Generative Engine Optimization (GEO) is the practice of structuring and evidencing your content so that AI search tools such as ChatGPT, Perplexity and Google AI Overviews cite, quote or recommend your business inside their answers. It matters because these tools now answer many questions directly, and being named in the answer is the new form of search visibility.
What is GEO and why does it matter now
GEO is the discipline of earning a place inside an AI-generated answer, rather than a position in a ranked list of links. The researchers who coined the term describe generative engines as systems that answer queries by synthesising information from multiple sources and summarising them with a large language model. When the answer itself names your firm, quotes your guidance or links your page, you are visible even if the user never scrolls to a traditional result.
The audience for these tools is now large. OpenAI's own usage research reports that ChatGPT passed 100 million logged-in weekly active users after one year and had more than 700 million by the end of July 2025 (OpenAI, September 2025). On the Google side, AI Overviews reached over 2 billion monthly users by mid-2025, and Alphabet said in July 2026 that AI Mode had surpassed 1 billion monthly active users and that AI features in Search send "billions of clicks to websites every week" (Alphabet, July 2026). For a buyer researching a supplier, the first summary they read is increasingly written by a model.
How generative engines choose which sources to cite
Generative engines retrieve a set of candidate pages, then write one answer and attach citations to the sources they relied on. Google describes a "query fan-out" technique in which AI Overviews and AI Mode issue multiple related searches across subtopics and data sources to develop a response, which can surface a wider set of supporting links than a single query would.
Two practical consequences follow:
- Retrieval comes first. If a crawler cannot reach your page, the model cannot cite it. OpenAI says sites opted out of OAI-SearchBot will not be shown in ChatGPT search answers, and Perplexity recommends allowing PerplexityBot in robots.txt so a site can appear in its results.
- Content quality decides inclusion. Among retrieved pages, the ones that state a clear answer, back it with evidence and are easy to quote are better candidates for the final response. The GEO research below is the best public evidence on which edits move the needle.
What the GEO research actually found
The GEO paper by Aggarwal and colleagues (authors from Princeton University, IIT Delhi and other institutions; accepted to KDD 2024) tested nine content edits on GEO-bench, a benchmark of 10,000 queries, and measured how much of each source made it into generated answers (full paper). Its headline result is that the best methods raised visibility by up to 40%, and a test on Perplexity.ai showed improvements of up to 37%.
| Content edit tested | What it means | Reported effect (GEO paper) |
|---|---|---|
| Cite Sources | Add citations to credible sources | Among the top methods: 30–40% relative improvement on position-adjusted word count |
| Quotation Addition | Add relevant quotes from credible sources | Among the top methods (same 30–40% range) |
| Statistics Addition | Replace qualitative claims with quantitative statistics | Among the top methods (same 30–40% range) |
| Fluency and easy-to-understand edits | Improve readability and simplify language | Smaller gains than the top three |
| Keyword Stuffing | Add more query keywords, as in classic SEO | "Little to no improvement" |
Source: Aggarwal et al., GEO: Generative Engine Optimization, arXiv 2311.09735. The paper also reports 15–30% relative improvement for the top methods on its subjective impression metric.
One finding is especially relevant to smaller firms. When all sources were optimised, lower-ranked websites gained the most: the Cite Sources method produced a 115.1% visibility increase for sites ranked fifth in the search results, while the top-ranked site's visibility fell by 30.3% on average. The authors suggest this gives independent businesses a way to compete with larger players who dominate traditional rankings.
Treat these numbers with care. The experiments used a fixed setup (five Google results per query and a GPT-3.5 model generating the answer), and the authors note that effectiveness varies by domain. Commercial engines change frequently, so the direction of the findings is more durable than the exact percentages.
How GEO differs from SEO
GEO is an extension of SEO, not a replacement for it. Google is explicit that the same fundamentals apply and that there is no special schema.org markup, AI text file or other machine-readable file needed to appear in its AI features (Google Search Central). What changes is the unit of success and the kind of content that wins.
| Dimension | Traditional SEO | GEO |
|---|---|---|
| Goal | Rank a page highly in a list of results | Be quoted, cited or recommended inside an AI-written answer |
| Main measure | Ranking position, impressions, clicks | How often and how prominently answers cite or mention you |
| What helps most | Relevance, links, technical health | All of SEO, plus clear answers, citations, quotations and statistics |
| Keyword stuffing | A violation of Google spam policies | Little to no effect on generative answers (GEO paper) |
| Crawler access | Googlebot and Bingbot | Also OAI-SearchBot, PerplexityBot and similar search crawlers |
| Prerequisite | Indexable pages | Indexable pages that are eligible for a snippet (Google) |
A practical GEO checklist for Singapore businesses
For most Singapore SMEs, GEO starts with five actions that also improve conventional SEO and the experience of human readers.
- Open every key page with the answer. Put a 40–60 word, self-contained answer to the page's main question in the first paragraph, so a model can quote it without surrounding context.
- Add evidence that can be checked. The GEO paper's top three methods were citations, quotations and statistics. Link each figure to its primary source and state its date. Never invent or round up a number to sound authoritative.
- Allow search crawlers you want to appear in. Review robots.txt and any firewall rules so that Googlebot, Bingbot, OAI-SearchBot and PerplexityBot can reach public pages. Blocking AI training is a separate decision: OpenAI documents GPTBot, a different crawler used for training, and the settings are independent.
- Describe your organisation consistently. Use the same legal name, address, services and contact details across your site and directory listings. JSON-LD markup such as LocalBusiness, Service, FAQPage and Article gives search engines "explicit clues about the meaning of a page" (Google), though Google says it is not a special requirement for AI features.
- Cover a topic in depth. Publish connected pages that answer the follow-up questions a buyer asks (cost, timeline, risks, alternatives). Query fan-out means a cluster of focused pages gives an engine more chances to find a relevant passage.
Some teams also publish an llms.txt file, a proposed markdown summary for language models. It is optional: Google's guidance says no AI text files are needed for its AI features.
How to measure GEO results
There is no single GEO ranking report, so measurement combines search data with structured spot checks. Google advises verifying your site in Search Console, and sites appearing in AI Overviews and AI Mode are included in the overall search traffic shown in its Performance report (Google Search Central).
- Keep a fixed list of 20–30 buyer questions and check monthly whether ChatGPT, Perplexity and Google's AI features mention or link your firm.
- Track referral traffic from AI assistants in your analytics tool, alongside Search Console clicks and impressions.
- Note which pages get cited and what they have in common, then apply the pattern to weaker pages.
Expect variation: generative answers differ between sessions and change as models are updated, so look at trends over several months rather than single checks.
Where to start
A good first step is an honest audit of your ten most commercially important pages: does each one open with a direct answer, cite its evidence, load fast and allow the right crawlers? Namtech Pte Ltd, a Singapore AI transformation consultancy, builds these principles into its corporate website work and covers AI search readiness in the AI Opportunity Audit. For a related view of how AI is changing operations, see AI agents for business.
GEO rewards the same things good buyers reward: a clear answer, checkable evidence and a source worth trusting.