"Do I need a new strategy for AI search?" is the question I get most from clients this year. Google AI Overviews and AI Mode, ChatGPT search, Perplexity and Claude now answer many questions directly — sometimes without a click.
The honest answer: mostly no — but the fundamentals matter more than ever. Google's own AI search guidance (published in May 2026) frames optimising for generative search as simply good SEO. What changed is how unforgiving AI systems are of slow, thin or hard-to-parse sites.
Here's the checklist I use, and that I applied to this very website.
1. Let the right crawlers in
You can't be cited if you're blocked. Check robots.txt and your CDN/firewall rules. Many AI companies run separate bots for training, indexing and live fetching:
User-Agent: *
Allow: /
Disallow: /dashboard
User-Agent: GPTBot
User-Agent: OAI-SearchBot
User-Agent: ClaudeBot
User-Agent: PerplexityBot
Allow: /
In Next.js, generate it from app/robots.js so it stays in version control.
2. Server-render everything you want cited
AI crawlers are often less patient with JavaScript than Googlebot. If your content only appears after client-side fetching, it may never be seen. On this site, project pages and blog posts are statically generated — the full text is in the initial HTML.
Quick test: curl https://yoursite.com/page | grep "a sentence from your content".
3. Structure content so it can be quoted
AI answers are built from passages. Make yours easy to extract:
- One clear
<h1>, logical<h2>/<h3>hierarchy. - Short paragraphs that make one point each.
- Answer the question in the first sentence of a section, then elaborate.
- Use lists and tables for comparisons.
- Add an FAQ section for questions people actually ask.
4. Structured data (JSON-LD)
Schema.org markup removes ambiguity about what a page is and who it's about. For a personal or business site, the essentials:
PersonorOrganization(withsameAslinks to your LinkedIn, GitHub, etc.)WebSiteArticle/BlogPostingfor posts, with author and datesBreadcrumbListFAQPagefor FAQ sections
<script
type="application/ld+json"
dangerouslySetInnerHTML={{ __html: JSON.stringify(articleJsonLd) }}
/>
5. Be an entity, not just a website
AI systems try to understand entities — people, companies, products — and how they connect. Help them:
- Use the same name and headline everywhere (site, LinkedIn, GitHub, X).
- Link your profiles with
rel="me"andsameAs. - Get mentioned on other credible sites: client websites ("Built by…"), community posts, interviews.
Earned mentions are the closest thing AI search has to backlinks.
6. llms.txt: nice to have, not magic
/llms.txt is a proposed Markdown file summarising your site for language models. It isn't an official standard, and Google has not said it uses it. I still publish one on this site because it's cheap and gives assistants a clean, accurate summary. Just don't expect it to move rankings on its own.
7. Freshness and accuracy
AI answers favour content that's current and specific. Put dates on posts, update them when facts change, and avoid vague filler. A specific paragraph backed by real numbers from your own work is far more citable than generic advice.
8. Core Web Vitals still matter
Speed affects rankings, user experience and crawl efficiency. Aim for good LCP, CLS and especially INP. Compress images, ship less JavaScript, and render on the server.
9. Measure the new traffic
- Google Search Console shows AI Overview impressions within normal search data.
- Watch referrals from
chatgpt.com,perplexity.aiand similar in your analytics. - Search your name and your key topics in several assistants every month and note who gets cited.
The checklist
- Crawlers allowed in
robots.txtand CDN rules - Content server-rendered
- Clean heading structure and quotable passages
- JSON-LD for Person/Organization, Article, Breadcrumbs, FAQ
- Consistent identity across the web +
sameAs - Optional
llms.txt - Dated, specific, updated content
- Good Core Web Vitals
- Search Console + referral tracking
AI search didn't replace SEO. It raised the bar for doing the basics properly.
