AI and search: what content designers need to know
Google has published new guidance on optimising websites for generative AI features in search.
The fundamentals of content design still apply. Write clearly, structure content well, and keep the technical foundations solid. Most of the AI-specific 'hacks' circulating online don't work.
What to do
Write for humans
Focus on clear headings, short paragraphs, plain language, and structure that's easy to follow. Google's AI systems use the same search index and ranking systems as regular results. Content that works well for readers works well for AI.
If someone can scan your page and find what they need, Google's AI can too.
Focus on needs
Write content to meet user needs. Google calls this 'non-commodity' content. It's more likely to rank well and appear in AI results than generic pages.
Read Google's guide to helpful, people-first content for more detail.
Use clear headings
Front-load your headings. Put the most important word first. Each heading should clearly describe the content beneath it. Avoid vague headings like 'Things to consider'.
Google's systems use headings to work out what each section is about. A clear structure also makes it easier for Google to pull the relevant part of a long page and surface it as an answer in AI Overviews.
Check your metadata
Check your page titles, meta descriptions, alt text, and structured data. Google's systems use these to understand what your content is about. Titles and meta descriptions can show directly in search results. Alt text and structured data work behind the scenes. If any of them are inaccurate, Google may misunderstand your content or present it incorrectly.
Keep related content on one page
Cover a topic thoroughly on one page, answering the questions people are likely to have about it. Google's systems can find and surface the relevant section. Creating separate pages for every possible question can violate their scaled content abuse policy.
Don't keyword-stuff
Write naturally. Google's AI understands synonyms and meaning, so it can connect the words people use in search with your content even when the exact words don't match. Answering the question well is more important than keyword-stuffing.
Use semantic HTML
Use semantic HTML to make your pages easier to parse for screen readers, browser agents, and search systems. It doesn't have to be perfect. Google can read messy HTML, but clean structure helps everyone. Google's AI models use publicly accessible, crawlable content, so make sure yours is.
You do not need special AI markup, LLMS.txt files, or markdown versions of your pages.
Using AI to create content
Google doesn't ban AI-generated content. But it must meet the standards of the Search Essentials and their spam policies.
Mass-produced content, whether by AI or humans, that exists primarily to manipulate rankings violates the policy on scaled content abuse.
Read Google's guidance on AI-generated content.
Common misunderstandings
Use these tables to help answer common misunderstandings about AI and search.
Content and rankings
Subject expert concern | What Google says |
|---|---|
'We need to increase traffic. Can we use AI to produce more content?' | More content doesn't mean more traffic. Mass-producing pages to chase rankings is what Google's scaled content abuse policy targets, whether AI-generated or human-written. Fewer pieces with real expertise will perform better. |
'We're not appearing in AI results. Should we be writing differently?' | AI results use the same search index and ranking systems as regular results. There's no separate style to learn. If the content is well structured, clearly written, and answers the question, it's already optimised for AI. |
'We're not ranking for this keyword. Can we add it more times?' | Repeating a keyword won't help. Google's AI understands synonyms and context, so it can match searchers with your content even when the exact words differ. Focus on answering the question well. |
'We need more online visibility. Let's get mentioned everywhere.' | Visibility matters, but not all mentions are equal. Google calls planted references 'inauthentic mentions' and says chasing them does less than people expect. Genuine coverage from real sources carries more weight. |
AI-specific tactics
Subject expert concern | What Google says |
|---|---|
'How do we get into AI Overviews?' | AI Overviews pull from Google's existing search index using the same ranking systems as regular search. There's no separate tactic. Good content and good SEO are the route in. |
'This page is too long. Should we split it up for AI?' | Length isn't the problem. Google's systems can extract the relevant section from a single page. One well-structured page is better than ten short pages. |
'People search for lots of different questions. Don't we need a page for each one?' | Google's AI matches intent, not exact words. A thorough page that covers the topic will reach users without you needing a separate page for every variation. |
'I've read we need an LLMS.txt file.' | Google has said explicitly you don't need special AI text files, markup, or markdown. Standard HTML is enough. |
'Do we need special schema markup for AI?' | Structured data isn't required for AI features. It's still useful for rich results in regular search. There is no special schema for AI. |