Generative Engine Optimization (GEO): How to Get Cited by ChatGPT, Gemini & Google AI Overviews

Quick answer: Generative Engine Optimization (GEO) is the practice of making web content easy for AI systems to find, trust and quote, so that it is cited inside AI-generated answers from tools such as ChatGPT, Google Gemini, Google AI Overviews and Perplexity. It builds on traditional SEO rather than replacing it, but it measures success differently: by citations and mentions inside an answer, not only by rankings and clicks.

For most of the past two decades, visibility online meant ranking on a page of blue links. Today, millions of questions are answered directly by AI systems that read many sources, write a single response and credit only a handful of them. Ahrefs, an SEO software company, studied 300,000 keywords and found that when a Google AI Overview appears, the top-ranking page receives about 58% fewer clicks than it otherwise would.

That shift has produced a new set of questions: how to get cited by ChatGPT, how to appear in Google AI Overviews, and how to optimize for AI search without wasting effort on tactics that do not work. This guide brings together the available research, official guidance and industry data to explain what GEO is, how AI engines choose sources, and what the evidence says about each tactic.

What Is GEO?

Generative Engine Optimization is the set of techniques used to improve how often, and how favorably, content appears in responses generated by large language model (LLM) based search tools. Where traditional SEO aims for a position in a list of results, GEO aims for a citation or mention inside the synthesized answer itself.

The term was introduced in a 2023 research paper, GEO: Generative Engine Optimization, by researchers from Princeton University, Georgia Tech, the Allen Institute for AI and IIT Delhi. It was later presented at KDD 2024, a major data-mining conference. The paper defined generative engines as systems that retrieve content from the web, synthesize an answer with a language model, and attach citations to the sources used.

You may also see related terms, including answer engine optimization (AEO), LLM SEO and AI search optimization. They describe largely the same activity, and the differences are mostly a matter of labeling rather than method. One practical note: “GEO” also abbreviates “geographic” in some marketing contexts, so searches for the term can return local-search results.

Why GEO Matters Now

AI search has reached mainstream scale

At Google I/O in May 2026, Google said AI Overviews reach more than 2.5 billion monthly users and that AI Mode, its conversational search experience, had passed 1 billion monthly users roughly a year after launch. OpenAI reported 900 million weekly active users for ChatGPT in February 2026.

Clicks are changing

Several large studies point in the same direction: when an AI summary appears, fewer people click through to websites.

• Ahrefs compared December 2023 with December 2025 across 300,000 keywords and found AI Overviews correlate with a 58% lower click-through rate (CTR) for the top-ranking page. An earlier version of the study, from April 2025, had measured 34.5%. The negative effect extended down the page, from about 51% at position two to about 19% at position ten.
• Pew Research Center, analyzing the browsing of about 900 U.S. adults (68,879 searches) in March 2025, found people clicked a traditional search result in 8% of visits when an AI summary appeared, compared with 15% when it did not. Only about 1% of visits involved a click on a link inside the AI summary.
• Seer Interactive tracked 5.47 million queries across 53 brands and found organic CTR of 0.61% on queries with an AI Overview, versus 1.62% on queries without one. A later write-up of the same study notes that CTR on AI Overview queries partly recovered, from 1.3% in December 2025 to 2.4% in February 2026, so the trend is a decline with volatility rather than a straight line.

The studies differ in method and scope, so the exact figures vary. The direction, however, is consistent: ranking well is no longer a guarantee of traffic, which is why being cited inside the answer has become a separate goal.

How AI Engines Choose Sources

Understanding source selection is the foundation of any Generative Engine Optimization strategy. The research shows that AI citations follow different patterns from classic rankings, and that each engine behaves differently.

Rankings and citations only partly overlap

An Ahrefs analysis of 863,000 keywords and 4 million AI Overview URLs (February 2026) found that only 38% of pages cited in Google AI Overviews also ranked in the top 10 for the same query, down from 76% about seven months earlier. Other studies report different overlaps, and one benchmark of 12,500 queries found 83% of AI Overview citations came from pages outside the top 10. Methods and query sets differ, so the safest summary is that top-10 rankings and AI citations overlap far less than many marketers assume, and that the overlap has been shrinking.

Engines rarely agree with each other

• ai analyzed 379,321 Claude citations across 16,406 domains and found Claude and ChatGPT shared only 13% of cited domains and 4.2% of top 100 cited URLs.
• A separate analysis found ChatGPT and Google AI Overviews overlap on only about 14% of cited URLs.
• Muck Rack’s analysis of more than 25 million links found 96% of ChatGPT responses contained citations, compared with 82% for Gemini and 55% for Claude.

This means a page cited by one engine may not be cited by another, and visibility has to be assessed engine by engine.

Earned media and third-party mentions carry weight

Muck Rack’s Generative Pulse study found that 84% of the links cited by ChatGPT, Claude and Gemini in its May 2026 edition were earned media, with journalism alone accounting for 27%. Paid or advertorial content made up 0.3%. Muck Rack defines earned media broadly, to include journalism, academic research, government sources, encyclopedic sites and third-party corporate content, and its sample covers ChatGPT, Claude and Gemini only. Separately, Ahrefs reports that branded web mentions correlate most strongly with a brand appearing in AI answers (Spearman correlation 0.664).

Not every study agrees on who gets cited. A Yext analysis of 6.8 million citations (July-August 2025, local and brand-related queries across ChatGPT, Gemini and Perplexity) found 86% came from sources brands manage, such as their own websites, listings and reviews. The two findings look contradictory but measure different things: different query types, time periods and definitions of what counts as earned or brand-managed. Readers should treat each as specific to the study that produced it.

Placement and freshness also play a role

One analysis reported that content updated within 30 days received 3.2 times more AI citations than older content, and another reported that 40–60% of cited sources change from month to month on Google AI Mode and ChatGPT. These are industry figures rather than peer-reviewed results, but they are consistent with the idea that freshness and ongoing monitoring matter.

Different engines, different retrieval

• ChatGPT’s search features have been reported to draw heavily on Bing-style web results, and OpenAI’s own crawler (OAI-SearchBot) indexes pages for its search features.
• Google AI Overviews and AI Mode draw on Google’s own index, and Google describes using a technique called “query fan-out,” which issues multiple related searches across subtopics to build a response.
• Perplexity retrieves live results and shows inline citations.

What Google Says About GEO

On May 15, 2026, Google published its first consolidated guide for site owners, titled “Optimizing your website for generative AI features on Google Search.” Its central message is that optimizing for AI Overviews and AI Mode remains part of SEO, because those features rely on Google’s core ranking and quality systems, retrieval-augmented generation and query fan-out.

GEO vs SEO: where Google draws the line

The guide is the clearest official statement on GEO vs SEO for Google’s own products. According to coverage of the guide, it states that site owners do not need:

• special AI text files such as llms.txt,
• breaking content into tiny “AI-friendly” chunks,
• AI-only rewrites or special structured data, or
• any particular page length (Google says there is no ideal length).

The guide also notes that crawlability still matters, and it advises structuring pages for people using the same semantic HTML that works for machines.

What the guide does not cover

Google’s guidance applies to AI Overviews and AI Mode. It does not necessarily describe how the standalone Gemini app, ChatGPT, Claude or Perplexity choose sources, and it does not offer a way to control inclusion in AI features separately from regular search. That is the space where broader GEO research is still developing.

In practice, how to appear in Google AI Overviews comes down to the same fundamentals as ranking in Google Search: content that is indexed, helpful, accurate and well structured, from a source with demonstrable expertise.

What the Research Shows

The most cited study in the field is the original Princeton-led GEO paper. The researchers tested nine content-modification methods across 10,000 queries, using a custom retrieval pipeline and validating results on Perplexity.

• Adding quotations, statistics and cited sources produced the largest improvements, roughly 30–40% relative gains in one of the paper’s visibility metrics.
• Keyword stuffing performed worse than the unmodified baseline, by roughly 10% on Perplexity.
• Fluency and readability improvements helped, and combining fluency optimization with statistics did better than either alone.
• The authors also reported that lower-ranked sites tended to benefit more than sites already at the top.

Important caveats

The headline figure of “up to 40% more visibility” is often repeated without context. A 2026 critical survey of GEO research covering 2023–2026 describes it as a relative maximum on one metric under a specific configuration, not a general promise.

The survey also concludes that general heuristics transfer poorly across engines, that competition from other sites can erode gains, and that visibility should be measured across multiple engines, prompts and time periods rather than reduced to a single result.

Other 2026 research has focused on diagnosing why pages fail to get cited, grouping failures into parsing problems (such as malformed HTML), fetching problems (such as truncated content) and generation-stage problems (such as missing entity information or mismatch with user intent).

In short, the research supports a direction, which is to make content specific, sourced and easy to extract. It does not support guaranteed percentages.

GEO Tactics That Work

This section groups common tactics by strength of evidence. It is intended as a practical reference for anyone building a Generative Engine Optimization strategy, and for readers wondering how to optimize for AI search without relying on myths.

Stronger evidence

1. Add specific facts. Replace vague claims with statistics, dated figures and named sources. This is the best-supported lever in the Princeton research.
2. Include attributed quotations and citations. Credible third-party quotes and references were among the highest-performing changes in the original study.
3. Answer the question early. Given that a large share of ChatGPT citations came from the top third of pages, leading each section with a direct, self-contained answer makes extraction easier.
4. Write clearly. Fluent, well-organized text improved results in the research, and Google’s guidance also encourages writing for people first.

Moderate or indirect evidence

5. Build mentions beyond your own site. The Muck Rack data suggests coverage in editorial and third-party sources is heavily represented in AI citations.
6. Keep important pages current. Freshness figures vary by study, but regular updates support both traditional ranking and AI retrieval.
7. Cover a topic thoroughly. Because Google’s engines use query fan-out, a page or cluster that addresses related sub-questions has more chances to be retrieved.

Weak or unsupported

8. Keyword stuffing. Shown to hurt in the original research.
9. txt. See the technical section below; no measurable effect has been demonstrated.
10. Over-chunking. Google says breaking content into tiny pieces for AI is unnecessary.

For readers asking how to get cited by ChatGPT specifically, the evidence points to a combination of being accessible to OpenAI’s search crawler, publishing distinctive, well-sourced information, and being mentioned across credible third-party sources, rather than to any single trick.

Technical Checklist

Check crawler access

OpenAI operates separate crawlers with different purposes:

• GPTBot gathers content that may be used for model training.
• OAI-SearchBot indexes pages for ChatGPT’s search features.
• ChatGPT-User fetches pages when a user asks ChatGPT to open them.

Because these are controlled independently in robots.txt, blocking one does not necessarily block another. An independent scan of the top 5,000 websites found that 535 blocked GPTBot, and 238 of those also blocked OAI-SearchBot, which can remove a site from ChatGPT’s search results. OpenAI recommends allowing OAI-SearchBot if a site wants to be discoverable in its search features. Publishers who wish to opt out of training can block GPTBot alone.

Blocking is not a perfect barrier. A BuzzStream study of 4 million citations found that 82.4% of top news sites blocking OAI-SearchBot still appeared in the citation dataset, which the researchers suggested may reflect earlier indexing or other retrieval paths.

Sites using a firewall or content delivery network should also confirm that those tools are not blocking the crawlers they intend to allow.

Make content readable by machines

Crawlers may not execute all JavaScript. Important text should be present in the page’s HTML, with clean headings, descriptive titles and standard semantic markup. Research on citation failures lists malformed HTML, excess page noise and truncated content among the common reasons otherwise strong pages are not cited.

The llms.txt reality check

llms.txt is a proposed file that lists a site’s key content for AI systems. SE Ranking analyzed about 300,000 domains, found the file on roughly 10% of them, and found no measurable relationship between having it and being cited more often. Google has said it does not require such files for AI Overviews or AI Mode, and OpenAI’s documentation focuses on robots.txt controls. The file is low-cost to add but should not be expected to influence citations.

Example: allow search crawling, restrict training

User-agent: OAI-SearchBot
Allow: /

User-agent: GPTBot
Disallow: /

This pattern allows ChatGPT search indexing while opting out of training use. Each site should decide according to its own policy, and should verify the current crawler names in each provider’s documentation, since they can change.

Is AI Traffic Worth It?

Volume is still small

Ahrefs’ tracker of 74,752 websites found that all AI chatbots combined sent 3.5 million visits in March 2026, about 0.28% of total web traffic. ChatGPT accounted for most of it. Other sources, such as SE Ranking, put AI assistants at a similar fraction of a percent, with Google organic search still hundreds of times larger.

Quality can be higher

On Ahrefs’ own website, AI search visitors made up 0.5% of traffic but 12.1% of signups. Other analyses report that AI-referred visitors convert at higher rates than average organic visitors, though the size of the gap varies widely.

The evidence is mixed

• A Stripe cohort found B2B SaaS companies converting AI traffic at 2.7% versus 1.4% for organic, but the pattern reversed for e-commerce, where organic converted at 2.1% versus 1.6% for AI.
• Analysts have noted that higher conversion may reflect selection, since people who click through from an AI answer may already be further along in their decision, rather than proof that the channel itself is superior. No published study has fully controlled for this.
• Some AI-influenced visits arrive without a referrer and show up as direct traffic, so the true size of the channel is hard to measure.

The balanced reading is that AI referral traffic is small but growing, and its value depends heavily on industry and goal.

How to Measure GEO

Because there is no equivalent of a rank tracker that works identically across engines, measurement relies on several complementary methods:

1. Prompt tracking. Define a set of questions your audience is likely to ask, run them regularly across ChatGPT, Gemini, Perplexity, Claude and Google’s AI features, and record whether the brand or site is cited or mentioned. Results vary between runs, so repeated sampling is more reliable than a single check.
2. Referral analytics. In an analytics tool such as GA4, create segments for known AI referrers (for example, chatgpt.com and perplexity.ai) to track visits that arrive through citations.
3. Search Console and SERP monitoring. Watch impressions, clicks and the presence of AI Overviews for priority queries, keeping in mind that impressions can rise while clicks fall.
4. Brand mention monitoring. Track mentions in news, forums and review sites, since third-party sources feature prominently in AI citations.
5. Record the date, engine and prompt for every observation. Citation sets change often, and researchers recommend treating visibility as a measurement across engines, prompts and time rather than a single score.

30-Day GEO Action Plan

The following plan summarizes how the findings above can be turned into a practical Generative Engine Optimization strategy for a typical content site. It is a starting framework, not a guarantee of results.

Baseline and access: Week 1

• Review robots.txt and firewall rules for OAI-SearchBot, PerplexityBot and Google crawlers.
• Confirm key pages render fully in HTML.
• Build a list of 20–30 target questions and record which sources each engine currently cites.

Strengthen priority pages: Week 2

• Open each key page with a direct answer to its main question.
• Add dated statistics, attributed quotes and references to primary sources.
• Remove filler and keyword repetition.

Extend and refresh: Week 3

• Fill gaps in related sub-questions, since AI systems may retrieve content for follow-up queries.
• Update older pages with current figures and add visible update dates.
• Identify credible third-party publications, directories and communities relevant to the topic.

Re-test and compare: Week 4

• Re-run the target prompts and compare citations by engine.
• Check AI referral traffic and Search Console trends.
• Note which changes coincided with improvements, remembering that correlation is not proof.

Key Takeaways

• GEO is the practice of earning citations inside AI-generated answers, and it is built on the same foundations as good SEO: accessible, accurate, well-structured content.
• AI Overviews reach billions of users, and studies from Ahrefs, Pew Research and Seer Interactive consistently find fewer clicks to websites when an AI summary appears.
• Rankings and citations overlap only partly, and engines frequently cite different sources, so visibility must be assessed engine by engine.
• The strongest evidence supports adding specific statistics, attributed quotes and cited sources, and writing clear, direct answers. Keyword stuffing performed worse than doing nothing in the original research.
• Technical access matters: confirm that search-related crawlers such as OAI-SearchBot are not blocked, and do not expect llms.txt to change results.
• AI referral traffic is small, with the value of each visit depending on the industry, and measurement is imperfect.
• Treat any claimed percentage gain, including the well-known “up to 40%” figure, as context-specific rather than a guarantee.

Are you planning a conference to gain more geospatial insights? Connect with the DigitalConfex team today. 👉 Let’s Connect

Sources and Further Reading

• Aggarwal et al., GEO: Generative Engine Optimization, KDD 2024
• Google Search Central, Optimizing your website for generative AI features on Google Search (May 15, 2026) and AI features and your website
• Ahrefs, AI Overviews and click-through rate study (published February 2026) and AI Chatbot Traffic research
• Pew Research Center, study of Google users and AI summaries (July 2025)
• Seer Interactive, AI Overview CTR study (January 2025–February 2026)
• Muck Rack, Generative Pulse (May 2026)
• SE Ranking, llms.txt analysis of about 300,000 domains
• ai, analysis of Claude citations (June 2026)
• BuzzStream, study of AI citations and crawler blocking (March 2026)
• Tian et al., Diagnosing and Repairing Citation Failures in Generative Engine Optimization (arXiv:2603.09296)
• Optimizing Visibility in Generative Engines: A Critical Survey of Generative Engine Optimization (2023–2026) (arXiv:2607.14035)

Statistics in this article are drawn from the sources above and reflect data available as of October 2026. AI search products and crawler policies change frequently, so readers should verify current figures and documentation before acting on them.