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

Last month I noticed something odd in my analytics. Traffic from Google was down 15%, but direct traffic and referrals from domains I did not recognise were up. The mystery referrals turned out to be AI answer engines — Perplexity, ChatGPT Browse, and a handful of smaller players — citing my content in their responses. Users were reading my words, on someone else's interface, and some of them clicked through. The game has changed. Traditional SEO is not dead, but a new channel is growing fast, and most developers are not optimising for it.

---

The Shift from Search to Answers

Traditional search works like this: user types a query, Google returns ten blue links, user clicks one. AI search works differently: user asks a question, an AI reads dozens of sources, synthesises an answer, and optionally cites the sources it used. The user may never visit your site at all — or they may visit it because the AI specifically recommended it.

This changes the optimisation calculus. In traditional SEO, you optimise to rank — to be in the top 10 results. In AI search, you optimise to be cited — to be one of the 3-5 sources the AI pulls from when constructing its answer. Ranking and citation are related but not identical. A page can rank #1 on Google and never be cited by AI, and vice versa.

The platforms that matter today:

• Google AI Overviews — the AI-generated summary box that appears above traditional search results
• ChatGPT Browse — when ChatGPT searches the web to answer a question
• Perplexity — a dedicated AI search engine that always cites sources
• Bing Copilot — Microsoft's AI-powered search experience
• Claude — Anthropic's AI, which can browse the web when asked

---

How AI Search Engines Consume Your Content

Understanding how these systems work under the hood helps you optimise for them. The process typically follows four steps:

1. Crawling. AI systems either maintain their own crawl index or query an existing search index (Google, Bing) to find candidate pages for a query.
2. Retrieval. From the candidate pages, the system selects a subset to read in full — typically 5-20 pages.
3. Comprehension. The AI reads the full text of the retrieved pages and extracts the relevant information.
4. Citation. The AI constructs an answer and cites the sources that contributed to it. Not every retrieved page gets cited — only those the AI considered most useful.

The key insight: the AI reads your entire page, not just the meta tags or first paragraph. Unlike Google's crawler, which focuses on signals like title tags and links, AI systems perform deep content analysis. The quality, structure, and clarity of your entire page content matters.

---

What Makes Content Citable by AI

After analysing which of my pages get cited by AI and which do not, clear patterns emerge. AI citation favours content that is:

Specific and concrete. AI systems prefer pages that make definitive statements with specific numbers, dates, or technical details. "UUID version 4 uses 122 random bits" is citable. "UUIDs are a popular way to generate unique identifiers" is not — it is too vague to be useful in an answer.

Well-structured. Content organised with clear headings, bullet points, numbered lists, and comparison tables is easier for AI to parse and excerpt. A well-structured page with h2 and h3 sections acts like a pre-indexed knowledge base that the AI can navigate efficiently.

Authoritative and up-to-date. AI systems consider freshness signals — publication dates, update dates, and whether the content reflects current reality. A guide written in 2022 about a rapidly changing topic is less likely to be cited than one updated in 2026.

Directly answering questions. If your page includes a clear question (in a heading or bold text) followed immediately by a concise answer, AI systems can extract that Q&A pair cleanly. This is the same reason FAQ sections perform well in traditional search — they map directly to how people ask questions.

Free of clutter. Popups, interstitials, excessive ads, and navigation-heavy layouts degrade the signal-to-noise ratio for AI parsers. The cleaner your content, the better.

---

Meta Tags That Matter for AI

AI search engines rely on traditional metadata to understand and categorise your pages before reading the full content. The meta tags you set for traditional SEO pull double duty for AI search.

Meta Tag	AI Relevance	Why It Matters
<title>	High	Used to assess topic relevance before reading the full page
meta description	Medium	Provides a summary that AI may use to decide whether to retrieve the full page
link canonical	High	Tells AI which URL to cite (prevents citation of duplicate URLs)
meta robots	Critical	noindex prevents crawling entirely — by AI and traditional search
article:published_time	Medium	Freshness signal for time-sensitive queries
article:modified_time	Medium	Shows the content is actively maintained
Schema.org JSON-LD	High	Structured data helps AI understand content type and relationships

The canonical URL is particularly important for AI citation. If the AI cites https://example.com/page?utm_source=twitter instead of https://example.com/page, the citation is technically correct but the URL is ugly and tracking-polluted. A clean canonical prevents this. Use the Meta Tag Analyser to verify your canonical tags are properly set.

---

Schema.org Markup for AI Discoverability

Structured data gives AI systems explicit, machine-readable information about your content. While AI can infer that a page is an article by reading it, a JSON-LD block saying "@type": "Article" with "author", "datePublished", and "dateModified" removes all ambiguity.

The Schema.org types that most benefit AI discoverability:

Article

{
  "@context": "https://schema.org",
  "@type": "Article",
  "headline": "SEO for AI Search",
  "author": { "@type": "Person", "name": "Your Name" },
  "datePublished": "2026-04-06",
  "dateModified": "2026-04-06",
  "publisher": { "@type": "Organization", "name": "Your Site" }
}

This tells AI systems the author, date, and publisher — trust signals that influence whether the content is cited.

FAQPage

{
  "@context": "https://schema.org",
  "@type": "FAQPage",
  "mainEntity": [{
    "@type": "Question",
    "name": "How do AI search engines find content?",
    "acceptedAnswer": {
      "@type": "Answer",
      "text": "AI search engines either maintain their own crawl index or query existing search indexes like Google or Bing..."
    }
  }]
}

FAQ schema is particularly powerful for AI because it pre-structures content in a Q&A format that maps directly to how users query AI systems.

HowTo

Step-by-step instructions marked up with HowTo schema are easy for AI to excerpt. If your page explains a process — how to configure cron expressions, how to set up Open Graph tags — HowTo markup makes the steps machine-readable.

---

Google AI Overviews

Google AI Overviews (formerly Search Generative Experience) is the most impactful AI search feature because of Google's market share. When a user searches for an informational query, Google may show an AI-generated summary above the traditional results, citing 2-5 sources.

How sources are selected:

• Google uses its existing search ranking as a starting point — top-ranking pages are more likely to be cited
• Content that directly answers the query in clear, structured prose gets prioritised
• Pages with Schema.org markup, especially FAQ and HowTo, appear more frequently
• Fresh content with recent publication or modification dates is preferred for time-sensitive queries

What this means in practice: if you already rank well for a query, you have a good chance of being cited in the AI Overview. If you do not rank well, the AI Overview is unlikely to discover your content. Traditional SEO remains the foundation — AI optimisation is a layer on top.

The traffic impact is mixed. Some site owners report that AI Overviews reduce click-through rates because users get the answer without clicking. Others report that being cited in an AI Overview drives more clicks than a traditional position #3 ranking because the citation carries an implicit endorsement. The net effect depends on your content type: informational pages may lose traffic, while pages requiring deeper engagement (tools, product comparisons, tutorials) tend to gain.

---

ChatGPT Browse and Perplexity

Both ChatGPT Browse and Perplexity perform live web searches when answering queries, but they work differently from Google AI Overviews.

Perplexity always cites sources and provides numbered references. It tends to cite more sources per answer (5-8) and is more willing to cite smaller, specialised sites. Perplexity's citation format links directly to the source, driving measurable referral traffic. Of all AI search engines, Perplexity is currently the most traffic-friendly for content creators.

ChatGPT Browse searches the web when the user explicitly asks for current information or when the query requires recent data. Citations are inline and linked. ChatGPT tends to cite fewer sources (2-4) and favours well-known, authoritative domains. Getting cited by ChatGPT requires either strong domain authority or highly specific, unique content that other sources do not cover.

Common traits of pages cited by both:

• Clear, factual content with specific data points
• Well-structured with headings that signal topic coverage
• Accessible without login walls or aggressive anti-scraping measures
• Fast to load (both systems have crawl timeouts)

---

Content Structure That AI Loves to Quote

AI systems excerpt content at the paragraph and list-item level. The structure of your content directly affects whether — and how — it gets quoted.

Headers as Topic Signals

Use h2 and h3 headings that describe the content of the section, not clever or vague headings. "Image Specifications Per Platform" is a better heading than "Getting the Picture Right" because AI systems use headings to locate relevant sections for specific queries.

The Answer-First Pattern

For informational content, state the answer in the first sentence of each section, then elaborate. This mirrors the inverted pyramid structure used in journalism:

Good: "UUID version 4 generates identifiers using 122 bits of cryptographically random data, making collisions statistically impossible for practical purposes. Here is how the generation works..."

Less effective: "When considering unique identifiers, there are many factors to evaluate. One common approach is UUID version 4, which we will now explore in detail..."

The first version is instantly extractable. The second requires the AI to read through filler to find the answer.

Comparison Tables

Tables are exceptionally well-cited by AI because they contain structured, comparative data that maps cleanly to comparison queries. When a user asks "What is the difference between X and Y?", a table with a direct comparison is the ideal source to cite. I have seen my JSON vs YAML comparison and UUID format comparisons consistently cited because the table format matches the query intent perfectly.

Lists and Checklists

Numbered lists and bullet points get extracted cleanly by AI. A "5 steps to..." or "common mistakes" list is easy for AI to quote verbatim or paraphrase with attribution. When appropriate, structure your expertise as lists rather than prose paragraphs.

---

E-E-A-T in the Age of AI Search

Google's E-E-A-T framework (Experience, Expertise, Authoritativeness, Trustworthiness) influences which pages get cited in AI Overviews. AI systems inherit the quality signals that Google already calculates.

Experience — content that demonstrates first-hand experience ("I tested this on 50 sites and found...") is more likely to be cited than generic summaries. AI systems can distinguish between original insight and regurgitated information.

Expertise — author credentials, consistent publication history on a topic, and technical depth signal expertise. A developer blog with 20 articles about SEO carries more weight than a page with a single SEO post.

Authoritativeness — backlinks, brand mentions, and domain reputation still matter. AI systems use the same web graph signals to assess source quality.

Trustworthiness — HTTPS, clear authorship, publication dates, and absence of misleading content. Pages that show when they were last updated signal active maintenance.

---

Practical Checklist for AI Search Optimisation

1. Set proper meta tags. Title, description, canonical, robots — the basics still matter. Generate them with the Open Graph Generator and validate with the Meta Tag Analyser.

2. Add Schema.org JSON-LD. At minimum: Article type with datePublished, dateModified, and author. Add FAQPage or HowTo where applicable.

3. Structure content with clear headings. Use descriptive h2/h3 headings. Each section should be self-contained enough to be excerpted independently.

4. Lead with answers. State the key fact or answer in the first sentence of each section. Elaborate after.

5. Include comparison tables. Structured data in table format gets cited more often than the same information in prose.

6. Publish and update dates. Always include article:published_time and update article:modified_time when you make meaningful changes.

7. Keep content accessible. No login walls, no aggressive anti-bot measures, no JavaScript-only rendering. Use server-side rendering or static generation — if social media crawlers cannot read your page, neither can AI crawlers.

8. Be specific. Concrete numbers, benchmarks, code examples, and real-world scenarios get cited. Generic overviews do not.

9. Build topical depth. Publish multiple related articles that link to each other. A cluster of 5-10 articles on a topic signals expertise more strongly than a single comprehensive guide.

10. Monitor AI referral traffic. Check your analytics for referrals from perplexity.ai, chat.openai.com, and similar domains. This is your signal that AI optimisation is working.

---

The Bottom Line

AI search is not replacing traditional search — it is adding a new discovery channel alongside it. The good news is that most AI search optimisation is just good content practices applied more deliberately: clear structure, specific information, proper metadata, and genuine expertise. If you are already creating high-quality, well-structured content with proper meta tags and Open Graph markup, you are already most of the way there. The pages that AI cites are the same pages that humans find useful — they just need to be structured so that a machine can find the useful parts efficiently.