build.

Titles, headings, body text, links and canonicals are in the server-rendered HTML; some fetchers never run JavaScript.
Every view has its own URL; no content lives behind a # fragment, a click, a scroll or a login.
Primary content is not lazy-loaded above the fold and is not hidden in tabs or accordions by default.
CSS and JavaScript needed to render the page are not blocked in robots.txt.

Every important page has at least one internal link, written as a real <a href>, not an onclick handler.
Link text says what the target is; no ‘click here’.
One topic has one page; you do not publish a separate thin page for each phrasing of the same question.
Filtered, sorted and internal-search URLs are kept out of the index.

Each page has one primary topic, and the title, H1 and meta description say the same thing.
The answer to the page’s main question comes first, in one or two plain sentences, before any background.
Facts that matter (what, who, where, price, dates) are written in text on the page, not only in images, video or PDFs.
Your business, person and product names are spelled the same everywhere on the site.

JSON-LD describes only what is visible on the page, uses the most specific schema.org type, and passes a validator.
Core Web Vitals pass at the 75th percentile in field data: LCP up to 2.5 s, INP under 200 ms, CLS under 0.1.
The mobile page carries the same content, headings, structured data and meta tags as desktop.

Google Search Console and Bing Webmaster Tools are verified, and sitemaps are submitted in both.
Clicks from answer engines are tracked by referrer or utm_source (ChatGPT appends utm_source=chatgpt.com).
AI-answer visibility is judged from console reports and repeated manual samples, never one answer.
If you publish an llms.txt, it returns 200 and matches the live site; no search engine documents it as a ranking, indexing or citation input.