GEO

How to Optimise for ChatGPT Search: What We've Learned

Matt Briggs5 min read
ChatGPT interface abstract representing optimisation for AI search

ChatGPT Search has shifted from an interesting experiment to a mainstream research tool used by hundreds of millions of people. Unlike traditional Google search, ChatGPT presents answers — not lists of links — and the pages it cites are the new front page.

This guide covers what we've learned about how ChatGPT selects its sources and what you can do to appear in them.

How ChatGPT Search Works

ChatGPT Search operates differently from a traditional search engine. When a user asks a question with web search enabled, ChatGPT issues multiple sub-queries, retrieves content from the web in real time, and synthesises that content into a single conversational answer. The sources it uses are cited inline and listed beneath the response.

OpenAI operates two distinct crawlers that you need to understand:

GPTBot — Used primarily for training data collection. Visiting your site doesn't mean you'll be cited in search.

OAI-SearchBot — Used for real-time search retrieval. This is the crawler that determines whether your content appears in ChatGPT Search responses.

The distinction matters because some site owners block GPTBot (to prevent their content being used for model training) while keeping OAI-SearchBot allowed for search visibility. Check your robots.txt carefully — many blanket "block AI crawlers" configurations unintentionally block OAI-SearchBot too.

What ChatGPT Looks for in Sources

Based on testing and published research, ChatGPT Search favours sources that demonstrate these characteristics:

Direct relevance to the specific question — ChatGPT parses queries with high semantic precision. Your content needs to address the specific question, not just the general topic. A page titled "SEO Guide" that mentions technical SEO in passing is less likely to be cited for a technical SEO query than a page that directly addresses the specific question.

Factual clarity and specificity — Vague content is de-prioritised. Content with clear, verifiable claims, specific statistics with attributions, and concrete examples performs significantly better in retrieval.

Clear content structure — ChatGPT extracts content at the passage level. Headers that map directly to questions, short declarative paragraphs, and explicit definitions make content easier to extract and cite accurately.

Domain trust — Sites with strong backlink profiles and consistent publishing history are preferred. This correlates with Google domain authority but isn't identical.

Content freshness — For time-sensitive topics, recently updated content is preferred. Add an explicit "Last updated" date and keep dateModified in your Article schema current.

Crawler Access: Check Your robots.txt

The first thing to check is whether OAI-SearchBot can access your site. A common mistake is using catch-all rules that block all unrecognised user agents:

User-agent: *
Disallow: /

Or blocking all bots except Google:

User-agent: GPTBot
Disallow: /

User-agent: *
Allow: /

A well-configured robots.txt for ChatGPT visibility should explicitly allow OAI-SearchBot:

User-agent: GPTBot
Disallow: /

User-agent: OAI-SearchBot
Allow: /

Also create an llms.txt file at the root of your domain. This file (following the llmstxt.org specification) tells AI systems what your site covers and which pages are most important — essentially an AI-optimised sitemap.

Optimisation Tactics for ChatGPT Citations

Answer Questions Directly at the Start of Each Section

ChatGPT retrieves passages, not full pages. For each H2 section in your content, the first 1–3 sentences should directly answer the question implied by that heading. This is the "inverted pyramid" approach — conclusion first, supporting detail second.

Source Every Statistic

ChatGPT users and researchers are quality-conscious. Content with unsourced statistics is treated as lower quality. Format every quantitative claim as: "According to [Source] ([Year]), X% of [finding]."

This serves two purposes: it increases citation probability and makes it more likely that your citation chain extends — ChatGPT may cite your source's source.

Use FAQ Sections with FAQPage Schema

A FAQ section mapping to 4–6 common questions on your topic is one of the highest-impact ChatGPT optimisations. Implement FAQPage JSON-LD schema alongside the visible content. The combination of explicit question-answer pairs and machine-readable schema significantly increases passage-level citation probability.

Build Topical Coverage

ChatGPT appears to weight domain-level topical authority similarly to Perplexity. A site with 15 well-researched articles on a specific subject is more likely to be cited than a site with one article on the same subject, even if both pages are individually high quality.

Develop pillar content and topic clusters around your core areas. The breadth and depth of your coverage signals authority.

Keep Content Updated

For queries about fast-moving topics — AI, digital marketing, technology, regulations — ChatGPT heavily favours freshly updated content. Build a regular update cycle for your most important pages. Add visible "Last updated" dates and keep dateModified in your Article schema current.

Write for Conversational Queries

ChatGPT users ask natural language questions. Use variations of the questions your audience actually asks as H2 and H3 headings. Think about how someone would phrase a query to ChatGPT, not just what keyword they might type into Google.

Measuring ChatGPT Citation Performance

  • Direct query testing — Search for your target queries in ChatGPT with web search enabled. Note whether your pages are cited and how accurately your content is represented.
  • Referral traffic — Track chatgpt.com and chat.openai.com as referral sources in GA4. Traffic volumes are typically low but engagement quality is high.
  • Monitoring platforms — Tools like Otterly.AI, Profound, and BrandMentions now offer automated ChatGPT citation tracking across major AI platforms.

FAQs

Can I block GPTBot but still appear in ChatGPT Search? Yes. GPTBot is for training data collection; OAI-SearchBot is for real-time search. You can block GPTBot in robots.txt while explicitly allowing OAI-SearchBot. Your ChatGPT Search citations will not be affected.

Does Google ranking affect ChatGPT citations? Partially. High domain authority correlates with citation probability, and ChatGPT likely uses some signals that overlap with Google's. But ChatGPT citation is not simply a function of Google rankings — well-structured, specific content on lower-authority sites is regularly cited over weaker pages from higher-authority domains.

What is the difference between ChatGPT Search and ChatGPT without search? ChatGPT without web search draws entirely from its training data (with a knowledge cutoff). ChatGPT Search retrieves and cites current web content in real time. Optimising for ChatGPT Search specifically requires content that is accessible to OAI-SearchBot and structured for AI extraction.

How quickly do optimisations take effect? OAI-SearchBot crawls frequently. Once your content is updated and recrawled — typically within days to a few weeks — changes to structure and freshness signals can affect citation rates relatively quickly.

For a full GEO strategy covering ChatGPT, Perplexity, and Google AI Overviews, get in touch or start with a free audit.

Matt Briggs — Operations Manager at Dynamically

Written by

Matt Briggs

Operations Manager

Matt is Operations Manager at Dynamically, overseeing project delivery, campaign workflows, and client reporting across SEO and PPC.

Back to Insights

Work with Dynamically

Ready to put these insights into practice?

Our Liverpool-based team works with UK businesses to grow organic search, improve paid media performance and build visibility in AI-powered search. Get a free audit to see exactly where your opportunities are.