Optimising for AI Overviews and ChatGPT search — what's actually changed.

The shift that's actually happening

Google AI Overviews now appear on approximately 15–20% of all US searches, with significantly higher coverage on informational and how-to queries. ChatGPT search is handling hundreds of millions of queries per month. Perplexity, Claude, Gemini and a growing list of AI-native search products are collectively handling a meaningful share of research and comparison queries.

Do AI Overviews cannibalise clicks? How do you get cited in AI-generated answers? Is this a threat to existing SEO or an extension of it? The honest answers: yes sometimes; through entity authority and structured expertise signals; and both, depending on what kind of SEO you've been doing.

What AI Overviews actually show — and who gets cited

Sources frequently cited in AI Overviews have three things in common:

• Strong topical authority — they rank well for many queries within a topic cluster, not just one keyword.
• Explicit expertise signals — content attributed to a human author with verifiable credentials or to an organisation with demonstrated domain expertise.
• Structured, citable content — clear questions, direct answers, numbered lists and concise definitions are more extractable than long-form prose with no hierarchy.

ChatGPT and third-party model citation — different signals

LLM-based search products operate differently from Google. Getting cited requires a different set of signals: presence in training data (a lagging signal — what you publish today affects citation in 12–18 months), brand entity recognition (built from frequency in high-quality content across multiple contexts), third-party citation patterns (mentions in 10 industry publications beat being cited only on your own site), and structured data (Organization, Person, FAQ schema make content more parseable and citable).

The practical implications for your content strategy

Prioritise topical clusters over individual keywords

AI models recognise topical authority at the subject level. A site that covers HVAC maintenance comprehensively across sub-topics and intent stages beats a site that ranks for one high-volume keyword.

Write for explicit expertise

Add bylines with verifiable credentials. Include author bios. Cite primary sources and original data. Content that reads as genuine expert work beats aggregated generics.

Structure content for extraction

H2/H3 subheadings that function as direct answers to specific questions. Numbered lists for processes. Concise summary or definition early in articles. Optimised for AI extraction and featured snippets simultaneously.

Build brand entity presence

Earned media, guest articles, inclusion in industry directories, third-party mentions that establish your brand as a real, verifiable entity. The more contexts your brand appears in as a genuine authority, the more strongly it registers in model training data.

Implement comprehensive schema

FAQPage schema with genuine Q&A pairs. Article schema with proper author attribution and datePublished. Organization schema with complete entity data. Helps Google indexing and AI parsing.

What's not changing

Most of the fundamentals haven't changed. Businesses that rank in 2026 still have technically clean sites, genuine topical expertise, earned authority signalled through quality backlinks, and user signals indicating the content is genuinely useful. The businesses most exposed to AI search disruption are the ones whose traffic came from thin informational content that was never genuinely useful — content that ranked because it was keyword-optimised, not because it was the best answer. The businesses best positioned are the ones who were doing genuine expertise-first content before anyone was talking about AI Overviews.