---
title: "What Is GEO? Generative Engine Optimization, Explained"
description: "GEO is how content gets cited by ChatGPT, Perplexity and AI Overviews. Origin of the term, tactics that lift presence inside answers up to 40%, where to start."
lang: en
canonical: https://upgradepro.eu/blog/what-is-geo
published: 2026-07-16
tags: ["GEO", "content"]
---

# What Is GEO? Generative Engine Optimization, Explained

> GEO is how content gets cited by ChatGPT, Perplexity and AI Overviews. Origin of the term, tactics that lift presence inside answers up to 40%, where to start.

Generative engine optimization (GEO) is the practice of structuring web content so that AI systems — ChatGPT, Perplexity, Google's AI Overviews — retrieve it, use it and cite it in their generated answers. The term comes from a November 2023 academic paper (arXiv 2311.09735, published at KDD 2024), which showed that tactics such as citing sources, adding quotations and including statistics can raise a page's presence inside a generated answer by up to 40% — measured as Position-Adjusted Word Count, not traffic or clicks.

## What does GEO actually mean?

GEO means optimizing content for generative engines: systems that answer questions directly instead of returning a list of links. The [original paper](https://arxiv.org/html/2311.09735v3) defines a generative engine as a system that takes a query, retrieves relevant documents and generates a natural-language answer grounded in those sources — combining a conventional search engine with generative models. ChatGPT search, Perplexity, Google's AI Overviews and AI Mode all fit that definition.

The practical difference from classic SEO is the goal. SEO competes for a position on a results page and a click. GEO competes for presence inside the answer itself: being the source the engine quotes, paraphrases or links when a customer asks a question your business can answer.

## Where does the term GEO come from?

GEO was coined in the paper "GEO: Generative Engine Optimization", submitted to [arXiv](https://arxiv.org/abs/2311.09735) on November 16, 2023 and published at the KDD 2024 conference. The authors — Pranjal Aggarwal, Vishvak Murahari, Tanmay Rajpurohit, Ashwin Kalyan, Karthik Narasimhan and Ameet Deshpande — span IIT Delhi, Princeton and other institutions. The study is popularly cited as "the Princeton paper", although the first author is from IIT Delhi.

The research is unusually concrete for a marketing topic. The team built GEO-bench, a benchmark of 10,000 queries (8,000 for training, 1,000 for validation, 1,000 for testing) drawn from 9 datasets across 25 subject domains, and ran 9 optimization methods against it.

One nuance matters before repeating the headline number. The paper reports that GEO can boost visibility by up to 40% in generative engine responses — but visibility here means Position-Adjusted Word Count: how much of the generated answer draws on your content, weighted by where it appears. It is not a claim about traffic or clicks. Any vendor quoting "40% more traffic with GEO" is misreading the source.

## Which GEO tactics work — and which do not?

Three methods clearly won in GEO-bench: Cite Sources, Quotation Addition and Statistics Addition, with relative improvements of 30-40% in Position-Adjusted Word Count, according to the [GEO paper](https://arxiv.org/html/2311.09735v3) (June 2024). Old-school keyword stuffing delivered little to no improvement. On that last point, the authors are blunt:

> "while widely used for Search Engine Optimization, we find such methods offer little to no improvement"
> — Aggarwal et al., GEO paper, June 2024

| Tactic tested in GEO-bench | Result (Position-Adjusted Word Count) | Verdict |
| --- | --- | --- |
| Cite Sources — add citations to credible sources | Top-3 method, 30-40% relative gain | Do it |
| Quotation Addition — add quotations from sources | Top-3 method, 30-40% relative gain | Do it |
| Statistics Addition — add sourced statistics | Top-3 method, 30-40% relative gain | Do it |
| Keyword stuffing | Little to no improvement | Skip it |

None of the winning tactics is exotic. They describe one editorial style: evidence-dense pages where claims are attributed, numbers have named origins, and quotations carry authority. That is a writing discipline, not a technical trick.

## Why does GEO favor smaller websites?

Because generative engines redistribute visibility instead of reinforcing the existing ranking. In the paper's experiments with the Cite Sources method, websites sitting in position 5 improved their visibility by 115.1%, while websites in position 1 lost 30.3%.

> "Lower-ranked websites, which typically struggle for visibility, benefit significantly more from GEO."
> — Aggarwal et al., GEO paper, June 2024

Vendor data points the same way, with the usual caveat attached. In its own study, [Semrush](https://www.semrush.com/blog/ai-search-seo-traffic-study/) (July 2025) — an SEO tools vendor, not a neutral party here — reports that ChatGPT search cites webpages ranking in positions 21 or worse in traditional organic search for related queries almost 90% of the time. AI visibility does not replicate the classic ranking. For an SME that will never outrank a marketplace or a national publisher on Google, this is the single most important finding in the literature.

## How big is AI search right now?

Still smaller than classic Google search, but growing quickly. Sam Altman announced at OpenAI DevDay on October 6, 2025 that ChatGPT had passed 800 million weekly active users — up from 700 million in August 2025 and 500 million in March 2025, as reported by [TechCrunch](https://techcrunch.com/2025/10/06/sam-altman-says-chatgpt-has-hit-800m-weekly-active-users/). Perplexity processed 780 million queries in May 2025, growing 20% month over month, according to CEO Aravind Srinivas, as covered by [Search Engine Land](https://searchengineland.com/perplexity-780-million-monthly-queries-month-456725) (June 2025).

Predictions deserve more distance than measurements. A [Gartner](https://www.gartner.com/en/newsroom/press-releases/2024-02-19-gartner-predicts-search-engine-volume-will-drop-25-percent-by-2026-due-to-ai-chatbots-and-other-virtual-agents) press release from February 2024, as reported by the trade press, predicted that traditional search engine volume would drop 25% by 2026 under pressure from AI chatbots and virtual agents. That was a forecast, not a measurement, and later analyses dispute whether it came true — read it as a signal of where the market believed things were heading, nothing more.

In Europe the shift has a precise date: Google switched on AI Overviews in Italy, Spain, Germany, Austria, Belgium, Ireland, Poland, Portugal and Switzerland on March 26, 2025, per [Google's official blog](https://blog.google/intl/it-it/prodotti/arriva-ai-overview-anche-in-italia/). Adoption still has room to run: ISTAT's Rapporto Annuale 2026 found only 19.9% of Italians aged 16-74 used AI in 2025, versus an EU27 average of 32.7%, as reported by [Primaonline](https://www.primaonline.it/2026/05/21/473913/povera-italia-penultima-in-ue-per-uso-ai-tutti-sui-social-e-online-lavoro-trainato-da-over50-che-limitano-innovazione/) (May 2026). Positioning early is cheap while competitors wait.

## What happens to traffic when AI answers first?

Fewer clicks, better visitors — and both effects are measured, not guessed. [Pew Research Center](https://www.pewresearch.org/short-reads/2025/07/22/google-users-are-less-likely-to-click-on-links-when-an-ai-summary-appears-in-the-results/) (July 2025) tracked 900 US adults and 68,879 searches from March 2025. When an AI summary appeared, users clicked a traditional result in 8% of visits, versus 15% without a summary. Links cited inside the AI Overview were clicked in just 1% of visits. AI summaries appeared on 18% of the searches studied.

The visitors who do click through are worth more. According to [Semrush](https://www.semrush.com/blog/ai-search-seo-traffic-study/) (July 2025), the average AI search visitor is 4.4 times as valuable as the average visit from traditional organic search, measured by conversion rate. The same study projects that, for digital marketing topics, AI search could overtake traditional search in visits by early 2028.

The takeaway is not panic — it is instrumentation. If you do not know how often AI engines mention your business today, you cannot tell whether these shifts help or hurt you. We cover how to build that baseline without enterprise tooling in [how to measure your AI visibility](/blog/measure-ai-visibility).

## Does GEO replace SEO?

No. Indexing and snippet eligibility remain the entry ticket, and Google's own documentation is explicit about it:

> "There are no additional requirements to appear in AI Overviews or AI Mode, nor other special optimizations necessary."
> — [Google Search Central](https://developers.google.com/search/docs/appearance/ai-features), AI Features and Your Website

The same document notes that Google's AI features use query fan-out — running multiple related searches for one query — which surfaces a more diverse set of links than a classic results page. Existing controls such as nosnippet, max-snippet and noindex apply to AI features too.

So the technical foundation is unchanged; what wins on top of it has changed. Keyword-first copy loses to evidence-dense, answer-first pages. For a full side-by-side of what carries over and what does not, see [GEO vs SEO: what actually changes](/blog/geo-vs-seo).

## Is it called GEO, AEO or LLMO?

Several acronyms circulate, and they largely name the same work. [SEOZoom](https://www.seozoom.it/acronimi-seo/) (February 2026) catalogued the terminology in the Italian market: GEO, AEO and AIO are the most commercially adopted, with SEO for AI, LLMO and AISO also in circulation — the last one emerging in job postings.

| Acronym | Expansion | Status per SEOZoom (Feb 2026) |
| --- | --- | --- |
| GEO | Generative engine optimization | Most commercially adopted |
| AEO | Answer engine optimization | Most commercially adopted |
| AIO | Artificial intelligence optimization | Most commercially adopted |
| SEO for AI | Descriptive phrase, not an acronym | In circulation |
| LLMO | Large language model optimization | In circulation |
| AISO | Emerging label, expansion not settled | Appearing in job postings |

We use GEO because it comes from the primary literature — the KDD 2024 paper — rather than from a vendor's naming exercise. Pick one label and be consistent; the work underneath is identical.

## What are the three layers of GEO work?

In practice, GEO breaks down into three layers: machine-readable facts, citable content and measurement. It is the same backbone as [our method](/method) — baseline audit, implementation, before-and-after measurement.

1. **Machine-readable facts.** Make sure engines can crawl and parse you: indexed pages, snippet eligibility, clean structure. Optionally add llms.txt — a proposal published by Jeremy Howard of Answer.AI on September 3, 2024 ([llmstxt.org](https://llmstxt.org/)): a Markdown file at /llms.txt giving LLMs a curated index of your site at inference time. It is a proposal, not an adopted standard — treat it as a low-cost bet, never a requirement.
2. **Citable content.** Apply the paper's three winning tactics as an editorial standard: named sources, direct quotations, verifiable statistics on every commercial page. We break the format down in [how to write citable content](/blog/citable-content).
3. **Measurement.** Record who the engines cite for your key questions before you change anything, then re-measure after. Label every number honestly — measured, calculated or assumed — so you can tell real movement from wishful thinking.

## What should you do this week?

A baseline you can build in one afternoon, without buying any tool:

1. **Ask the engines your customers' questions.** Put your five most valuable buyer questions to ChatGPT, Perplexity and Google, and record who gets cited. That record is your baseline.
2. **Confirm eligibility.** Google requires nothing extra for AI Overviews — verify in Search Console that your key pages are indexed and snippet-eligible.
3. **Upgrade one page with the paper's tactics.** Add named sources, one quotation and two or three sourced statistics to your most important commercial page.
4. **Write answers, not keywords.** Turn headings into real questions and answer each one in its first two sentences. The GEO paper found keyword stuffing offers little to no improvement.
5. **Re-run the baseline in 30 days.** Same questions, same engines. Compare citations before and after — that difference is your GEO result, measured.

If you would rather have the baseline done for you — measured, sourced and labeled — that is exactly what our [AI visibility audit](/services/ai-visibility) delivers.

## Frequently asked questions

### What is GEO (generative engine optimization)?

GEO is the practice of optimizing content so that generative engines such as ChatGPT, Perplexity, Gemini and Google AI Overviews use it and cite it in their answers. The term comes from an academic paper submitted in November 2023 (arXiv 2311.09735, a Princeton and IIT Delhi collaboration published at KDD 2024), which defines a generative engine as a system that retrieves sources and generates a natural-language answer grounded in them.

### Does GEO replace SEO?

No. Google states in its official documentation that there are no additional requirements to appear in AI Overviews or AI Mode: a page needs to be indexed and eligible for snippets, exactly as in classic SEO. What changes is which tactics pay off. The GEO paper found that adding source citations, quotations and statistics improved visibility by 30-40%, measured as presence inside the generated answer (Position-Adjusted Word Count) rather than traffic or clicks, while traditional keyword stuffing offered little to no improvement.

### How much can GEO improve visibility?

The original paper measured improvements of up to 40% in how much of a generative answer draws on the optimized content, using a metric called Position-Adjusted Word Count. That is presence inside the answer, not traffic or clicks. The most relevant finding for small businesses: websites in position 5 gained 115.1% visibility with the cite-sources method, while sites in position 1 lost 30.3%.

### Is AI search worth it if it still sends little traffic?

The volume is still smaller than classic Google search, but it is growing fast: ChatGPT passed 800 million weekly active users in October 2025, and Perplexity processed 780 million queries in May 2025 alone. Visitors arriving from AI search also convert far better: 4.4 times as valuable as the average traditional organic visit, according to Semrush. Meanwhile, Pew found that when an AI summary appears, clicks on traditional results drop from 15% to 8% of visits.

### What is the difference between GEO, AEO and LLMO?

They largely describe the same work with a different emphasis. GEO stands for generative engine optimization, AEO for answer engine optimization, and LLMO for large language model optimization. According to SEOZoom, GEO, AEO and AIO are the most commercially adopted acronyms in the Italian market, with SEO for AI, LLMO and AISO also in circulation.

### When did AI Overviews launch in Europe?

Google rolled out AI Overviews to Italy, Spain, Germany, Austria, Belgium, Ireland, Poland, Portugal and Switzerland on March 26, 2025. AI adoption in Italy remains low: according to Primaonline, reporting on ISTAT's Rapporto Annuale 2026, only 19.9% of Italians aged 16-74 used artificial intelligence in 2025, against an EU27 average of 32.7%, which leaves an early-mover window for businesses that position themselves now.

## Sources

Every claim in this article links to the source it comes from, with its publication date.

- [Aggarwal et al., GEO: Generative Engine Optimization (arXiv 2311.09735, KDD 2024)](https://arxiv.org/abs/2311.09735) — 2024-06-28
- [Aggarwal et al., GEO paper — full text (HTML v3)](https://arxiv.org/html/2311.09735v3) — 2024-06-28
- [Pew Research Center — Google users are less likely to click on links when an AI summary appears](https://www.pewresearch.org/short-reads/2025/07/22/google-users-are-less-likely-to-click-on-links-when-an-ai-summary-appears-in-the-results/) — 2025-07-22
- [Semrush — AI Search SEO Traffic Study](https://www.semrush.com/blog/ai-search-seo-traffic-study/) — 2025-07-21
- [Google Search Central — AI Features and Your Website (living document, consulted July 2026)](https://developers.google.com/search/docs/appearance/ai-features) — 2025
- [Google Italy official blog — AI Overviews arrive in Italy and Europe](https://blog.google/intl/it-it/prodotti/arriva-ai-overview-anche-in-italia/) — 2025-03-26
- [TechCrunch — Sam Altman says ChatGPT has hit 800M weekly active users](https://techcrunch.com/2025/10/06/sam-altman-says-chatgpt-has-hit-800m-weekly-active-users/) — 2025-10-06
- [Search Engine Land — Perplexity hits 780 million monthly queries](https://searchengineland.com/perplexity-780-million-monthly-queries-month-456725) — 2025-06-06
- [Gartner press release — search engine volume predicted to drop 25% by 2026](https://www.gartner.com/en/newsroom/press-releases/2024-02-19-gartner-predicts-search-engine-volume-will-drop-25-percent-by-2026-due-to-ai-chatbots-and-other-virtual-agents) — 2024-02-19
- [llmstxt.org — the /llms.txt proposal (Jeremy Howard, Answer.AI)](https://llmstxt.org/) — 2024-09-03
- [SEOZoom — SEO for AI, GEO, AEO: definitions and acronyms](https://www.seozoom.it/acronimi-seo/) — 2026-02-05
- [Primaonline, on ISTAT Rapporto Annuale 2026 — AI use in Italy vs EU27](https://www.primaonline.it/2026/05/21/473913/povera-italia-penultima-in-ue-per-uso-ai-tutti-sui-social-e-online-lavoro-trainato-da-over50-che-limitano-innovazione/) — 2026-05-21