# GEO: How to Get Your Website Cited in AI Search and Assistants

Generative Engine Optimization without the hype: crawler access, citable facts, structured data, consistent entities, llms.txt and what you can measure.

by Alexander Herling Published 25 August 2026 6 min read

![Abstract illustration: connected nodes in shades of blue condensing into a cloud, with a single line leading out of it](https://degoya.digital/blog/geo.svg)

More and more people put their questions to an assistant rather than a search box. Google shows AI Overviews above its results; ChatGPT, Perplexity, Copilot and others answer questions with cited sources. That raises a new question for businesses: is my website mentioned in those answers? Generative Engine Optimization (GEO) is the umbrella term for the work that makes this more likely. A caveat up front: much of what circulates on the topic is speculation. This article sticks to what can be reasonably justified.

## How AI search arrives at an answer

Simplified, most systems work in two steps. First they retrieve documents relevant to the question, from a search index or via their own crawlers. Then a language model writes an answer from those documents and cites some of them as sources.

Three things follow from that:

1. **No discovery, no citation.** A page that is not crawled and indexed cannot be selected.
2. **What gets selected answers the question well.** Clear, unambiguous passages are easier to use than convoluted sales copy.
3. **Trust matters.** Systems favour sources that are considered reliable elsewhere too.

Google itself says AI Overviews require no special optimisation and that the usual search fundamentals apply. Good SEO is the prerequisite. Our [SEO tips for 2026](https://degoya.digital/en/blog/seo-tips-2026) help you check where you stand.

## 1. Control crawler access deliberately

Review your `robots.txt`. Many sites block AI crawlers wholesale, often without a conscious decision, for example because of a host's or CDN's default settings.

Some providers distinguish between crawlers for model training and crawlers for search:

| Provider   | Example user agents        | Note                                                                                                          |
| ---------- | -------------------------- | ------------------------------------------------------------------------------------------------------------- |
| Google     | Googlebot, Google-Extended | Google-Extended controls use for Gemini models, not Google Search. AI Overviews use the regular search index. |
| OpenAI     | OAI-SearchBot, GPTBot      | OAI-SearchBot for search in ChatGPT, GPTBot for training.                                                     |
| Microsoft  | Bingbot                    | Copilot draws on the Bing index.                                                                              |
| Anthropic  | ClaudeBot and others       | Separate agents for training and retrieval; see their documentation.                                          |
| Perplexity | PerplexityBot              | For search in Perplexity.                                                                                     |

Names and roles change from time to time. Check each provider's current documentation before making changes.

Also note that many of these crawlers do not execute JavaScript. Content that is only loaded in the browser is invisible to them. Statically generated or server-rendered pages have the edge here.

## 2. Write to be cited

An AI answer rarely takes a whole page; it takes individual statements. Write so that those statements can stand on their own:

- **Answer first.** Answer the question of a page or section in the first sentence; the reasoning follows.
- **Clear definitions.** "A wireframe is…" rather than "wireframes can be many things".
- **Concrete, verifiable facts.** Figures, dates, deadlines, legal references, each with a source.
- **Questions as subheadings** where it feels natural. They mirror the way people ask assistants.
- **Tables and lists** for comparisons, steps and requirements.
- **Date and author.** Visible on the page and in structured data. Freshness is a factor for many questions.

Your own experience and data are especially valuable. If you only summarise what is already everywhere, there is no reason to cite you in particular.

## 3. Structured data

Structured data based on schema.org describes in machine-readable form what a page is about. The most useful types:

- `Organization` or `LocalBusiness` with name, address, logo and `sameAs` links to your profiles
- `Person` for authors, with role and profiles
- `Article` or `BlogPosting` with publication and modification dates
- `Product` and `Offer` for shops, `Service` for services
- `FAQPage` for questions and answers

Keep expectations realistic. Structured data helps search engines classify content correctly and enables certain search result features. There is no evidence that it directly controls inclusion in AI answers. Google also limited FAQ rich results to a small set of site types in 2023 and stopped showing them altogether in May 2026. The markup still does no harm and makes your content less ambiguous.

## 4. Keep entities consistent

Language models and search engines think in entities: organisations, people, places, products. The more consistent the information about your organisation across the web, the more confidently a system can attribute it.

- The same company name, spelled the same way, on your site, business profile, directories and social media
- Identical address and phone number
- A clear, recurring description of what you do and for whom
- Maintained profiles on platforms that matter in your industry
- On multilingual sites: the same facts in every language

## 5. Mentions elsewhere

What others write about you often carries more weight than what you write yourself. Trade articles, interviews, industry directories, reviews, talks and partner pages create mentions that AI systems come across. This is classic PR and resembles link building, except that unlinked mentions may count as well.

## 6. llms.txt: optional, with no expectations

`llms.txt` is a format proposed in 2024: a Markdown file in the site root that summarises and links a site's most important content for language models. The idea makes sense, but the file is not a standard. Major search providers have not confirmed that they use it for search or citations, and Google has been rather dismissive.

Our take: if your system can generate it automatically at little cost, there is no harm in it. For developer documentation consumed directly by coding assistants it can be useful. Do not plan on it as a GEO measure with measurable impact.

## What to avoid

- **Hidden instructions to AI systems** in page text or metadata. This is manipulation, providers increasingly detect it, and it can damage your reputation.
- **Mass-generated pages** for every conceivable question. Google's spam policies treat this as scaled content abuse.
- **Different content for crawlers and people.** That is cloaking, even when it is labelled "AI optimisation".

## Measure what you can

Visibility in AI answers can only be measured to a limited extent so far. Useful sources:

- **Referrers in your analytics:** visits from chatgpt.com, perplexity.ai, copilot.microsoft.com and similar
- **Google Search Console:** clicks and impressions from AI Overviews are included in performance data. Since June 2026 Google has been rolling out separate reports for generative AI features, initially to a subset of sites and with impressions only
- **Spot checks:** regularly put your customers' ten most important questions to several assistants and note who gets mentioned
- **Specialised tools** that automate these queries. Their figures are approximations, because answers vary by user, location and time

## Where to start

First check your `robots.txt` and whether your content is readable without JavaScript. Then rework the five pages that should answer your customers' most common questions: answer first, facts with sources, date and author visible. Add structured data for your organisation and authors, and align your company details across the web. If you would like support, see our [SEO and GEO services](https://degoya.digital/en/services/seo-geo).

## Frequently asked questions

Is GEO different from SEO?

GEO builds on SEO. AI search systems rely on search indexes and their own crawlers, so indexing, good content and authority remain the foundation. GEO adds a citable writing style and consistent information about your organisation.

Should I add an llms.txt file?

You can; the effort is small. Do not expect a measurable effect, though: the file is a proposal, not a standard, and major providers have not confirmed that they use it for search or citations.

Can I block AI training and still appear in AI answers?

Partly. Some providers use separate crawlers for training and for search, so you can treat them differently in robots.txt. The rules differ by provider and change over time, so check their documentation regularly.

## About the author: Alexander Herling

Founder and owner of DEGOYA.digital. A deep affinity for technology since early childhood and his first contact with 8-bit computers make him a digital native from day one. The command line is his friend. 35 years of experience in marketing and web development, specialising in SEO techniques.

Related service

[AI & automation tools](https://degoya.digital/en/services/ai-automation)

[Back to the blog](https://degoya.digital/en/blog)

Source: https://degoya.digital/en/blog/geo-ai-search-visibility