Will the new site get found? A technical SEO and AI-search checklist
A rebuild can quietly delete the search visibility you already have. Ten technical checks to run before a relaunch, what helps AI assistants quote a page, and what nobody can honestly promise.
- Most relaunch traffic loss is technical and preventable: broken redirects, hidden content, duplicate URLs.
- AI assistants quote pages that state the answer early, plainly and with a named source.
- Run the audit on the staging site before launch, not on the live site after.
- Nobody can guarantee rankings or AI citations. Be wary of anyone who does.
- We have no traffic or citation result to show yet. We say what we checked, not what we won.
Quick answer
Check ten things before relaunch: that pages are crawlable without scripts, one clean URL per page, redirects from every old URL, canonical tags, a sitemap, structured data, fast loading, clear headings, answer-first page openings, and a visible author and date. No one can promise rankings or AI citations. They can only remove the technical blockers and publish pages worth quoting.
Disclosure: our co-founder Aadarsh Mishra also co-founded CrawlTheory, an SEO publication we have worked on. It is one of the sources of this checklist, so weigh it accordingly.
The short answer
A new site doesn’t lose traffic because the design changed. It loses traffic because something technical broke: old URLs that lead nowhere, text that only appears after a script runs, two URLs for the same page. Those are checkable before launch.
Why marketing heads ask this
You have visibility you’ve built over years, and a rebuild puts it at risk. The other half of the question is new: buyers now ask ChatGPT, Perplexity and Gemini before they ever open a search page, and you want to be the answer they quote. You need a checklist, not a promise.
What good looks like: ten checks before relaunch
| # | Check | Why it matters |
|---|---|---|
| 1 | Content is readable without scripts. View the page with scripts off, or fetch it as a crawler would. | If the text only exists after scripts run, some crawlers see an empty page. |
| 2 | One clean URL per page. | Duplicates split signals and confuse crawlers. |
| 3 | Every old URL has a redirect to its closest new page. | The single biggest cause of relaunch traffic loss. |
| 4 | Canonical tags are set and point to the right URL. | Tells search engines which version counts. |
| 5 | A sitemap exists, lists only real, indexable pages, and is submitted. | Helps discovery. |
| 6 | Crawlers aren’t blocked by mistake. Check the robots file, including crawlers used by AI tools. | Staging settings sometimes ship to production. |
| 7 | Titles, descriptions and headings are clear, with one main heading per page. | They are how both people and machines read the page. |
| 8 | Structured data describes real content such as article, FAQ, organisation. | Helps systems understand what a page is. |
| 9 | Pages load fast on a phone. | Slow pages lose readers before they read. |
| 10 | Each page opens with the answer, names its sources, and shows author and last-checked date. | This is what makes a page easy to quote. |
Run the list on the staging site. Compare it with the live site by URL: every old URL should map to a new one or to a deliberate removal.
What helps an AI assistant quote a page
No one outside those systems knows exactly how they choose. But the sensible things are the same ones that help a careful human reader:
- State the direct answer in the first lines.
- Use a heading for each question the page answers.
- Name where a fact came from, with a link.
- Show who wrote it and when it was last checked.
- Say plainly where the advice doesn’t apply.
- Let crawlers in.
How we’ve done it
Work on CrawlTheory, a publication Aadarsh co-founded, included an admin CMS, a full-site UX audit that surfaced 52 issues, a mega-menu redesign, a brand kit, and fixes for visibility to generative-search crawlers. The resource hub you are reading uses the same structure. We have no traffic or AI-citation result for either.
The audit is the part that carries over. We went through CrawlTheory page by page and found 52 issues. The CMS work then made the fixes systematic: SEO fields, FAQs and dates became fields on each page, and not text someone had to remember to type. The hub you are reading follows the same shape: answer first, sources named, a last-checked date on every page.
What we haven’t shown. We can’t say these changes raised traffic or got us cited. We haven’t recorded either, so we make no claim. What we can say is that we ran the checks, and what they were.
Where this doesn’t apply
If search is not how your customers find you, such as an invite-only B2B product sold through referrals, a long SEO checklist is the wrong priority. And if your content is thin, technical fixes won’t help: a perfectly crawlable page that says little will not be quoted. Technical SEO removes blockers. It doesn’t create demand.
Questions to ask any vendor, including us
- How will you map and redirect every old URL, and can I see the map before launch?
- Will you audit the staging site against this list, and share the result?
- What do you promise: specific work, or a ranking?
- Where do titles, descriptions and FAQs live: as fields, or inside the body text?
- How will we measure this after launch, and when do we look?
Next: can marketing publish without a developer and should one vendor do the tech and the go-to-market. Part of the guide Buying custom software and hiring a tech vendor.
The CrawlTheory work comes from our project notes and is tracked in our claim ledger. Aadarsh co-founded CrawlTheory. The ten checks and the reader-friendly list are our method and common practice, not a claim about any search engine’s ranking rules. We hold no traffic or citation measurement. No figure on this page comes from a third party.
More by GarunaCDX
Let's talk about your build.
Tell us what you are building. A founding engineer replies within one working day with how we would approach it.
