GEO (generative engine optimisation) is the work of making sure your content appears accurately, and is cited as a source, in AI-powered search such as ChatGPT, Perplexity, Claude and Google AI Overviews. It is not separate from SEO: these systems search the web before writing an answer and link to the pages they used. A site that is crawlable, fast, clearly written and consistent about its facts is the foundation for both.
How is GEO different from SEO?
The term comes from the paper GEO: Generative Engine Optimization, published in November 2023 and accepted at KDD 2024. It calls systems that compile an answer from several sources “generative engines” and groups the techniques that make content more visible in those answers under the name GEO.
Google’s guide to generative AI features in Search mentions GEO and AEO (answer engine optimisation), but says that from Google Search’s point of view, optimising for generative AI search is still SEO. AI Overviews and AI Mode ground their answers in pages retrieved from Google’s index (RAG) and split a question into related sub-queries to gather more results (query fan-out). Microsoft presents the AI Performance report in Bing Webmaster Tools as an early step towards GEO tooling.
Our working distinction: SEO aims for rankings and clicks; GEO adds being described accurately inside the answer and earning the source link. The technical foundation is the same.
How do AI assistants find their sources?
The flow is similar across assistants: turn the question into search queries, search the web or an index, read a few pages, write the answer and cite the pages used. A page that cannot be crawled or is not indexed cannot be chosen.
| Assistant | Where it finds sources | Crawler |
|---|---|---|
| ChatGPT search | OpenAI’s own crawling and, for some searches, partner search providers (OpenAI) | OAI-SearchBot |
| Perplexity | The web, in real time; every answer links its sources (Perplexity) | PerplexityBot |
| Claude | A web search tool, with citations (Anthropic) | Claude-SearchBot |
| Google AI Overviews and AI Mode | Google’s search index (Google) | Googlebot |
Google has no extra requirement: a page must be indexed and eligible to show in Search with a snippet, and no special markup or file is needed. For ChatGPT search, OpenAI asks you to allow OAI-SearchBot and make sure your host or CDN does not block its published IP addresses, and notes that placement is not assured.
Practical steps for your site
A crawlable, fast site
Google’s page on AI features lists the basics: crawling allowed in robots.txt and by your CDN or host, content reachable through internal links, a good page experience and important content available as text. Firewall or CDN bot protection can silently block crawlers you allowed in robots.txt, so check it.
On antargetech.com, rebuilt in October 2026, the home page scores 97 for Performance on mobile Lighthouse and 100 for Accessibility, Best Practices and SEO. The biggest gains came from preloading the above-the-fold image, shrinking images and removing unused CSS (from 727 kB to 134 kB).
Structured data
Structured data describes a page in a machine-readable way, and Google recommends JSON-LD. But Google says it is not required for generative AI search and there is no special schema.org markup to add; the markup simply has to match the visible text. Google has also stopped showing the FAQ rich result since 7 May 2026.
Every page on our site carries Organization, WebSite, WebPage and BreadcrumbList; service pages add Service and FAQPage, product pages SoftwareApplication and blog posts BlogPosting.
Consistent company information
An AI answer about a company combines several sources, so a name or address that differs between them can produce a contradictory summary. Google’s Organization structured data lets you define your name, address, logo and, through sameAs, your profiles elsewhere. For local businesses, Google says a Business Profile can help visibility in AI responses too.
For us, that means the same registered name (Ant ArGe Teknoloji Bilişim Sanayi ve Ticaret Ltd. Şti.), short name (AntArge) and address (Antalya Technokent R&D 2 Building, Konyaalti, Antalya) everywhere. Google also notes that chasing inauthentic “mentions” helps less than it might seem.
Answer first, then explain
Google recommends clear paragraphs and headings, and says you do not need to “chunk” text or write in a special way for AI. Every post on this blog answers its title’s question in the first two or three sentences, so neither readers nor systems quoting the page have to hunt for it. FAQ sections are a good place to answer what customers actually ask; our service pages have them.
Cite sources and add original information
In the GEO paper’s experiments, adding citations, quotations and statistics gave a relative visibility gain of 30–40% on the paper’s own metric, while keyword stuffing brought little or no improvement. These are lab results, but the direction is clear. Google likewise says original, first-hand content is likely to matter more in the long run than any of its other suggestions.
robots.txt and llms.txt
AI crawlers
Vendors use separate user agents for search and for training, and you can manage each in robots.txt. OpenAI states that you can, for example, allow OAI-SearchBot while blocking GPTBot.
| Vendor | User agent | Purpose |
|---|---|---|
| OpenAI | OAI-SearchBot | ChatGPT search; opted-out sites are not shown in its answers |
| OpenAI | GPTBot | Model training |
| OpenAI | ChatGPT-User | User-triggered visits; robots.txt may not apply |
| Anthropic | Claude-SearchBot | Improving Claude’s search results |
| Anthropic | ClaudeBot | Model training |
| Anthropic | Claude-User | Visits to answer a user’s question |
| Perplexity | PerplexityBot | Perplexity search; not used for training |
| Perplexity | Perplexity-User | User-requested visits; generally ignores robots.txt |
| Google-Extended | Gemini training and grounding in Gemini Apps and Vertex AI |
OpenAI and Perplexity say robots.txt changes can take about 24 hours to apply. Our robots.txt is open to both search and training crawlers; a simplified version:
# AI search crawlers
User-agent: OAI-SearchBot
User-agent: Claude-SearchBot
User-agent: PerplexityBot
Allow: /
# Training crawlers (change to "Disallow: /" to opt out)
User-agent: GPTBot
User-agent: ClaudeBot
User-agent: Google-Extended
Allow: /Allowing training crawlers is a business decision. If you want to appear in AI search, at least keep the search crawlers open.
What Google-Extended does not control
Google-Extended is not a separate crawler but a robots.txt token. Google’s documentation says it controls Gemini model training and grounding in Gemini Apps and Vertex AI; it does not affect inclusion in Google Search and is not a ranking signal. Appearance in AI Overviews and AI Mode depends on Googlebot access, controls such as nosnippet, data-nosnippet, max-snippet and noindex, and the Search generative AI setting in Search Console. Blocking Google-Extended will not remove you from AI Overviews.
llms.txt
llms.txt is a convention proposed by Jeremy Howard in September 2024: a Markdown file at the site root that gives AI tools a short summary and links to key pages. Only an H1 with the site name is required. It is an open proposal, not an official standard.
Do not assume search engines use it. Google says Google Search does not use llms.txt and that the file neither helps nor harms visibility, though keeping one for other services is fine. OpenAI and Perplexity publish llms.txt on their own documentation sites, but their crawler documentation says nothing about reading other sites’ files. We publish /llms.txt and /llms-full.txt as a tidy summary for tools, not as a ranking lever.
How do you measure it?
- Google Search Console: AI Overviews and AI Mode traffic is counted in the Performance report under “Web”. The generative AI performance report shows how often your links appeared in these features by page, country and device; Google says it reached all sites as of 31 August 2026.
- Bing Webmaster Tools: AI Performance shows how often your pages are cited in Microsoft Copilot and Bing’s AI summaries. Bing notes the numbers reflect citation frequency, not importance or ranking.
- Google Analytics 4: the default AI Assistant channel groups visits from sources such as ChatGPT, Gemini, DeepSeek, Copilot and Grok; AI Overviews and AI Mode stay in Organic Search. For other assistants, check session source for domains like perplexity.ai and claude.ai.
- Manual checks: regularly ask assistants the 10–20 questions your customers ask and record whether you are cited and described correctly.
Frequently asked questions
Is GEO replacing SEO?
No. Google says optimising for generative AI search is still SEO, built on the same ranking and quality systems. GEO adds a few habits in measurement and writing.
If I block training bots, will I disappear from AI search?
According to the vendors’ documentation, no: search and training crawlers are managed separately, PerplexityBot is not used for training, and Google-Extended does not affect Google Search.
Do I need an llms.txt file?
No. It is a proposal and Google Search does not use it. It is easy to add but does not replace crawlable pages and original content.
Conclusion
GEO is not a list of tricks; it is the work that lets search engines and AI assistants find, understand and confidently cite your site. Start with crawlability and speed, then accurate structured data and consistent company information, then original content that answers directly and cites its sources. Moving to a new platform? See our post on avoiding SEO loss during a website migration.
If you would like to review your site’s visibility in AI search together, start with our SEO and digital visibility page. To put your own content to work in AI applications, see our AI solutions.
