Short answer: what is GEO?
GEO (Generative Engine Optimization) is the work of making content more likely to be selected and cited as a source by AI-powered search experiences such as AI Overviews, AI Mode, ChatGPT search or Copilot. AEO (Answer Engine Optimization) is a close synonym that stresses answering questions directly. According to Google, optimizing for AI search is still SEO: the foundation remains helpful, original content on pages that are indexed and meet the technical requirements.
This guide draws on official documentation from Google, OpenAI and Bing and on the original GEO paper, to separate what has been confirmed from popular tactics without evidence.
SEO vs AEO vs GEO: what is the difference?
The term GEO comes from a paper by Aggarwal et al. posted to arXiv in November 2023 and later presented at KDD 2024. In its guide “Optimizing your website for generative AI features on Google Search” (published 15 May 2026), Google describes AEO and GEO as terms for work focused on visibility in AI search experiences and states that, from Google Search's perspective, this is “still SEO”. The table shows the different emphasis of each label; in practice the boundaries are not strict.
| Criterion | SEO | AEO | GEO |
|---|---|---|---|
| Goal | Rank and earn clicks from search results | Be chosen as the direct answer to a question | Be cited or mentioned in AI-synthesized answers |
| Surface | Lists of links, rich results | Answer passages, voice assistants, AI answers | AI Overviews, AI Mode, ChatGPT search, Copilot, Perplexity |
| Unit of competition | Page | Answer passage | Passages and entities (brand, product, author) |
| Measurement | Rankings, impressions, clicks | Presence in answers | Citations, cited pages, AI referral traffic |
| Shared foundation | Crawling, indexing, helpful content | Same as SEO | Same as SEO |
What Google says about AI Overviews and AI Mode
Google's “AI features and your website” documentation is explicit: there are no additional requirements or special optimizations needed to appear in AI Overviews or AI Mode. To be eligible as a supporting link, a page must be indexed and eligible to be shown in Google Search with a snippet. Both features may use a “query fan-out” technique, issuing multiple related searches across subtopics to build a response.
The main emphasis of Google's AI search optimization guide is non-commodity content: a unique point of view and first-hand experience rather than restating what already exists online. Google also warns that creating separate pages for every query variation or fan-out query, primarily to manipulate results, violates its scaled content abuse policy.
What the 2023 GEO paper actually found
“GEO: Generative Engine Optimization” (Aggarwal, Murahari, Rajpurohit, Kalyan, Narasimhan, Deshpande) introduced the GEO-bench benchmark and tested a range of content edits. The authors report that GEO can boost visibility in generative engine responses by up to 40%. The top methods, Cite Sources, Quotation Addition and Statistics Addition, achieved roughly 30–40% relative improvement on the Position-Adjusted Word Count metric, while keyword stuffing, a traditional SEO tactic, performed poorly. Effectiveness also varied by domain.
Read these results with care:
- “Up to 40%” is a best case under experimental conditions, not an average or a promise for any page.
- Experiments mainly used a generative engine built by the researchers, with an additional test on Perplexity; today's commercial systems may behave differently.
- The metrics measure how much of a source appears in the answer and where, not traffic or revenue.
- “Add statistics” only helps when the numbers are real and sourced. Inventing figures to look citable is a reputational risk and may breach policies.
How to write content AI systems cite
AI-assisted content production should include a human verification step; see an AI content workflow for marketing teams and labeling AI-generated content.
Answer first, explain second
Open each page and section with a two-to-three sentence direct answer, then go into detail. That serves readers and gives AI systems a self-contained passage to quote. Google says there is no need to artificially “chunk” content; structure it with clear headings and paragraphs as you would for people.
Cite sources and use verified figures
Every number, legal reference or important claim should have an openable source, preferably a primary one. This matches both the GEO findings and Google's guidance on helpful, reliable content.
Clear entities and structured data
Write your company, product and author names consistently across your site and external profiles. Structured data such as Organization, Article and BreadcrumbList is still useful for rich results, but Google states there is no special schema for AI search and structured data is not required. Mark up only what is visible on the page.
Keep content fresh
Show a genuine updated date and review content when laws, products or figures change. Bing recommends IndexNow to notify search engines when content is added, updated or removed, so AI answers reference the current version.
Technical: robots.txt for AI crawlers and llms.txt
Providers split their crawlers by purpose, so you can decide separately for each. Blocking a search crawler by mistake means your pages cannot be cited.
On llms.txt: it is a proposal published by Jeremy Howard on 3 September 2024 (v2 updated 10 August 2026) suggesting a Markdown file at the site root that summarizes and links to important content for language models. It is a community proposal, not an official standard. Google states that Google Search does not use llms.txt and that creating one neither helps nor harms visibility in Google Search. Create it if other systems you care about use it, but do not treat it as a main lever.
| Token | Provider | Purpose | Effect of blocking |
|---|---|---|---|
| Googlebot | Crawls for Google Search, including AI Overviews and AI Mode | Lose visibility in Google Search | |
| Google-Extended | Control token for use in Gemini apps and the Vertex AI API for Gemini | Not a separate crawler; per Google, no effect on Google Search | |
| OAI-SearchBot | OpenAI | Surfaces sites in ChatGPT search results | Pages unlikely to be cited in ChatGPT search |
| GPTBot | OpenAI | Crawls content that may be used to train models | Excluded from future training data; independent of OAI-SearchBot |
| Bingbot | Microsoft | Indexes for Bing, the basis for Copilot answers | Lose visibility in Bing and Copilot |
FAQ schema after 2026: still worth it?
In August 2023 Google announced that FAQ rich results would only be shown for well-known, authoritative government and health websites. According to a note on Google's FAQPage documentation (reported by Search Engine Land and other trade outlets), as of 7 May 2026 FAQ rich results no longer appear in Google Search; the Search Console report and Rich Results Test support were scheduled for removal in June 2026 and API support in August 2026.
The practical conclusion: visible Q&A sections are still valuable because they answer real user questions, but do not add FAQPage markup expecting a dropdown on Google. Existing markup does not need urgent removal; Google has said unused structured data does not cause problems.
Measuring visibility in AI search
Google notes that no third-party tool has access to its internal ranking or AI systems; be cautious of services that “promise” AI Overview placement.
- Google Search Console: traffic from AI Overviews and AI Mode is included in the Performance report (Web search type); Search Console does not currently offer a separate AI-citation report like Bing's.
- Bing Webmaster Tools: the AI Performance report (public preview since February 2026) shows citations, cited pages and grounding queries across Copilot and partners. Data is sampled and covers only Microsoft's ecosystem.
- ChatGPT: per OpenAI, referrals from ChatGPT search include utm_source=chatgpt.com, which you can track in analytics.
- Regular manual checks: a fixed set of questions about your brand and core topics, recording which sources are cited.
- Tie it to conversions: enquiries, sign-ups, consultation requests, not just clicks.
What not to do
The practices below either break search policies or have been shown to work poorly:
- Create hundreds of near-identical pages for each question variant or fan-out query.
- Buy or stage fake brand “mentions” on forums and social media.
- Invent statistics, quotes or clients to make content look citable.
- Stuff keywords; the GEO paper found this performs poorly.
- Accidentally block Googlebot, Bingbot or OAI-SearchBot through robots.txt or a firewall.
- Rewrite everything in an “AI voice”; Google says no special writing style is needed for AI search.
Frequently asked questions
What is GEO?
GEO (Generative Engine Optimization) is optimizing content so AI-powered search experiences select and cite it as a source in their answers. The term comes from a 2023 paper by Aggarwal et al. The goal is to be chosen as a source by AI Overviews, AI Mode, ChatGPT search or Copilot, not only to rank in a list of links.
Are AEO and GEO different from SEO?
They differ in surfaces and measurement, but share the same foundation. Google states that, from Google Search's perspective, optimizing for AI search is still SEO.
Do I need an llms.txt file?
No. llms.txt is a proposal, not a standard, and Google Search does not use it. Create one only if systems you care about use it.
How do I get my site into AI Overviews?
Per Google, a page must be indexed and eligible to show with a snippet; there are no additional technical requirements. Helpful, original, well-structured content matters most.
Does blocking GPTBot remove me from ChatGPT?
Per OpenAI, GPTBot (training) and OAI-SearchBot (search) are controlled independently. You can block GPTBot while allowing OAI-SearchBot.