Home / GEO and AI search visibility / Generative engine optimisation
generative engine optimization
Generative engine optimisation
Generative engine optimisation, usually shortened to GEO and spelt generative engine optimization in US usage, is the practice of making content more likely to be retrieved, used and cited by AI systems that write answers.
The term was popularised by a 2023 research paper by researchers including a Princeton team that tested how changes to page content affected visibility in generated answers.
Understand how generative answers use web pages, which page changes have evidence behind them and where the claims run ahead of the evidence.
In the paper's benchmark, adding citations, quotations from relevant sources and statistics were among the methods that most improved source visibility in generated answers. arXiv: GEO: Generative Engine Optimization (Aggarwal et al.) ↗
Google states there are no additional requirements or special optimisations to appear in AI Overviews or AI Mode; supporting links must be indexed and eligible for a snippet. Google Search Central: AI features and your website ↗
New to the topic? Read the background on artificial intelligence.
generative engine optimization
what is generative engine optimisation?
Optimising content for systems that generate an answer rather than list links.
A traditional results page lists links and lets the searcher choose; a generative engine reads sources and writes a single answer, sometimes with citations.
GEO asks what makes a page one of the sources that answer is written from, and whether the answer credits it.
Note that 'what is GEO' is also a common geography and satellite-orbit search, so this page uses the full phrase.
Generative engines write answers.
Sources may or may not be cited.
GEO targets inclusion and credit.
generative engine optimization
how do AI assistants choose which pages to use?
Most retrieve candidate pages with a search step, then write from the most relevant passages.
Assistants with web access typically run a search, fetch several pages and pass passages to the model.
That means ordinary search visibility, crawler access and fast, clean HTML still matter; a page that cannot be fetched cannot be quoted.
OpenAI, for example, documents separate crawlers for training and for search, and site owners can allow or block them individually.
Perplexity likewise documents PerplexityBot as the crawler that surfaces sites in its answers, so generative engine optimisation starts with letting the right crawlers in.
generative engine optimization
what does the research say helps?
The original GEO study found citations, quotations and statistics improved visibility in its tests.
The 2023 GEO paper by Aggarwal and colleagues reported that adding cited sources, quotations from relevant authorities and specific statistics improved how often content appeared in generated answers in its benchmark.
Those findings come from a research setting, not from any single commercial assistant, so treat them as direction rather than rules.
They also align with ordinary editorial quality: specific, sourced, clearly structured pages.
The paper, GEO: Generative Engine Optimization, is the most cited evidence base for generative engine optimisation, and its strongest methods were statistics, quotations and cited sources.
| Check | What to establish |
|---|---|
| 1 | Cite primary sources. |
| 2 | Use specific, attributable figures. |
| 3 | Structure answers clearly. |
generative engine optimization
what should a GEO page look like?
Question as heading, direct answer first, evidence next, detail last.
Each section opens with the question a buyer would type or speak, answers it in one or two sentences, then supports the answer.
Every factual claim carries a link to a primary source, and the page states its publish and update dates.
predeploy also serves each page as Markdown and plain text, which removes navigation and scripts for machine readers; this is our practice, not a requirement any provider has published.
Google also notes in its structured data introduction that markup clarifies meaning without guaranteeing a display, which is the right expectation for generative engine optimisation too.
generative engine optimization
where does generative engine optimisation run ahead of the evidence?
Claims about secret ranking signals, guaranteed citations and special files.
No major assistant publishes a ranking formula for citations.
Proposals such as llms.txt are community conventions; adopting one is cheap, but no provider has said it changes answers.
For the commercial side, see what a GEO agency should and should not promise, and how GEO vs SEO fit together.
No published citation formula.
llms.txt is a proposal.
Be wary of guarantees.
Make the decision concrete
What does the current task cost in time?
Choose a task connected to generative engine optimisation and enter its current workload. This gives you a baseline to discuss before buying training or commissioning a build.
A baseline is useful even if you decide to keep the process manual. Measure the work that matters, then compare quality and review effort as well as time.
Assess the workflow before changing it, or plan business AI training if the gap is how people use existing tools.
Your current workload
Estimate how much time the existing task takes before discussing generative engine optimisation.
16 hours per week
People × minutes × weekly tasks ÷ 60. A workload baseline from your inputs, not a prediction of AI time savings. Include review and support time when comparing a proposed workflow.
A useful AI investment leaves your team able to do the work and judge the result.
Continue with the right question
Where should you go next?
Use the overview for context, then follow the pages that match your delivery decision.
See our ai powered gtm service or read the geo and ai search visibility overview.
FAQ
Questions, answered
Evidence and further reading
Which sources support the guidance?
Source links support the specific background described here. The practical checklists and illustrative examples are predeploy’s guidance.
In the paper's benchmark, adding citations, quotations from relevant sources and statistics were among the methods that most improved source visibility in generated answers. arXiv: GEO: Generative Engine Optimization (Aggarwal et al.) ↗
Google states there are no additional requirements or special optimisations to appear in AI Overviews or AI Mode; supporting links must be indexed and eligible for a snippet. Google Search Central: AI features and your website ↗
Structured data gives Google explicit clues about page meaning; correct markup makes rich results possible but not guaranteed. Google Search Central: introduction to structured data ↗
llms.txt is a proposed, community-input convention for a Markdown file that helps AI agents use a website. llmstxt.org: the /llms.txt proposal ↗
OpenAI documents OAI-SearchBot for ChatGPT search and GPTBot for training, and states each robots.txt setting is independent. OpenAI: overview of OpenAI crawlers ↗
Perplexity documents PerplexityBot for surfacing and linking sites in search results, separate from user-triggered Perplexity-User fetches. Perplexity: Perplexity crawlers ↗
Background reading on the broader field of artificial intelligence. Wikipedia: artificial intelligence ↗
Choose your next step
Explore geo and ai search visibility
Start with a topic overview, or go directly to the question your team needs to answer.
GEO and AI search visibility
Be found, described and cited in AI answers as well as search.
Start with the decision
More questions in this topic
