How to Add SEO, GEO and AEO to an Existing Website Without a Rebuild

Short answer: SEO, GEO and AEO are added to an existing site in four layers, without touching the design: a technical foundation measured from outside and fixed (found), structured data and internal links that tell machines what each page is about (understood), verifiable copy AI systems can cite (trusted), and contacts measured under one identifier (chosen). Measurement decides which items are needed; the platform decides which can be done.

This guide does not answer "should you rebuild"; there is a separate decision guide for that. The assumption here: the foundation is sound and the layers are missing.

Measure first: how does the site look from outside?

An audit that never reads source code and crawls only over HTTP shows the site as a search engine sees it: which pages are indexable, where canonical points, whether hreflang is consistent, whether the sitemap and the crawled pages overlap, which pages receive no internal link at all (orphans), whether bots are blocked. Any proposal written without this measurement is a guess. How to read such a report is a separate article.

The measurement is combined with the data in your own Search Console and Analytics accounts. If they do not exist they are set up; if they do, access is granted and ownership stays with you.

Layer 1: Found

The technical foundation, in order:

• Indexability. Pages left on noindex, directories robots.txt closed by mistake, internal links returning 4xx/5xx.
• Canonical. Every page should point at itself or its correct version; a canonical that points at the homepage makes the page disappear in the search engine's eyes.
• hreflang. On a multilingual site every language version should list the others and itself; a missing or contradictory tag makes the versions compete with each other.
• Sitemap and robots. A sitemap should list only indexable URLs on its own host. A subdomain or a demo URL does not belong in it.
• Search-engine coverage. Google and Bing Webmaster verification, sitemap submission, instant notification of Bing and Yandex through IndexNow. A host's bot filter sometimes blocks Bing; saying "we are not on Bing" without checking this is a wrong diagnosis.
• Performance basics. Google indexes and ranks the mobile version (mobile-first indexing); heavy images and blocking scripts on mobile are cleared first.

This layer depends most on the platform. On a site whose source code or CMS is accessible it is done directly; if the site sits with another agency, access is opened for the duration of the work and the implementation is still done by the same hands; on closed platforms some items cannot be done and the proposal says "not possible on this platform". We defined those two boxes on the service page.

Layer 2: Understood

What a page tells a person it should also tell a machine.

• Structured data. Organization, Service, FAQPage, BreadcrumbList and the types that fit the sector (LocalBusiness, Product and so on). One consistent graph tied to a single #org node, not a copy-pasted block per page.
• Entity consistency. Business name, address, phone and service names written the same way on the site, on Google Business Profile and in the schema.
• Topic clusters and internal-linking architecture. Hub page → service (money) pages → guide articles. Every guide links to a service page; no money page is orphaned; the anchor text to the same page varies.

This layer can be done on most platforms, and it is the one most often missing.

Layer 3: Trusted

For AI systems and human readers alike: a claim that can be checked.

• Answer-first copy. The first paragraph answers the question; the ornament comes after.
• Verifiable claims. If there is a number, its source, tool and date sit next to it; a claim that is not measured does not stay on the page. We apply this on our own site in the verification block of the turkdizisi case: every claim carries the URL to check it and a prompt for auditing us.
• llms.txt. One plain-text file saying what the site is and which pages answer what.

These make citation easier; they do not guarantee it. Which source an AI answer picks varies, and no rate can be promised without measuring.

Layer 4: Chosen

If the site brings customers, there should be proof. Form, phone, email and WhatsApp are wired into one measurement mechanism: every contact carries a source and an identifier, marked in Google Analytics 4 with recommended events such as generate_lead (GA4: Recommended events). The chain is source → contact → qualified → won. Selling is the business's job; the measurement shows where the chain breaks.

Why this order?

Schema on a page that cannot be found, contact measurement on a page that is never cited: wasted. If the lower layer is not sound, the upper one does not work. So the proposal is written in this order too, and an item the measurement does not call for stays out of it even if it could be done.

What this guide does not say

It does not say which item is missing on your site; the external measurement does. Nor does it say what can be done on your platform; every platform changes with its versions, and the proposal is written after the measurement and after the platform is known.

Which layer is missing on your site?

How we build these four layers without rebuilding your site, with an itemised deliverables list, is on the SEO, GEO and AEO for your existing website page. What is needed is measured first by the Check-up.