How to get Google and AI to index your website
A beautiful website is useless if the robots that decide your visibility can't understand what you do.
TL;DR To get Google and AI to index your website, first remove indexing blocks (noindex above all), then mark up your entities in JSON-LD and align your local and language signals. Robots can't see your design: without semantics, there's no visibility.
“What is essential is invisible to the eye.”
— Antoine de Saint-Exupéry, 1943
Your new showcase site has been live for three months, and you're proud of it. The design is sharp, the photos are beautiful, and your clients' feedback has been enthusiastic. Everything says the work was done well. I work from a village at 1,650 metres: a lovely view and hard to find, I know the combination well.
Then you type your main service plus your region into Google. Nothing. You ask Perplexity or ChatGPT the same question, and they happily cite three of your competitors. Still nothing, not even your name in a footnote.
The site doesn't have a quality problem; it has a readability problem. For the machines that decide your visibility, you simply don't exist yet, and no advertising budget changes that.
Why publishing a site is no longer enough to be found
The web is saturated. Google no longer indexes every new page the moment it goes live: it sorts, it postpones, and it quietly ignores whatever doesn't seem worth the effort. Going live is no longer the end of the project; it's the start of an exam.
Crawl budgets are shrinking
Search engines handle billions of URLs and spend less and less time exploring sites that perform badly. Crawl budget is the number of pages a robot is willing to visit on your site before moving on. If your server is slow to respond, if your pages are heavy or if half your links lead to redirects, the robots turn back before reaching what matters. Every visit wasted on a redirect or a duplicate is one your service pages don't get.
Waiting for the algorithm to discover your content on its own is wishful thinking. A recent site with no inbound links and no history starts with the smallest budget there is, and that's precisely when every visit counts.
From isolated keywords to a network of entities
AI assistants and answer engines no longer look for simple keyword matches. They map knowledge graphs: who you are, what you offer, where, at what price, and who can confirm it. An entity is something that can be identified without ambiguity, like a company, a place or a service.
If your code is a tangle of generic containers with no clear hierarchy, no machine can pull out the essence of your offer. It's actually the most common flaw of AI-generated websites that stay invisible: visually flawless, semantically silent. Without explicit entities, you remain a digital ghost.
Check your indexing tags before anything else
Start by checking that no tag is stopping Google from indexing your pages. It's the most common deployment mistake, even on projects run by established agencies. It's also the cheapest to fix, as long as you know where to look.
Hunt down leftover noindex directives
First, check that your robots.txt file doesn't block the resources needed to render the page, such as stylesheets and scripts. Then make sure no noindex meta tag set during staging survived the launch. Watch out for the classic trap: a page blocked in robots.txt can't be crawled, so the robot will never see its noindex tag. The two mechanisms don't add up, they cancel each other out.
While you're at it, check that you're letting in the answer engines' robots. OAI-SearchBot, PerplexityBot and Claude-SearchBot feed the assistants' citations; blocking them out of habit amounts to asking not to be cited.
Finally, generate a dynamic XML sitemap, submit it in Google Search Console and Bing Webmaster Tools, then read the page indexing report. "Discovered – currently not indexed" means Google knows the address but didn't think it was worth visiting. "Crawled – currently not indexed" is harsher: it read the page and decided it didn't deserve a place. The first is a signals problem, the second a content problem.
Add structured data to every page
Add Schema.org markup in JSON-LD to every page. Large language models don't read between the lines. You have to give them the dictionary, and Schema.org markup is that dictionary: a vocabulary shared by Google, Bing and the assistants to describe a business, a service or an article without ambiguity.
Connect your Schema.org markup into a graph
Add complete JSON-LD markup that links your business entity (LocalBusiness or Organization) to your services (Service), your website and the person who runs the business. Spell out your service areas, your prices and your direct contact details. Each node carries a stable identifier, and those identifiers refer to each other from page to page: that's what turns a collection of tags into a graph.
By providing a standardized structure, you remove ambiguity and give AI exactly the material it needs to cite you. Markup doesn't replace content, it makes it verifiable, and that's the whole point of a website optimized to be cited by AI. Let's be honest: no structured data guarantees a rich result or a position. Google says so itself. It improves your chances of being understood, not your right to rank.
Then test every page template, not just the homepage, with Google's Rich Results Test and the Schema.org validator. A single syntax error in one JSON-LD block is enough to get the whole thing ignored, and nothing on screen warns you. It's the kind of flaw you only find by looking for it.
Consolidate your local and language signals
Make your site, your Google profile and the directories that list you say the same thing. A web page cut off from the rest of the digital ecosystem lacks legitimacy. Engines cross-check what your site claims against what your Google profile, the directories and the sites that mention you say, and they extend their trust when everything matches.
Align your NAP and your hreflang tags
Make sure your core data, your NAP (Name, Address, Phone), is strictly identical on your site, your Google Business Profile and the main Swiss directories. One different abbreviation or an old number is enough to create two entities where there should be one, and to split your credibility in half.
I learned it on my own listing. The studio is at Chemin de la Mounta 24 in St-Luc, as spelled in the official address register; my Google Business Profile showed "Ch. de la Monta 24". An abbreviation and one missing letter are enough to turn one address into two, and the fix takes far longer than the mistake took to settle in.
If your site speaks to both French- and German-speaking Switzerland, set up flawless hreflang tags: one URL per language, with each version pointing to all the others, itself included. That's how the engines know which version to show each user, instead of pitting your versions against each other.
Visibility is an architecture decision, not a finishing touch
You don't add discoverability at the end of a project with a handful of keywords scattered at random. When the structure is wrong, every later fix costs more than getting it right from the start, and most invisible sites are invisible for reasons decided in the first week.
Build the code for robots first
That's the very foundation of a modern SEO strategy: semantics shape the structure of the pages from the first day of development. Hierarchical headings, HTML markup that says what each block is, structured data generated with the content rather than pasted on afterwards. Planned from the start, it costs almost nothing; added later, it often means redoing half the templates.
You also have to accept a limit. Indexing is only the front door, the first of the three types of SEO: it gives you the right to be assessed, not the right to win. Once you're indexed, it's how relevant your content is to the question asked that makes the difference, and no technical setting replaces that.
Ready to speak the language of answer engines?
If the algorithm doesn't understand you, your clients won't find you, however good your work is. The good news is that everything above can be checked, fixed and measured. Clean up your code, structure your data, and let the machines do the rest.
👉 Let's audit your pages' technical visibility together. Get in touch