Why GEO appeared
People increasingly get answers without visiting a site. AI Overviews answers inside the results
page; Perplexity and ChatGPT Search synthesise answers from several sources. In that model, holding
the first organic position is not enough — the goal is to be one of the sources the model trusts.
GEO is the discipline that answers the question: how do you get the model to cite your content?
How a model picks sources
Generative engines follow a similar pattern: take the query, retrieve candidate sources (through
conventional search or pretrained knowledge), then synthesise an answer from the fragments they
trust most.
The trust signals:
- Source authority — backlinks, mentions in reputable publications, E-E-A-T
- Fragment relevance — how precisely the opening sentences answer the question
- Structure — clear claims, facts with numbers, headings that match the question
- Technical accessibility — not blocked to crawlers, fast, with an llms.txt
Important: AI Overviews cites pages that already rank in the top ten far more often than pages
that do not. Good classical SEO is a necessary but insufficient condition for GEO.
Practical principles
Content
- Answer the question in the first two or three sentences of a paragraph, with no preamble. Models
extract fragments, so the opening matters most. - Specific data beats generalisation. A concrete benchmark with a number gets cited; a statement
that something is important for business does not. - Structure through FAQ sections: the questions in H2 or H3 should match what people actually ask.
- Authors with biographies and a visible professional background raise the page’s E-E-A-T.
The technical side
| Element | What to do |
|---|---|
| llms.txt | Publish a file at the root describing the site and the sections open to LLM crawlers |
| Schema.org | Mark up Article, FAQ, HowTo and Product — structured facts for the model |
| robots.txt | Do not block the AI crawlers if you want citations |
| Canonical | Ensure the canonical version of a page is the one that answers the query |
| Load speed | Core Web Vitals as an indirect authority signal |
Monitoring
Unlike a ranking position, share of voice in AI answers is not measured by standard SEO tools. The
approaches:
- A manual audit of 20–50 key questions across the main assistants
- Dedicated tools that track AI answer share over time
- Tracking brand and category queries specifically, since those are where commercial intent sits
GEO priorities in e-commerce
For an online store the highest-potential queries are category and comparison questions: which
laptop to choose for work, the best running shoes. That is where people are asking for advice and
where models synthesise answers most actively.
Concretely:
- Publish buying guides — expert material with selection criteria, not just listings
- Mark up comparison tables with Schema.org
- Publish FAQs on category pages with specific answers
- Keep content current — models prefer fresh sources