kimo

GEO: how to show up in AI answers — and measure it

Generative engine optimization (GEO) is the practice of making your content more likely to be retrieved, cited and accurately represented in AI-generated answers. The term comes from a KDD 2024 paper that found adding citations, quotations and statistics raised a source’s visibility in generative answers by up to about 40%. Google says no special optimization is needed for its AI features beyond good SEO. Measurement is partial today: you can count AI assistant referrals, Search Console clicks and some engine-reported citations, but not your share of every answer.

Sam Okafor
Growth lead12 min read8 sources

What is generative engine optimization?

Classic SEO optimizes for a ranked list of links. A generative engine retrieves several sources, then writes one synthesized answer with a handful of citations. Visibility stops being a position on a page and becomes a question of whether you are cited, how much of the answer draws on you, and where your citation appears. That shift is what the original research set out to measure. For the short definition, see our glossary entry on GEO.

What does the research actually say?

The paper GEO: Generative Engine Optimization by Aggarwal, Murahari, Rajpurohit, Kalyan, Narasimhan and Deshpande (accepted to KDD 2024) introduced the term and a benchmark, GEO-bench, of 10,000 queries across domains, each paired with web sources.1 The authors measured visibility with two metrics: Position-Adjusted Word Count (how much of the answer is drawn from a source, weighted by where it appears) and Subjective Impression (a model-judged score combining sub-metrics such as relevance, influence, uniqueness and perceived prominence).1

They then rewrote sources with nine strategies and measured the change. The headline result is that the best methods improved visibility by up to about 40% over the unmodified baseline.1

Position-Adjusted Word Count by GEO method (GEO-bench)
Figure. Source: Aggarwal et al., GEO: Generative Engine Optimization, Table 1 (GEO-bench test set). Higher means more of the generated answer drew on the optimized source.

Four findings matter for practitioners:

  • Evidence beats repetition. Quotation addition, statistics addition and citing sources were the strongest methods, while keyword stuffing scored below the baseline (17.8 vs 19.5).1
  • Lower-ranked sources gain the most. Citing sources raised visibility by 115.1% for sources ranked fifth in retrieval, but lowered it by 30.3% for top-ranked sources.1 GEO helps challengers more than incumbents.
  • Results transfer, imperfectly. On Perplexity.ai, quotation addition improved Position-Adjusted Word Count by 22%, and statistics addition improved Subjective Impression by 37%.1
  • Domain matters. The authors stress that effectiveness varies by domain, so there is no universal recipe.1

What does Google say about optimizing for AI Overviews?

Google’s documentation on AI features is unusually direct. There are no additional requirements to appear in AI Overviews or AI Mode, and no special optimizations are necessary. To be eligible as a supporting link, a page must be indexed and eligible to be shown in Search with a snippet.2 Google also says you do not need to create new machine-readable files, AI text files or special markup to appear in these features.2

What Google does recommend is ordinary, solid practice: allow crawling, use internal links, offer a good page experience, keep important content in text, and make structured data match the visible page.2 Underneath that sits its guidance on helpful content: content made primarily for people rather than to manipulate rankings, with the clearest possible answers to who created it, how it was produced and why it exists.3 Google notes that trust is the most important part of E-E-A-T, and that E-E-A-T itself is not a specific ranking factor.3

Read together, the research and Google’s guidance point in the same direction. Content that states verifiable facts, attributes them, and answers the question clearly is easier for any system, human or machine, to use.

How is GEO different from SEO?

DimensionSEOGEO
Unit of visibilityRanked link on a results pageCitation and share of a synthesized answer
Primary leverRelevance, authority, crawlabilitySame, plus quotable evidence and clear claims
Where it is reportedSearch Console, rank trackersPartly in Search Console and analytics, partly not at all
Typical failureNot indexed, thin contentRetrieved but not cited, or cited inaccurately
Who benefits mostEstablished, well-linked domainsSpecific, evidence-rich pages, including smaller sites
GEO builds on SEO rather than replacing it: a page that is not retrieved cannot be cited.

A practical GEO playbook

Page-level checklist

  • Answer the title’s question in the first two sentences.
  • Use descriptive H2s, several phrased as the questions people ask.
  • Add specific statistics with a citation next to each one.
  • Quote named experts only when the quote is verbatim and sourced.
  • Define terms crisply in one or two sentences before elaborating.
  • Use tables for comparisons; they are easy to extract and hard to misread.
  • Show authorship and how the content was produced (who, how, why).
  • Keep structured data consistent with what is visible on the page.

Then check access. Each AI system documents its own crawlers, and blocking the wrong one can remove you from answers you wanted to be in. Google states that Google-Extended does not affect inclusion in Google Search,8 while OpenAI states that sites opted out of OAI-SearchBot will not be shown in ChatGPT search answers.7

Crawler or tokenWhat it controlsEffect if blocked
GooglebotCrawling for Google Search, including AI featuresNot eligible for Search or AI Overviews
Google-ExtendedA robots.txt product token (not a separate crawler) for training future Gemini models and grounding in Gemini Apps and Vertex AINo effect on Google Search inclusion or ranking
OAI-SearchBotDiscovery for ChatGPT search featuresNot shown in ChatGPT search answers (navigational links may remain)
GPTBotContent that may be used to train OpenAI modelsSignals content should not be used for training
Sources: Google’s common crawlers documentation and OpenAI’s crawler documentation (see Sources).

To limit what Google shows from a page, the controls are the usual ones: nosnippet, data-nosnippet, max-snippet or noindex.2

What can you measure about GEO today?

This is where most GEO advice overreaches. Here is an honest inventory:

SignalWhereWhat it tells youLimits
AI Overview and AI Mode clicksSearch Console Performance, Web search typeClicks and impressions from Google AI featuresIncluded in overall Web data. All links in one AI Overview share one position.
AI Assistant sessionsGA4 default channel groupVisits referred by assistants such as ChatGPT, Gemini or CopilotExcludes Google AI Overviews and AI Mode. Visits without a referrer land elsewhere.
CitationsBing Webmaster Tools AI Performance (preview)Total citations, average cited pages, grounding queries, page-level countsCovers Copilot, Bing AI summaries and some partners. Grounding queries are a sample.
Prompt panelYour own scheduled checksWhether you are cited for the prompts you care aboutA sample, not a census. Answers vary by user, location and time.
Sources: Search Console Help, Google’s AI features documentation, GA4 Help and the Bing Webmaster blog (see Sources).

In detail: Google states that AI feature traffic is included in the Search Console Performance report under the Web search type.2 A click on an external link in an AI Overview counts as a click, and every link in the overview is assigned the same position.4 In GA4, the AI Assistant channel captures arrivals from sources like ChatGPT, Gemini, DeepSeek, Copilot or Grok, and explicitly excludes Google’s AI Overviews and AI Mode, which stay in Organic Search.5 Bing’s AI Performance report, in public preview since February 2026, reports total citations, average cited pages, grounding queries and page-level citation activity across Copilot, AI summaries in Bing and select partners.6

What you cannot measure: how often you appear across all AI answers, how many people saw an answer that cited you inside an assistant with no click, and whether every answer represents you accurately. Anyone selling a precise "AI share of voice" across engines is extrapolating from a sample. Say so in your reporting.

Formula

Prompt-panel citation rate=Prompts where your domain is cited ÷ Prompts checked

where
Prompts checked
A fixed, versioned list of questions buyers ask, run on a schedule per engine.
Cited
Your domain appears as a linked source in the answer, not merely mentioned.

Keep the panel stable (50 to 200 prompts is a practical range for most B2B teams), version it when you change it, and track the rate per engine and per topic. Pair it with an accuracy review: a monthly read of the answers that cite you, flagging anything wrong.

How to track GEO in Kimo

GEO belongs next to SEO, not in a separate tool. In Kimo, connect Search Console and GA4, and the AI Assistant channel appears automatically in the SEO view and the Command center. Upload Bing citation exports or your prompt-panel results as a sheet, and Kimo models them as a citations table keyed by page, so you can see which pages earn citations and signups. The SEO funnel template adds the join to your product database.

A GEO measurement stack: search data, analytics, citation exports and your product database.

For the broader picture of how AI visibility fits into one marketing overview, read One dashboard for every channel and the Search Console and GA4 guide.

Frequently asked questions

Is GEO replacing SEO?

No. Generative engines retrieve sources before they write, so a page that is not crawlable, indexed and relevant cannot be cited. GEO adds a layer on top: making retrieved content easy to quote accurately.

Do I need an llms.txt file or special markup for AI Overviews?

Google says you do not need new machine-readable files, AI text files or special markup to appear in AI Overviews or AI Mode. Other engines may use such files, but there is no Google requirement.

What content changes help most in AI answers?

In the GEO study, adding quotations, statistics and citations to credible sources helped most, while keyword stuffing did not help. Clear answer-first structure and accurate, attributed facts are the practical takeaway.

Can I see AI Overview traffic separately in Search Console?

Google reports AI feature traffic as part of the overall Web search type in the Performance report, and all links within one AI Overview share a single position.

How do I track ChatGPT traffic in GA4?

GA4’s default channel group includes an AI Assistant channel for visits referred by assistants such as ChatGPT, Gemini or Copilot. Visits that arrive without a referrer cannot be attributed this way.

Does blocking Google-Extended remove me from AI Overviews?

Google says Google-Extended does not affect inclusion in Google Search. It controls use for training future Gemini models and grounding in Gemini Apps and Vertex AI.

Sources

8 references
  1. GEO: Generative Engine Optimization (Aggarwal et al., KDD 2024) (opens in a new tab)
    arXiv2024arxiv.org

    GEO-bench (10,000 queries), visibility metrics, method results, rank-dependent effects.

  2. AI features and your website (opens in a new tab)
    Google Search Centraldevelopers.google.com

    No additional requirements; eligibility; no special files; reporting in Web search type; snippet controls.

  3. Creating helpful, reliable, people-first content (opens in a new tab)
    Google Search Centraldevelopers.google.com

    People-first content; Who, How and Why; E-E-A-T and trust.

  4. What are impressions, position, and clicks? (opens in a new tab)
    Search Console Helpsupport.google.com

    Click, impression and position rules for AI Overviews and AI Mode.

  5. [GA4] Default channel group (opens in a new tab)
    Google Analytics Helpsupport.google.com

    AI Assistant channel definition and exclusions.

  6. Introducing AI Performance in Bing Webmaster Tools Public Preview (opens in a new tab)
    Bing Webmaster Blog2026blogs.bing.com

    Citation metrics and surfaces covered.

  7. Overview of OpenAI crawlers (opens in a new tab)
    OpenAIdevelopers.openai.com

    OAI-SearchBot, GPTBot and ChatGPT-User.

  8. Google’s common crawlers (opens in a new tab)
    Google for Developersdevelopers.google.com

    Google-Extended does not affect inclusion in Google Search.

External sources were accessed at the time of writing. Kimo product details, customers and figures in examples are illustrative unless a source is cited.

  • #GEO
  • #SEO
  • #AI search
Found this useful? Pass it on.
Written by
Sam Okafor
Growth lead at Kimo · 3 articles

Writes about GEO, SEO, AI search, Attribution.

Kimo people and customers mentioned are illustrative; example charts use simulated data unless a source is cited.

Put it to work

Go deeper

Whitepaper

One View of Marketing

Unified measurement across social, search, paid and earned media, in the age of AI search.

24 pages
Template

SEO → signup funnel

Impressions to signups, by page and query, joined to your product database.

4 min setup
Live demo

Open the Command center

Social, SEO, paid, email, PR and AI-search visibility in one live overview.

Simulated data · no sign-up

All resources
Definition
Marketing

Generative Engine Optimization (GEO)

Generative Engine Optimization (GEO) is the practice of optimizing content so AI answer engines such as ChatGPT, Perplexity and Google AI Overviews retrieve, cite and accurately represent it in their generated responses.

Kimo team
3 min read
Whitepaper
Marketing

One View of Marketing

Unified measurement across social, search, paid and earned media, in the age of AI search.

Nadia Benali
24 pages

Every channel. One dashboard.

Social, SEO, paid, email, PR and AI-search visibility, normalized into one command center your whole team reads the same way.