The Top 5 Ways to Increase your GEO Ranking.

There is just one small problem: GEO does not really have rankings.

Here at LeadFabric we've been exploring how brands can come up as reccomendations in ChatGPT, Claude and other LLMs. Google itslef has even started to prioritise AI summaries rather than their own search results... but there's one hitch for everyone who's an SEO guru: search engines like Google used to give users a list, but generative AI gives users an answer and reasoning besides it.

An AI engine might cite your website, mention your brand, paraphrase your expertise, recommend your services - or sail past you without so much as a polite nod. Even when it cites you, your source could provide one supporting fact or shape half the answer.

This makes Generative Engine Optimisation, or GEO, less of a race towards “position one” and more of a battle for answer participation, which is made up of three aspects:

• How often your brand appears (mentions)
• How prominently it appears and
• How much the answer relies on your information (citations)

This means that while SEO has a bit of overlap with GEO, the conventional approach to rankings do not fit generative engines entirely. Visibility now depends on several factors, including citation position, relevance, uniqueness and how much of the answer a source influences. 

So, when marketers talk about increasing a GEO ranking, what they should really mean is:

Increasing the likelihood that AI systems can find, understand, cite and recommend your brand.

Here are five practical ways to do exactly that based off of our experience and research.

The five ways to improve GEO visibility

1. Make your website accessible to AI search engines.
2. Stop publishing content a model could write without you - public content it can't.
3. Support your claims with extractable evidence.
4. Build authority outside your own website.
5. Measure prompts, not just positions - measure the sources shaping your visibility.

Let's check them out one at a time.

1. Make your website accessible to AI search engines

It sounds painfully obvious, yet plenty of brands are trying to charm ChatGPT while blocking its search crawler at the front door.

Before rewriting your content, check that search engines can crawl, render and index it. Your important pages should have clear internal links, sensible canonical tags, accurate XML sitemaps and accessible textual content. Your robots.txt file, CDN and firewall should not accidentally block the crawlers you need.

Google states that a page must already be indexed and eligible to appear with a snippet in traditional Search before it can appear as a supporting link in AI Overviews or AI Mode. It does not require a separate “AI version” of the page or any special GEO markup.

OpenAI uses OAI-SearchBot to discover and surface pages within ChatGPT’s search features. Crucially, OAI-SearchBot operates separately from GPTBot, which OpenAI uses for potential model training. This means a business can allow its content to appear in ChatGPT search while choosing not to make that content available for training.

Your technical checklist should therefore cover:

• Googlebot, Bingbot and OAI-SearchBot access
• Indexation and canonicalisation
• Internal links to important commercial and educational pages
• Accurate sitemaps
• Server-side rendering or reliable JavaScript rendering
• Clear page titles, headings and metadata
• Fast, mobile-friendly pages
• Important information presented as readable text

And before somebody adds an llms.txt file and declares the GEO strategy complete: Google explicitly says it does not use llms.txt for Search or its generative AI features. Creating one will neither help nor harm your Google visibility.

It may support a particular service in future. It is not, however, a substitute for a healthy website.

2. Stop publishing content a model could write without you.

You don't want to be article #3556 of "What is email marketing".

The internet will survive without another one.

Generative engines do not need more lightly rearranged summaries of information that already exists everywhere else. They can produce those summaries themselves, probably before your first coffee is ready.

They need sources that add something.

Google’s current guidance for generative AI search places particular emphasis on non-commodity content: original viewpoints, first-hand experience, specialist knowledge and information that goes beyond a generic summary. It specifically recommends creating content based on what you genuinely know rather than recycling what others have already published.

For a B2B organisation, this might include:

• Original research or anonymised benchmark data
• Lessons from real implementations
• A clearly named methodology
• Commentary from experienced subject-matter experts
• Detailed use cases
• Decision frameworks
• Honest limitations and trade-offs
• Before-and-after examples
• Insights drawn from customers, campaigns or buying groups

So, for example, instead of publishing “What is intent data?”, explain why intent data fails when Marketing and Sales use different definitions of “in-market.”

Instead of another broad “Guide to implementing 6sense,” explain the five implementation decisions that determine whether Sales actually uses it - and even better, what you've seen in your own organisation while implementing 6sense.

Instead of telling readers that AI is changing marketing - news that may come as a shock to anyone recently rescued from a Siberian cave - show them how AI changes a specific buying process, measurement framework or go-to-market decision within your own company, or one you are working with.

The tools matter - but the experience behind the tools matter more.

3. Support your claims with extractable evidence.

If you've ever tried to return a Black Friday impulse buy, you'll know that every claim needs a receipt. The same is true with GEO. 

AI systems need information they can extract and reuse. Your job is to make that information easy to find.

A strong GEO page should contain clear definitions, specific facts, relevant statistics, expert commentary, comparisons, examples and practical steps. It should also explain where its information came from.

That does not mean sprinkling random numbers through the copy like parsley. Every statistic should support the argument, name its source and provide enough context for the reader to understand it.

The original GEO study, accepted at KDD 2024, found that adding relevant citations, quotations and statistics could increase source visibility by as much as 40% under its benchmark conditions. Keyword stuffing, meanwhile, produced little benefit and sometimes made performance worse.

Now, that 40% is not a universal uplift guarantee. The research used defined experimental conditions, and commercial AI systems continue to change, but the broader lesson remains useful:

Evidence always beats padding.

Good evidence can include:

• A precise definition
• A dated statistic from a primary source
• An attributed expert quotation
• A table comparing approaches
• A step-by-step process
• A documented customer result
• A transparent research methodology
• A concrete example of the principle in action

Structure helps too, but not because AI systems possess a secret fondness for FAQ pages.

A 2026 observational study covering 602 prompts across ChatGPT, Google and Perplexity found that pages with greater apparent influence tended to contain modular sections, strong semantic alignment and extractable material such as definitions, numbers, comparisons and procedures. It also found that converting content into a question-and-answer format did not improve influence by itself - it was the structure underpinning them.

Use clear headings. Answer important questions directly. Give each section a distinct purpose. Remove the throat-clearing copy that says very little over several impressive-looking paragraphs.

Make the page easy for a human to navigate and easy for a machine to interpret. These goals are not competing.

4. Build authority outside of your website.

Your website can say you are brilliant. It would be slightly awkward if it did not.

But buyers (and the systems advising them) will also look for evidence from sources you do not control.

Generative engines often construct answers by retrieving and synthesising information from several websites. Google describes its AI features as using query fan-out, where one complex question can trigger several related searches across different topics and sources.

That means your wider digital reputation matters.

Earn accurate, substantive coverage through:

• Respected industry publications
• Partner and technology ecosystem pages
• Customer stories
• Event and conference websites
• Expert interviews and podcasts
• Relevant business directories
• Review platforms
• Professional communities
• Research reports
• Trade associations

One 2026 preprint analysed 167,551 URL-grounded citations relating to 128 brands across multiple markets and languages. Within that dataset, 85.7% of citations pointed to third-party websites rather than brand-owned domains. The study does not prove that every model or industry follows the same ratio, but it provides a strong reminder that AI systems build brand narratives from far more than the brand’s homepage.

The answer is not to manufacture mentions across low-quality websites.

Instead, create material other people genuinely want to reference. Contribute expertise. Publish useful research. Work with credible partners. Give customers a reason to discuss the results you created together.

You should also keep basic brand facts consistent across the web:

• What does the company do?
• Which customers does it serve?
• Where does it operate?
• Which technologies does it support?
• What expertise does it possess?
• How should its name and services be described?

You cannot completely control what AI says about your brand, but you can improve the evidence available to it.

5. Measure prompts, not just positions.

A single AI answer is not a baseline.

Responses can differ between platforms, markets, languages, prompt wording and points in time. Different engines may retrieve different sources or use those sources in different ways. We recommend measuring whether a page gets selected as a citation separately from how deeply the final answer uses it.

Start with a controlled set of prompts based on real commercial questions.For a B2B organisation, that prompt set should cover several kinds of discovery:

Category prompts

What are the best B2B go-to-market agencies in Europe?

Problem prompts

How can an enterprise identify more in-market buying groups?

Capability prompts

Which partners can implement and operationalise 6sense?

Comparison prompts

What should I look for when choosing an account-based marketing agency?

Recommendation prompts

Which European agency can connect buyer intelligence, marketing technology and campaign execution?

Run the same prompts repeatedly across the relevant AI platforms. Keep the location, language and wording consistent so that you can compare like with like.

For each prompt, record:

• Whether the brand appeared
• Whether it received a recommendation or a passing mention
• Where it appeared within the answer
• Which sources the engine cited
• Which owned pages appeared
• Which third-party pages appeared
• Whether the description was accurate
• Which competitors or alternatives appeared alongside it
• Whether visibility changed over time

Then investigate the source gap.

When another brand appears and yours does not, ask:

• What evidence supports that recommendation?
• Which publication, customer page or directory did the model use?
• Does the cited page answer the question more precisely?
• Is your equivalent information missing, vague or inaccessible?
• Does your brand lack third-party corroboration?

Traffic still matters, but it tells only part of the story. Google now directs site owners to a Generative AI performance report in Search Console, while OpenAI automatically adds utm_source=chatgpt.com to referral links from ChatGPT search.

Track those visits and conversions.

Just do not confuse “no click” with “no influence.” A buyer may encounter your brand in an AI answer, discuss it internally and return through another channel later.

GEO measurement must therefore connect visibility to the wider buying journey - not merely count blue links.