Generative Engine Optimisation (GEO): The 2026 Guide

Search used to end with ten blue links. Now it increasingly ends with an answer. AI Overviews, AI Mode, ChatGPT, Perplexity and Gemini read the web, write a response, and cite a handful of sources. If you are one of those sources you are recommended. If you are not, you are invisible for that query, however well your page ranks.

Generative engine optimisation, GEO, is the practice of making your content the source those engines choose. This guide explains how the engines decide, what the research actually shows works, and the framework I use to get clients cited.

What is generative engine optimisation?

GEO is optimising your content and brand so that AI powered search tools use it when they generate answers, ideally with a citation back to you.

You will see several overlapping terms. Here is how they relate:

In practice these are one discipline with different labels, and I have written separately about answer engine optimisation where the target is the snippet rather than the generated paragraph. What matters is the goal: when someone asks an AI about your area of expertise, your brand appears in the answer.

Why GEO matters now

The scale is no longer theoretical. Google put AI Overviews at over 2.5 billion monthly active users at I/O in May 2026, with AI Mode passing one billion monthly users in its first year.

Three shifts make GEO urgent.

1. Clicks are concentrating on cited sources. Pew Research, using browsing data from March 2025, found people clicked a traditional result on 8 percent of visits to search pages carrying an AI summary, against 15 percent when there was not one. The clicks that remain flow disproportionately to the sources the AI names.

2. Ranking no longer guarantees citation. Ahrefs looked at 863,000 SERPs in January 2026 and found that 37.9 percent of cited URLs also appeared in the first ten organic results, down from roughly 76 percent in July 2025. BrightEdge, measuring differently, puts the overlap as low as 17 percent. Either way, citation has become a separate visibility layer from ranking. You can rank third and not be cited, or rank on page two and be cited.

3. AI visitors convert. Traffic from AI is still a small share of total visits, but it is high intent. Adobe found AI traffic to retail sites converted 42 percent better than other traffic in March 2026, reversing a 38 percent deficit a year earlier. Someone who arrives after an AI recommended you has already been partly sold.

Sources: Google I/O 2026, Pew Research, Ahrefs, BrightEdge, Adobe Q2 2026 AI Traffic Report.

How AI engines choose which sources to cite

Understanding the mechanics removes most of the mystery. Most AI search tools follow roughly the same pipeline:

1. Query understanding. The engine interprets the question and often splits it into sub queries, which Google calls query fan out. Best CRM for a small plumbing business might become separate searches for CRM features, pricing, trade specific tools and reviews.
2. Retrieval. It searches an index for pages matching each sub query.
3. Passage selection. It pulls the specific passages that best answer each part. Not whole pages: chunks.
4. Synthesis. The model writes an answer combining those passages.
5. Attribution. It cites the sources it relied on most.

An AI engine is a researcher writing a report on a deadline. It skims dozens of sources, grabs the paragraphs that answer each question cleanly, and cites the ones it trusted. Your job is to be the easiest, most trustworthy source to quote.

That has practical consequences. If your answer is buried in paragraph six after a long introduction, the researcher skims past. If your page makes a claim without evidence, it is riskier to quote. If nobody else on the web mentions you, it is harder to trust you.

What the research says works

GEO has a research base. The term comes from a 2023 paper by researchers at Princeton, Georgia Tech, the Allen Institute for AI and IIT Delhi, GEO: Generative Engine Optimization, later presented at KDD 2024. They tested content changes across thousands of queries to see which made a source more visible in generated answers.

The headline findings:

• Adding statistics, citing credible sources and including quotations were the strongest methods, improving visibility by up to around 40 percent in their tests. That is the ceiling rather than the average.
• Fluent, clear writing helped too.
• Keyword stuffing performed poorly. The tactic that built early SEO does little or nothing for AI.
• Lower ranked sites gained the most, which is good news for a small business competing against big brands.

The logic is intuitive. A model generating a factual answer prefers passages it can support. A specific number with a named source is safer to use than a vague claim.

The five layer GEO framework

Work through these in order, because each depends on the one before.

1. Be retrievable

If the engine cannot access your content, nothing else matters.

• Allow AI crawlers in your robots.txt. Many sites block them by accident through security plugins. Check for GPTBot, OAI-SearchBot, PerplexityBot, ClaudeBot and Google-Extended.
• Get indexed in both Google and Bing. Several AI tools rely on Bing's index.
• Do not hide content behind JavaScript. Many AI crawlers do not render it well. Server rendered HTML is safest.
• Fix the SEO basics: speed, clean structure, working internal links. The SEO for small business playbook covers the foundations, and technical SEO is the service version.

2. Be extractable

Engines lift passages. Write so every section can stand on its own.

• Answer first. Put a direct two or three sentence answer immediately under each heading, then expand.
• Use question headings that match how people phrase queries.
• One idea per section. A section covering three topics gets pulled for none of them.
• Use tables and lists for comparisons, specs, steps and pricing. They extract accurately.
• Define things plainly. X is a Y that does Z is exactly what engines quote.
• Avoid pronoun dependence. A sentence that needs the one above it to make sense does not survive extraction.

Before and after:

Before: When it comes to pricing, there are lots of factors to think about, and every project is different, so it is hard to give an exact figure without seeing the property.

After: A dormer loft conversion in the West Midlands typically costs 45,000 to 65,000 pounds in 2026. The main cost drivers are roof structure, staircase position and whether you add an en suite.

The second version is quotable. The first gives an engine nothing to use.

3. Be citable

• Include specific statistics with named sources and dates.
• Cite authoritative sources where you make claims.
• Add original data. Your own job data, customer survey results or pricing analysis is content nobody else has, which makes you the only possible source.
• Show expertise: named authors with real credentials, firsthand experience, photos of actual work.
• Keep it current, with a visible last updated date. Stale numbers get dropped.

4. Be corroborated

This is the layer most businesses miss. AI engines do not just read your site, they check what the rest of the web says about you. A business mentioned consistently across reviews, directories, press, forums and industry sites looks far more credible than one that only talks about itself.

• Earn mentions on third party sites: industry roundups, best of lists, trade publications, local press. That is digital PR work.
• Be present where discussions happen. Forums are heavily cited by several engines. Genuine participation, not spam.
• Keep your details consistent everywhere. Conflicting information makes engines less confident recommending you.
• Build reviews on multiple platforms. The local SEO guide covers this in depth.

5. Be an entity

Engines understand the world in terms of entities, meaning people, companies and places, and the relationships between them. Make yours unambiguous.

• Add Organization schema with your name, logo, contact details and sameAs links to your profiles and listings.
• Add Person schema for authors, linking to their profiles.
• Use your brand name consistently across every platform.
• Build a clear About page stating plainly who you are, what you do, where, and since when.

What does not work, or is not worth your time yet

• Keyword stuffing. The research is clear: it does not help.
• Mass producing thin AI content. Engines are built to find the best source, and a thousand interchangeable pages do not beat one definitive one.
• Hidden instructions aimed at AI crawlers. It is manipulation, it is detectable, and it risks your whole site's credibility.
• llms.txt as a priority. It is a proposed standard for guiding AI tools to your key content. Cheap to add, but no major AI search engine has confirmed it influences citations. Do it last, not first.
• Treating GEO as separate from SEO. Every retrievable, extractable, citable page is also a better SEO page. Run them as one strategy.

How to measure GEO

GEO needs different KPIs from traditional SEO:

• Citation rate: how often you are cited for your target prompts
• Mention rate: how often you are named, even without a link
• AI share of voice: your mentions compared with competitors for the same prompts
• Sentiment: whether AI describes you accurately and positively
• AI referral traffic and conversions: visits and leads from the assistants

I walk through a full measurement setup, including prompt tracking and a GA4 configuration you can copy, in how to rank in ChatGPT and AI Overviews.

Where to start this week

1. Check your robots.txt is not blocking AI crawlers.
2. Verify your site in Bing Webmaster Tools.
3. Pick your five most important pages and rewrite the first paragraph under each heading as a direct answer.
4. Add one statistic with a named source to each.
5. Ask ChatGPT and Perplexity your ten most important customer questions and record who they cite.

That last step usually tells you more about your GEO position than any tool will.

Frequently asked questions