Optimising web content for AI Snippets, and AI Chat queries (the new SEO)
My first reaction to AI snippets and AI Chat queries was indignant: If no one’s clicking, there’s no engagement data. So how does the algorithm rank which content is the most valued by users? Is it just relying on old data? Will new content ever get surfaced? I wondered if the internet might get stuck at a point in time: AI recycling old information, fresh content being pointless.
AI Snippets aren’t like the ‘structured snippets’ that disrupted SEO a few years ago. Back then, you’d get an excerpt from a web page, with a link you could click to see the full context. There was only one page that could get that top spot, so we were all gunning for the most traffic, going for the niche long tail keywords, and tracking it like a race. Now, AI snippets are plating up summarised content from multiple pages - often with shaky accuracy. There’s no need for a user to click through to the original source, unless they know enough about the topic to spot if something in the snippet looks obviously wrong. Not to mention that as content creators, we have no visibility into the traffic or interaction we get from it either.
As an industry, we’ve now had more time to look at how AI chat and snippet tools actually work, and how they choose and rank the content they use. There’s still a lot to figure out, but some useful patterns are emerging.
1. Indexing on Bing matters
If your site isn’t indexed on Bing, it won’t appear in ChatGPT or other AI tools that rely on Bing’s infrastructure.
2. Understand how AI fetches content. You’re not just optimising for keywords anymore, you’re also optimising for how the AI interprets the query and selects its sources.
AI doesn’t search the web in the traditional sense. When someone enters a query:
The model first interprets the query using natural language processing. It defines its meaning, and identifies the users intent to figure out what sort of content it should be looking for.
Intent Mapping: The identified intent is mapped to a specific action or task that the chatbot should perform.
Then it uses a tool like WebGPT to fetch structured, easy-to-read content that fits.
Response Generation: Based on the user's intent and the chatbot's knowledge base, a suitable response is generated.
What AI seems to favour:
Wikipedia and G2: These show up consistently as top sources, so it helps to have a presence there where possible, and to have a customer review strategy in place.
Listicles and scannable content that’s cleanly structured and easy to summarise.
Meta descriptions that answer the query. Unlike traditional SEO where you might treat meta descriptions like ad copy, AI prefers direct, explanatory text that tells it what’s on the page.
Brief, LLM-readable summaries. Adding a short paragraph (a couple of tweets' worth) at the bottom of your page that explains what the page is about (including the answer to the query).
Long, descriptive URLs. Avoid random codes or numbers. AI models pick up on clean, meaningful slugs.
Fresh content. AI tools seem to pick up new or updated content within a few days (much quicker than Google indexing). Update your pages regularly and keep timestamps current.
A few other things worth noting:
Low-traffic pages can still rank. AI doesn’t necessarily prioritise what’s popular. If the structure is clean, current and the content is useful, it will still surface.
Too many backlinks might be a bad thing. Oddly, content with over 10 backlinks was shown to perform worse in AI rankings in one study. Worth keeping an eye on.
In the context of B2B…
What’s the point of a website, really? It’s still there to help us win leads, support the sales process, show credibility, and tell our story. Even though AI snippets and chats may reduce clicks, traffic has never been the whole point. If our prospects are using AI tools to find information, we need to make sure our content is set up to be found and understood in those environments, even if we can’t always trace the journey back to a lead.
Come at us with thoughts - hello@cusp.co.nz