AI Search Visibility (GEO / AEO)

A customer now, instead of typing “best hotel in Antalya” into Google, asks ChatGPT and gets a single answer. That answer mentions three brands. If yours isn’t among them, that customer never saw you at all — and you can’t even measure it.

There are three names for this space, all closely related: AEO (Answer Engine Optimization) means showing up in answer engines, GEO (Generative Engine Optimization) means AI citing you as a source, and AIO refers to Google’s own AI Overviews layer. In practice, they’re all different faces of the same work.

Why does AI choose a particular source?

Language models prefer reliable, clear, and easily machine-parseable sources when generating answers. That boils down to three things: the content directly answering the question, claims being concrete and verifiable, and the page being structurally clean.

Pages written in marketing language that say “we’re the industry leader” without giving a single number are almost never cited by AI. Pages that give a clear answer to a clear question, with dates and figures, come out on top.

Publish the corrected identity where machines can read it

Once an incorrect fact is located, make the authoritative version clear to readers and machines. Structured data can describe that version; it cannot force a model to adopt it.

Establish one consistent organisation identity

Use the appropriate organisation markup on the home page or an organisation page. Include applicable, public facts: registered and trading names, contact details, logo and official profile links in sameAs. A sameAs list identifies relevant official references; it is not an instruction to distrust every other source. Add LocalBusiness data only where it fits the business and the visible page.

Validate supported markup with Google’s Rich Results Test, address critical errors and assess relevant warnings. The code and visible text must agree. Updating an address only in JSON-LD leaves the reader-facing contradiction unresolved.

Notify discovery systems without promising indexing

Request recrawling of changed pages through Search Console and keep the sitemap accurate. IndexNow can notify participating search engines, which share submissions within the protocol; each engine still decides whether and when to crawl or index. Google uses its own submission route rather than IndexNow. Verify the current participant list before designing an integration.

Repeat the visibility checks

Google’s AI features vary by supported country, language and query. Check the current availability page for Turkey and Turkish rather than promising an AI summary for every search. Repeat the same documented questions after correction, separating factual accuracy, brand mention, citation and referral traffic.

Sources checked in October 2026: Organization structured data · IndexNow FAQ · AI Overviews.

The technical side: is your site readable for AI agents?

A new measurement area has emerged: how well an AI agent understands your page. Chrome’s Lighthouse has an experimental “Agentic Browsing” category for this, but according to Chrome’s documentation it does not give a 0-100 score; it shows the ratio of readiness checks passed. It looks at things like the accessibility tree (names, labels, visibility), CLS, whether an llms.txt exists at the domain root, and WebMCP tool registration, and testing it requires Chrome 150 or later (Chrome documentation). Because the standards are still emerging, I read it as a checklist, not a ranking signal. A page with broken heading hierarchy, unlabeled links, and no marked main content area looks normal to a human but reads as a mess to an agent.

Structured data (schema) is decisive here. FAQPage, Organization, Person, Service, and Article markup are the most direct way to tell a machine “this is what’s on this page.” Set up correctly, it makes life easier for both Google and language models.

Measurement: how will you know you’re showing up?

The hardest part of this space is measurement. Classic rank tracking doesn’t work here because there’s no single “position.” What I do is regularly ask the questions actually asked in your industry to different AI engines and track whether your brand gets mentioned.

Over time this produces a curve: which questions you show up in, which ones your competitors dominate, what changed after the work was done. It’s not a perfect measurement, but it’s enough to make decisions on, and it’s the best available right now.

This work has three separate parts

What I described above isn’t a single package; it’s three stages you can take separately depending on your needs:

• AI Visibility Audit — diagnosis. Which questions you’re currently mentioned in, where your competitors show up instead, and why. Written report.
• Preparing a Site for AI Search — implementation. llms.txt, crawler permissions, entity schema, citable content structure.
• AI Search Brand Tracking — monitoring. Monthly measurement, competitor comparison, change report.

Most projects start with an audit; jumping straight to implementation without seeing where you currently stand is working blind.

How do we start? First, we talk about what you want and what’s realistic. If it works for you, I send the scope and price in writing, then we start. You can reach out via the contact page.

If you would like a first look at your own site, the free analysis tool scans your address and reports the technical and content signals a machine can measure (it does not measure whether AI answers mention you). If you would rather go through the findings together and turn them into a prioritised to-do list, the Visibility Session is a paid, two-hour one-to-one working session.

Frequently Asked Questions

Is GEO replacing SEO?

No, it builds on top of it. AI engines crawl the same web pages too; GEO doesn’t work without a solid SEO foundation. SEO makes you discoverable, GEO gets you inside the answer. They’re not separate budget lines — they’re two layers of the same work.

How long until you see results?

It’s less predictable than classic SEO because language models’ training and crawling cycles work differently. In engines that do live search (Perplexity, ChatGPT’s search mode), changes can reflect within weeks; settling into the model’s own knowledge takes much longer.

My competitors show up in ChatGPT, I don’t. Why?

The three most common reasons: not enough mentions of you elsewhere online, your site not having content that directly answers questions, and the machine technically failing to parse the page. I check all three individually in the initial analysis. Also check whether your business name, contact information and service descriptions are consistent across the web.

What’s your own site’s standing on this?

I set up what I describe here on my own site first: llms.txt, a connected schema graph, Q&A markup and open AI crawler access. Because Lighthouse’s Agentic Browsing category is experimental and reports a pass ratio rather than a score, I don’t promise you a “score”; I show what I measured and how I measured it.