AI SEO: making your site readable, and citable, by Google and assistants
People, Google and assistants read your site. None of them runs your JavaScript as a favour.
The work is making your site legible to all three without writing anything twice.
What changes when the reader is a model
A model does not browse your navigation. It lands on one page, reads it whole, and decides whether it can be cited. That rewards what was already good SEO — content in the HTML, headings that describe, data that matches what is on screen — and punishes what was already noise.
What I work on
Structured data that matches the page
Markup declaring questions, prices or reviews that are not on screen is the pattern search engines treat as hidden content. Every block has to say what the page shows.
An entity graph without dangling identifiers
If your name is shared or your brand resembles another, the job is leaving no doubt about which one you are — and making sure every identifier points at something that exists.
Passages that stand on their own
A claim and its source in the same place, so a model can use the passage without inventing the rest of the context.
How you would know it is working
Access, exposure and citation are three different things, measured with three different instruments.
Access
Your own server logs: which AI crawlers arrive, what they request, and what they receive. First-party data, not a third party's estimate. If you are on Cloudflare, AI Crawl Control gives you this directly.
Exposure
Search Console's generative AI features report: how often a link to your site was shown inside a Google AI response.
Citation
Bing Webmaster Tools, AI Performance: which pages Copilot cites and which queries retrieve them.
Those are the only two first-party sources that exist today. ChatGPT, Claude and Perplexity report nothing, and that gap is only covered by sampling prompts in a clean session. Anyone telling you otherwise is selling you something.
When this is not for you
If your site does not rank in ordinary search yet, this is not your problem to solve: assistants lean on search indexes, so the failure is upstream. If you sell through direct referral rather than discovery, same. And if your content says nothing that someone else does not say equally well, no amount of structured data will make a model prefer your version.
What I do not promise
I cannot make ChatGPT cite you, and neither can anyone selling you that. What I can do is make your site legible, coherent and verifiable, which is the part that depends on us.
If you would rather do it yourself, the how-to is in how to rank in AI Overviews.
Can Google and the assistants actually read your site?
I check it in the free diagnosis: structured data, entity, and what a model can read from your pages.
Get a free diagnosis