How to be found on AI: the new SEO world

That is the new buyer journey, and it already converts better than the old one. AI search visitors sign up at 11 times the rate of organic search traffic, 1.66% versus 0.15%, according to a Microsoft Clarity study of 1,200+ sites in late 2025. The highest-converting acquisition channel of the year, and around 78% of marketers cannot measure whether their brand appears in AI answers at all.

The reason most teams have no data: GA4 and Search Console were built for clicks. About 93% of Google AI Mode chats produce zero clicks. The buyer got their answer, decided what to buy, and never visited a site that could measure it.

Google ranking and AI citation are not the same thing

When someone asks ChatGPT for a recommendation, ChatGPT does not look at your Google rank. It pulls from Bing. Perplexity built its own index of around 200 billion URLs. Claude pulls from Brave Search. Google's AI Mode pulls from Google. Different indices, different signals, different winners.

Whitehat SEO ran 100,000 prompts and found only 11% citation overlap between ChatGPT and Perplexity. About 68% of brands appear on a single platform. So a founder who spends six months "winning SEO" can still be invisible on the platform their buyer actually opens.

One spot-check tells you nothing

If you have already opened ChatGPT, typed "best [your category]" and felt reassured: that test is broken. SparkToro and Gumshoe.ai gave 600 volunteers nearly 3,000 queries to run and found less than a 1% chance that two AI responses recommended the same brand list. Same prompt. Same model. Different answer. One check is noise.

What real measurement looks like:

• Citation rate across hundreds of prompt variations, not one
• Share of voice against your top three competitors
• Sentiment, meaning not just whether you are mentioned but how you are described
• Source mapping: which third-party articles, podcasts and directories drive your citations
• Cross-platform coverage, so you know where you are absent

What this changes about the content you write

Two findings worth knowing. First, about 40% of sources cited in Google's AI Overviews rank outside Google's standard top 10. AI authority correlates with Google authority but is not identical. Second, Kevin Indig's analysis of 3 million ChatGPT responses found 44% of citations come from the first 30% of an article. If your insight sits below a thousand words of preamble, AI does not see it.

Generic schema does not save you either. Growth Marshal looked at 730 AI citations and found pages with boilerplate schema get cited less often (41.6%) than pages with no schema at all (59.8%). Only fully detailed schema crosses 61.7%. Half-effort on schema is worse than skipping it.

What to do this quarter

1. Run your own baseline. Pick five real buyer questions in your category. Run them across ChatGPT, Claude, Perplexity and Google AI Mode. Log who is cited. Repeat the same five queries the next day. The variance you see is the lesson.
2. Front-load the answer. Put the conclusion in the first paragraph of every page, not the last. AI engines pull from the top of the article. Old SEO conventions were wrong about this.
3. Get cited externally. Three podcasts this quarter, one industry roundup, one journalist quote. Around 85% of AI citations come from someone else's site, so you need someone else to publish you.
4. Track citation rate weekly. A few hundred prompts a month costs around $3 in API credits. Build the loop yourself, or pick a purpose-built tool when you need cross-platform analysis at scale. Weekly cadence beats monthly.