How to Optimise for AI Search: AI Overviews, ChatGPT and Perplexity

Search is changing shape. Google shows AI Overviews at the top of many results and offers a conversational AI Mode. ChatGPT, Perplexity, Claude, Microsoft Copilot and others answer questions directly and cite web pages as sources. For many queries, people now read an AI-generated summary before – or instead of – clicking a blue link.

That raises a new question for site owners: how do you become one of the sources these systems quote? The practice is sometimes called generative engine optimisation (GEO) or answer engine optimisation. The good news is that it builds directly on solid SEO.

How AI search engines find sources

Most AI search tools work in a similar way:

1. They interpret the question, often breaking it into several sub-queries
2. They retrieve relevant pages – from a search index, a live web search or their own crawl
3. They read those pages and generate an answer
4. They cite the pages that supported the answer

So to be cited, a page must first be retrievable (crawlable and indexed), then readable, and finally useful enough to support an answer.

Google AI Overviews and AI Mode

Google’s AI features draw on its regular search index and ranking systems. Google has said there are no special requirements beyond normal SEO best practices: a page must be indexed and eligible to be shown with a snippet. Pages that rank well and directly answer questions are the most likely to appear.

ChatGPT, Perplexity, Claude and others

These tools use a mix of search partnerships and their own crawlers, such as OAI-SearchBot (OpenAI), PerplexityBot and Claude-SearchBot (Anthropic). If your robots.txt blocks these bots, your pages may not be used as sources in their search answers.

Step 1: Make sure AI crawlers can access your content

Check your robots.txt. Many sites added blanket blocks for AI bots in 2023–2024 without distinguishing between purposes:

• Training crawlers (GPTBot, ClaudeBot, Google-Extended, CCBot) collect content for model training
• Search crawlers (OAI-SearchBot, Claude-SearchBot, PerplexityBot) fetch pages to answer and cite in search
• User-triggered fetchers (ChatGPT-User, Claude-User, Perplexity-User) read a page when a person asks about it

You can block training while allowing search. Blocking Google-Extended does not affect Google Search or AI Overviews.

Step 2: Put your content in the HTML

Most AI crawlers fetch the raw HTML and do not run JavaScript. If your text is rendered in the browser by a JavaScript framework, those crawlers may see an empty page. Use server-side rendering or static generation for important content.

Step 3: Allow snippets

The nosnippet and max-snippet:0 directives stop Google from showing text from your page – and Google has said they also keep content out of AI Overviews. Only use them on content you genuinely do not want quoted. The data-nosnippet attribute lets you exclude specific sections instead of a whole page.

Step 4: Write answer-first content

AI systems look for passages that clearly answer a question. Make those passages easy to find:

• Use question-style headings that mirror how people ask (“How long does a loft conversion take?”)
• Answer in the first sentence below the heading, then add detail
• Keep paragraphs focused on one idea each
• Use lists for steps and tables for comparisons – they are easy to extract
• Define terms clearly the first time you use them

This is the same structure that wins featured snippets and helps human readers scan.

Step 5: Be specific and original

Generative answers summarise what is commonly said. To be cited rather than blended in, offer something others do not:

• Original data, surveys, test results or pricing information
• First-hand experience, photos and examples
• Clear expert opinions and recommendations
• Precise figures, dates and definitions with sources

Pages that simply repeat what everyone else says give an AI no reason to cite them specifically.

Step 6: Build entity clarity and trust

AI systems need to understand who you are and whether you are credible:

• Use consistent names, descriptions and details for your brand across your site, social profiles and directories
• Add Organization, Person and Article structured data with authors and dates
• Show author bios with relevant expertise
• Earn mentions and links from reputable sites in your field – the same signals that build authority for search

Step 7: Keep content fresh

For questions where facts change – prices, regulations, product line-ups, statistics – AI search tends to favour recently updated sources. Update key pages and show the last updated date.

What about llms.txt?

llms.txt is a proposed convention (introduced in 2024) for a Markdown file at /llms.txt that summarises a site and links to its most important pages for language models. Some AI tools and developer assistants read it. It is not an official standard, and Google has said it does not use it for Search. It is cheap to add, so documentation sites and SaaS products in particular may find it worthwhile – but it will not substitute for crawlable, high-quality content.

Measuring AI search visibility

Measurement is still developing:

• Referral traffic from chatgpt.com, perplexity.ai, copilot.microsoft.com and similar shows up in analytics
• Google AI features traffic is included in Search Console’s overall Performance data rather than reported separately
• Manual checks: regularly ask the main AI tools the questions your customers ask and note whether and how you are cited
• Brand searches: rising branded search can reflect visibility in AI answers

A quick AI search checklist

• AI search crawlers allowed in robots.txt
• Main content present in raw HTML
• No unintended nosnippet directives
• Question headings with direct answers
• Lists and tables for steps and comparisons
• Original information and first-hand experience
• Organization, Article and author structured data
• Visible author and last-updated date
• Key pages kept current

Content formats that AI search tends to cite

Some formats are especially easy for AI systems to use and attribute:

• Definitions and explanations – a clear one- or two-sentence definition at the start of a section
• Step-by-step instructions – numbered steps with specific details
• Comparisons – tables comparing options on consistent criteria
• Statistics and data – original numbers with clear context and dates
• Pros and cons – balanced lists for products, methods or decisions
• FAQs – genuine questions with direct answers

Notice that these are also the formats readers find most useful. Writing for AI search is, to a large extent, writing clearly for people.

Before and after: making a passage quotable

Before:

When it comes to the question of how often you should be descaling your espresso machine, there are a lot of different factors that come into play, and many people have different opinions on this topic, but generally speaking it depends.

After:

Descale an espresso machine every one to three months. Descale monthly if you have hard water or make more than three drinks a day, and every three months with soft or filtered water. Most machines also show a descaling alert.

The second version answers in the first sentence, gives specific numbers and covers the conditions – exactly what an AI answer (or a busy reader) needs.

Local businesses and AI search

AI assistants increasingly answer local questions (“best Italian restaurant near Leeds station that takes bookings”). They draw on business listings, reviews and websites. To be included:

• Keep your Google Business Profile, Apple Business Connect and Bing Places listings complete and consistent
• Collect genuine reviews that mention specifics
• Make opening hours, prices, location and booking details easy to find on your site
• Add LocalBusiness structured data that matches your listings

What not to do

• Do not hide text or create AI-only content. Showing different content to crawlers than to users (cloaking) violates search engine guidelines.
• Do not mass-produce thin pages targeting every possible question. Scaled low-value content is treated as spam, whether written by people or AI.
• Do not chase prompts instead of people. The questions your customers ask are the best guide to the content you need.

Key takeaways

• AI search engines retrieve, read and cite web pages, so SEO fundamentals still come first.
• Allow AI search crawlers in robots.txt; blocking training crawlers does not affect search visibility.
• Put content in the raw HTML – most AI crawlers do not run JavaScript.
• Write answer-first content with question headings, lists and tables.
• Original data, first-hand experience and clear entity information make pages worth citing.
• llms.txt is optional and emerging; it does not replace crawlable, high-quality content.

Frequently asked questions

Is GEO different from SEO?

It is an extension of SEO rather than a replacement. Crawlability, indexing, helpful content and authority remain the foundation; GEO adds emphasis on being quotable and accessible to AI crawlers.

Do AI Overviews reduce clicks?

For some simple informational queries, users get their answer without clicking. For more complex or commercial queries, cited sources can still earn valuable visits. Focusing on content that goes beyond a quick answer helps.

Should I block AI crawlers?

It depends on your goals. Blocking training crawlers protects content from model training without hurting search. Blocking search crawlers removes you from those AI answers.

Can I pay to appear in AI answers?

Organic citations in AI answers cannot be bought. Some platforms are experimenting with ads alongside AI answers, which are separate and labelled.